CN113251670A

CN113251670A - Hot blast stove control and training method, device, equipment, hot blast stove system and medium

Info

Publication number: CN113251670A
Application number: CN202110587915.7A
Authority: CN
Inventors: 冯恺睿; 张永强; 王崇鹏; 魏华剑; 张煊; 钟智敏
Original assignee: Hkust Intelligent Iot Technology Co ltd; Jiangsu Yonglian Huike Iot Technology Co ltd
Current assignee: Hkust Intelligent Iot Technology Co ltd; Jiangsu Yonglian Huike Iot Technology Co ltd; CSG Smart Science and Technology Co Ltd
Priority date: 2021-05-28
Filing date: 2021-05-28
Publication date: 2021-08-13
Anticipated expiration: 2041-05-28
Also published as: CN113251670B

Abstract

The application provides a hot blast stove control method, a device, electronic equipment, a hot blast stove system and a medium, wherein each first stove condition state of the hot blast stove at a time period is obtained; the first furnace condition state includes: empty coal data and furnace condition temperature data; predicting the predicted air supply temperature and the predicted total gas consumption at the second future moment of the first moment according to the first furnace condition states of the time intervals; and determining a target air coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption. According to the embodiment of the application, the intelligent hot blast stove can predict accurate data at the future time, a target empty coal control strategy for optimizing air supply temperature and total gas consumption is correspondingly provided, more flexible and accurate hot blast stove adjustment is realized, and the energy consumption is effectively saved while the air supply temperature is maintained.

Description

Hot blast stove control and training method, device, equipment, hot blast stove system and medium

Technical Field

The present application relates to the technical field of control systems for ferrous metallurgy material manufacturing, and in particular, to a method and an apparatus for controlling a hot blast stove, an electronic device, a hot blast stove system, and a medium.

Background

The hot blast stove is an important thermal equipment for providing hot air for blast furnace ironmaking.

The hot blast stove is periodically and circularly operated. In each working cycle, the working cycle is divided into a combustion period and an air supply period. During the combustion period, gas and air are fed into the burner of the hot blast stove in certain proportion, the gas is combusted to heat the heat accumulator in the hot blast stove, the combustion product (i.e. flue gas) is discharged from the chimney through the flue, and the inlet of the burner is closed after the hot blast stove is heated to the required temperature to convert the combustion product into air supply. During the air supply period, cold air sent from the cold air pipeline enters the hot blast stove, the cold air is heated into hot air when passing through the heat accumulator, and the hot air is sent into the blast furnace through the hot air outlet and the pipeline. When the heat accumulator can not heat the cold air to the temperature required by the blast furnace, the air supply is stopped, and the combustion inlet is opened to shift to the combustion period of the next working cycle.

In order to optimize the gas consumption, the automatic control system adopts a plurality of preset fixed air-coal ratios to control the operation of the hot blast stove so as to switch among the air-coal ratios according to the actual stove condition. However, the preset fixed air-coal ratios are not actually the optimal air-coal ratios, and the optimization of the gas consumption still has a huge optimization space. In addition, 3-4 hot blast stoves are adopted to alternately work in the combustion and air supply period to supply air to the blast furnace. Therefore, when the hot blast stoves are alternately replaced, the fixed air-coal ratio does not meet the requirement of actual gas consumption, so that the condition that a plurality of stoves need gas simultaneously but the gas quantity in a gas main pipe is insufficient can be caused, and the temperature in the stoves fluctuates.

In the current automatic control mode of the hot blast stove, only consideration is given to the stability of the temperature in the stove, and how to optimize the gas consumption is not considered so as to achieve the purposes of energy conservation and emission reduction.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present application provides a hot blast stove control method, apparatus, electronic device, hot blast stove system, and medium, which solve the problems of the prior art.

To achieve the above and other related objects, a first aspect of the present application provides obtaining first furnace condition states of a hot blast stove over a period of time; the first furnace condition state includes: empty coal data and furnace condition temperature data; the empty coal data comprises at least one of: control parameters of a gas regulating valve and an air regulating valve of the hot blast stove; gas and air flow; predicting the predicted air supply temperature and the predicted total gas consumption at a second time in the future of the first time according to each first furnace condition state of the time period, wherein the predicting comprises the following steps: predicting a second furnace condition state and total gas consumption at a second moment according to each first furnace condition state of the time interval; and predicting according to the second furnace condition state to obtain the predicted air supply temperature; the second furnace condition state and the total gas consumption at the second moment are predicted according to the first furnace condition state, and the prediction is completed through a furnace condition prediction model; the predicted air supply temperature is obtained according to the second furnace condition state prediction and is completed through an air supply temperature prediction model; the total gas consumption is predicted according to the second furnace condition state, and the total gas consumption is predicted through a total gas consumption prediction model; determining a target empty coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption, wherein the control strategy comprises the following steps: calculating corresponding comprehensive evaluation results based on various situations of the predicted air supply temperature and the predicted total gas consumption under the second furnace condition state which is changed into the second time by adopting each alternative air-coal control strategy under the first furnace condition state; wherein the comprehensive evaluation result comprises: the comprehensive result between the gas consumption evaluation generated by the air coal control strategy at the first moment and the expected value of the prediction evaluation based on the predicted air supply temperature and the predicted total gas consumption at the second moment is obtained; and determining a target empty coal control strategy according to the optimal one in the comprehensive evaluation results.

In some embodiments of the first aspect, the furnace condition temperature data comprises at least one of: the dome temperature of the hot blast stove; flue temperature.

In some embodiments of the first aspect, the first furnace condition state further comprises: gas pipeline pressure.

In some embodiments of the first aspect, the second furnace condition status comprises predicted furnace condition temperature data; the predicted furnace condition temperature data includes at least one of: predicting the vault temperature; and predicting the flue temperature.

In some embodiments of the first aspect, there is at least one of:

case 1: the furnace condition prediction model is realized based on a fusion model of a plurality of sub-models;

case 2: the air supply temperature prediction model is realized based on a fusion model of a plurality of sub models;

case 3: the total gas consumption prediction model is realized based on a fusion model of a plurality of sub-models.

In some embodiments of the first aspect, the plurality of submodels of the fusion model in at least one of case 1 to case 3 are respectively assigned with prediction weight parameters representing magnitudes of contributions of the submodels to the prediction result of the fusion model; the prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments.

In some embodiments of the first aspect, the stove control method comprises: predicting the predicted data of the future time according to the actual data of the time interval, the combination of the actual data and the predicted data of the time interval or the predicted data of the time interval so as to obtain the data of each time in one working cycle of the hot blast stove; the actual data and the predicted data correspond to at least one of a furnace condition state, air supply temperature data and total gas consumption.

In some embodiments of the first aspect, the predictive assessment based on the predicted supply air temperature and the predicted total gas consumption comprises: a difference between a first representation relating to an error between the predicted supply air temperature and the requested supply air temperature and a second representation relating to a total gas consumption.

In some embodiments of the first aspect, the pre-association between each condition state and the corresponding predictive assessment at each time in a time period is stored as pre-stored data; determining a target empty coal control strategy according to the optimal one of the comprehensive evaluation results, wherein the method comprises the following steps: according to the furnace condition state at the first moment, inquiring the predicted associated prediction evaluation of various second furnace condition states in the prestored data; calculating expected values according to the predicted associated prediction evaluation of the various second furnace condition states, and calculating various comprehensive evaluation results by combining various alternative air coal control strategies at the first moment; and calculating to obtain the target empty coal control strategy based on the optimal comprehensive evaluation result.

In some embodiments of the first aspect, the calculation of the expected value comprises a random result caused by a random fluctuation parameter.

To achieve the above and other related objects, a second aspect of the present application provides a training method for a predictive model of a hot blast stove, comprising: obtaining a training data set based on the furnace condition state of the hot blast furnace in the historical time period; each piece of training data in the training data set comprises input data and corresponding reference data; at least one predictive model is trained based on the training data set for enabling the at least one predictive model to predict at least a predicted supply air temperature and a predicted total gas consumption at a future time based on a furnace condition state over a period of time.

In some embodiments of the second aspect, the predictive models are multiple, including:

a furnace condition prediction model for predicting a second furnace condition state at a future time from a first furnace condition state for a period of time;

the air supply temperature prediction model is used for predicting and obtaining the predicted air supply temperature according to the second furnace condition state at the same moment;

a total gas consumption prediction model for predicting a predicted total gas consumption at a future time according to a first furnace condition state of a time period;

wherein at least one of the furnace condition prediction model, the air supply temperature prediction model and the total gas consumption prediction model is trained; the input data for each piece of training data of the furnace condition prediction model includes: a time period of furnace condition status, the reference data comprising: the furnace condition status at a time after the time period; the input data for each piece of training data of the supply air temperature prediction model includes: the reference data of the furnace condition state at one moment comprises: the air supply temperature at the same time; the input data for each piece of training data of the total gas consumption prediction model includes: a time period of furnace condition status, the reference data comprising: total gas consumption at a time after the period.

In some embodiments of the second aspect, the training method is at least one of:

In some embodiments of the second aspect, the plurality of submodels of the fusion model in at least one of case 1 to case 3 are respectively assigned with prediction weight parameters representing magnitudes of contributions of the submodels to the prediction result of the fusion model; the prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments.

To achieve the above and other related objects, a third aspect of the present application provides a hot blast stove control device comprising: the acquisition module is used for acquiring a first furnace condition state of the hot blast stove in a period of time; the first furnace condition state includes: empty coal data and furnace condition temperature data; the empty coal data comprises at least one of: control parameters of a gas regulating valve and an air regulating valve of the hot blast stove; gas and air flow; the prediction module is used for predicting the predicted air supply temperature and the predicted total gas consumption at the future second moment at the first moment according to the first furnace condition state, and comprises: predicting a second furnace condition state and total gas consumption at a second moment according to each first furnace condition state of the time interval; and predicting according to the second furnace condition state to obtain the predicted air supply temperature; the second furnace condition state and the total gas consumption at the second moment are predicted according to the first furnace condition state, and the prediction is completed through a furnace condition prediction model; the predicted air supply temperature is obtained according to the second furnace condition state prediction and is completed through an air supply temperature prediction model; the total gas consumption is predicted according to the second furnace condition state, and the total gas consumption is predicted through a total gas consumption prediction model; the strategy determination module is used for determining a target air-coal control strategy of the hot blast stove at the first moment according to the optimization trends of the predicted air supply temperature and the predicted total gas consumption, and comprises the following steps: calculating corresponding comprehensive evaluation results based on various situations of the predicted air supply temperature and the predicted total gas consumption under the second furnace condition state which is changed into the second time by adopting each alternative air-coal control strategy under the first furnace condition state; wherein the comprehensive evaluation result comprises: the comprehensive result between the gas consumption evaluation generated by the air coal control strategy at the first moment and the expected value of the prediction evaluation based on the predicted air supply temperature and the predicted total gas consumption at the second moment is obtained; and determining a target empty coal control strategy according to the optimal one in the comprehensive evaluation results.

To achieve the above and other related objects, a fourth aspect of the present application provides a training device for a predictive model of a hot blast stove, comprising: the training data acquisition module is used for acquiring a training data set based on the furnace condition state of the hot blast furnace in the historical time period; each piece of training data in the training data set comprises input data and corresponding reference data; and the training module is used for training at least one prediction model based on the training data set so as to enable the at least one prediction model to at least predict the predicted air supply temperature and the predicted total gas consumption at the future moment according to the furnace condition state of a period of time.

To achieve the above and other related objects, a fifth aspect of the present application provides an electronic device comprising a processor and a memory; the memory to store computer program instructions; the processor executing computer program instructions that invoke the memory performs the stove control method of any one of the first aspects or the training method of any one of the second aspects.

To achieve the above and other related objects, a sixth aspect of the present application is a computer readable storage medium having stored therein program instructions, which, when executed by a processor, implement the hot blast stove control method of any one of the first aspects or the training method of any one of the second aspects.

To achieve the above and other related objects, a stove system according to a seventh aspect of the present application includes: at least one hot blast stove with a gas regulating valve and an air regulating valve; the hot blast stove control system is in communication connection with and controls the gas regulating valve and the air regulating valve and collects the state of the stove condition from the hot blast stove; an electronic device comprising a processor, a memory, and a communication interface; the memory to store computer program instructions; the processor executing computer program instructions invoked in the memory to perform the hot blast stove control method of any one of the first aspect; the communication interface is in communication connection with or integrated with the hot blast stove control system and is used for acquiring the acquired furnace condition state and outputting the target air coal control strategy to the hot blast stove control system.

In some embodiments of the seventh aspect, the stove control system has an overall control loop subject to stove temperature control, the overall control loop comprising: a sub-control loop taking the gas flow as a control object; and a sub-control circuit for controlling the air flow rate.

In summary, the present application provides a method, an apparatus, an electronic device, a hot blast stove system and a medium for controlling a hot blast stove by obtaining each first furnace condition state of the hot blast stove at a time period; the first furnace condition state includes: empty coal data and furnace condition temperature data; predicting the predicted air supply temperature and the predicted total gas consumption at the second future moment of the first moment according to the first furnace condition states of the time intervals; and determining a target air coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption. According to the embodiment of the application, the intelligent hot blast stove can predict accurate data at the future time, a target empty coal control strategy for optimizing air supply temperature and total gas consumption is correspondingly provided, more flexible and accurate hot blast stove adjustment is realized, and the energy consumption is effectively saved while the air supply temperature is maintained.

Drawings

Fig. 1 shows a schematic flow diagram of a method for controlling a hot blast stove in an embodiment of the present application.

Fig. 2 shows a schematic flow chart of the step S102 in fig. 1 for predicting the supply air temperature and the total gas consumption in one embodiment.

Fig. 3 shows a schematic flow chart of obtaining a target empty coal control strategy in an embodiment of the present application.

Fig. 4 shows a schematic flow chart of calculating a target empty coal control strategy by querying prestored data in the embodiment of the present application.

Fig. 5 shows a schematic flow chart of a training method for a predictive model of a hot blast stove in an embodiment of the present application.

Fig. 6 shows a block schematic diagram of a control device of a hot blast stove in an embodiment of the present application.

Fig. 7 shows a block schematic diagram of a training device for a predictive model of a hot blast stove in an embodiment of the present application.

Fig. 8 is a schematic structural diagram of an electronic device in an embodiment of the present application.

Fig. 9 shows a schematic structural diagram of a hot blast stove system provided in an embodiment of the present application.

Fig. 10 shows a schematic control signal flow diagram of a control system of a hot blast stove in the embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application is provided by way of specific examples, and other advantages and effects of the present application will be readily apparent to those skilled in the art from the disclosure herein. The present application is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present application. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

Embodiments of the present application will be described in detail below with reference to the accompanying drawings so that those skilled in the art to which the present application pertains can easily carry out the present application. The present application may be embodied in many different forms and is not limited to the embodiments described herein.

In order to clearly explain the present application, components that are not related to the description are omitted, and the same reference numerals are given to the same or similar components throughout the specification.

Throughout the specification, when a device is referred to as being "connected" to another device, this includes not only the case of being "directly connected" but also the case of being "indirectly connected" with another element interposed therebetween. In addition, when a device "includes" a certain component, unless otherwise stated, the device does not exclude other components, but may include other components.

When a device is said to be "on" another device, this may be directly on the other device, but may also be accompanied by other devices in between. When a device is said to be "directly on" another device, there are no other devices in between.

Although the terms first, second, etc. may be used herein to describe various elements in some instances, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, the first interface and the second interface, etc. are described. Also, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms "comprises," "comprising," "includes" and/or "including," when used in this specification, specify the presence of stated features, steps, operations, elements, components, items, species, and/or groups, but do not preclude the presence, or addition of one or more other features, steps, operations, elements, components, species, and/or groups thereof. The terms "or" and/or "as used herein are to be construed as inclusive or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: a; b; c; a and B; a and C; b and C; A. b and C ". An exception to this definition will occur only when a combination of elements, functions, steps or operations are inherently mutually exclusive in some way.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the singular forms "a", "an" and "the" include plural forms as long as the words do not expressly indicate a contrary meaning. The term "comprises/comprising" when used in this specification is taken to specify the presence of stated features, regions, integers, steps, operations, elements, and/or components, but does not exclude the presence or addition of other features, regions, integers, steps, operations, elements, and/or components.

Terms representing relative spatial terms such as "lower", "upper", and the like may be used to more readily describe one element's relationship to another element as illustrated in the figures. Such terms are intended to include not only the meanings indicated in the drawings, but also other meanings or operations of the device in use. For example, if the device in the figures is turned over, elements described as "below" other elements would then be oriented "above" the other elements. Thus, the exemplary terms "under" and "beneath" all include above and below. The device may be rotated 90 or other angles and the terminology representing relative space is also to be interpreted accordingly.

Although not defined differently, including technical and scientific terms used herein, all terms have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. Terms defined in commonly used dictionaries are to be additionally interpreted as having meanings consistent with those of related art documents and the contents of the present prompts, and must not be excessively interpreted as having ideal or very formulaic meanings unless defined.

The hot blast stove belongs to a large energy consumption family in the iron making process, and the consumed heat accounts for more than 20 percent of the whole iron making process. Because the heat source of the hot blast stove is coal gas combustion, how to intelligently adjust the proper air-coal ratio to achieve the situation of required air supply temperature and greatly save energy consumption to reduce carbon emission, the technical problem to be solved is urgently needed in the industry.

In view of this, the embodiment of the present application provides a hot blast stove control method, which outputs an optimized air-coal control strategy to solve the problems in the prior art, and also achieves the goals of energy saving and emission reduction under the condition of maintaining the required air supply temperature standard.

Specifically, the hot blast stove control method provided in the embodiment of the present application uses the predicted air supply temperature at the second time, which is accurately predicted by the prediction model based on machine learning, as a reference, and further generates a corresponding dynamic air-coal control strategy, thereby getting rid of the manner of fixing the air-coal ratio in the prior art.

Fig. 1 shows a schematic flow chart of a hot blast stove control method in the embodiment of the present application.

In this embodiment, the hot blast stove control method comprises:

step S101: a first condition state of the hot blast stove in a period of time is obtained.

The first condition state selects hot blast stove data relating to the supply air temperature for predicting the supply air temperature in a subsequent step. In some embodiments, the first fire condition state comprises: air coal data and furnace condition temperature data.

In some more specific embodiments, the empty coal data includes at least one of: control parameters of a gas regulating valve and an air regulating valve of the hot blast stove; gas and air flow. In a specific implementation example, the control parameters of the gas regulating valve and the air regulating valve of the hot blast stove may include valve opening degrees and the like, and the opening degrees of the gas regulating valve and the air regulating valve may affect the flow rates of gas and air in the pipeline, thereby affecting the temperature of the heat accumulator and then affecting the air supply temperature of the hot blast stove; therefore, the actually collected gas and air flow are influenced by the control parameters (such as the opening degree) of the opening degrees of the gas regulating valve and the air regulating valve, and also influence the air supply temperature of the hot blast stove.

The empty coal data can comprise control parameters of the gas regulating valve and the air regulating valve and flow of gas and air, so that the machine learning model can more accurately learn the actual relevance between the data. In a specific implementation example, the opening degrees of the gas regulating valve and the air regulating valve can be obtained from the valves, and the flow rates of the gas and the air can be obtained through flow sensors.

In some embodiments, the furnace condition temperature data includes at least one of: the dome temperature of the hot blast stove; flue temperature. The air supply temperature of the hot blast stove is related to the vault temperature and the flue temperature, and the prediction of the air supply temperature by taking the vault temperature and the flue temperature of the hot blast stove as input can be more accurate. In a specific implementation example, the dome temperature and flue temperature of the stove may be obtained by temperature sensors at the relevant locations.

In some embodiments, the first furnace condition state may further include: gas pipeline pressure. Specifically, in the process of switching the combustion period and the air supply period of the hot blast stove in the working period, the pressure of a gas pipeline changes; in other cases, the gas pipeline pressure may fluctuate, which may have an effect on the temperature inside the hot blast stove, so that the gas pipeline pressure may also be used as a reference data in order to obtain a more accurate prediction result. In a specific implementation example, the gas pipeline pressure may be obtained by a pressure sensor at the relevant location.

In some embodiments, the period may include a plurality of time instants. For example, with one time per minute, the time period may be several minutes, tens of minutes, or tens of minutes. For example, the actual opening degrees of the gas regulating valve and the air regulating valve of the hot blast stove in 5 minutes, the gas and air flow rates, the vault temperature and the flue temperature of the hot blast stove, the gas pipeline pressure and the like are collected to be used for predicting the air supply temperature at the second moment (for example, the 6 th minute after 5 minutes).

Step S102: and predicting the predicted air supply temperature and the predicted total gas consumption at the future second moment of the first moment according to the first furnace condition state.

In some embodiments, the predicted supply air temperature data and the predicted total gas consumption at the second time can be further predicted by predicting a second furnace condition state at the second time according to the first furnace condition state. The total predicted gas consumption refers to the accumulated gas consumption from a certain moment to a second moment.

Specifically, as shown in fig. 2, a schematic flow chart of the step S102 in fig. 1 for predicting the air supply temperature and the total gas consumption in an embodiment is shown.

The step S102 may further include:

step S201: and predicting a second furnace condition state at a second moment and predicting the total gas consumption according to the first furnace condition state.

Step S202: and predicting the air supply temperature according to the second furnace condition state.

In some embodiments, the second furnace condition status may include predicted furnace condition temperature data; the predicted furnace condition temperature data includes at least one of: predicting the vault temperature; and predicting the flue temperature. It should be noted that the second furnace condition state may include only the predicted furnace condition temperature data, that is, when the furnace condition temperature is predicted, only at least one of the predicted dome temperature and the predicted flue temperature is predicted, and the supply air temperature may be predicted only based on the predicted furnace condition temperature data, thereby simplifying the calculation.

In some embodiments, predicting a second furnace condition state at a second time according to the first furnace condition state in step S201 may be performed by a furnace condition prediction model, and predicting a predicted total gas consumption at the second time according to the first furnace condition state in step S201 may be performed by a total gas consumption prediction model; the prediction of the supply air temperature according to the second furnace condition state prediction in step S202 may be implemented by a supply air temperature prediction model. The furnace condition prediction model, the total gas consumption prediction model and the air supply temperature prediction model can be realized through a machine learning model. Of course, in other examples, the furnace condition prediction model, the total gas consumption prediction model and the air supply temperature prediction model do not need to be implemented as independent models, and two prediction models can be fused or combined into a whole.

In some embodiments, part or all of the furnace condition prediction model, the total gas consumption prediction model and the air supply temperature prediction model can be implemented by a fusion model formed by fusing a plurality of sub-models.

Optionally, the types of the plurality of submodels may include one or more of a decision tree, unsupervised, time series, neural network, support vector machine, and the like. In an example of a specific implementation, the plurality of sub-models comprises some or all of the following models: random Sample Consensus (RANSAC) model, Linear Regression (Linear Regression) model, Ridge Regression (Ridgeregression) model, Bayesian Ridge Regression (Bayesian Ridge Regression) model, LASSO Regression model, Elastic network Regression (Elastic Net) model, eXtreme Random Tree (eXtreme Random Tree) model, Gradient Boosting (Gradient Boosting) model, tbCaoost Regression model, eXtreme Gradient Boosting (eXemtree Gradient Boosting) model, Orthogonal matching pursuit (Orthogonal matching pursuit) model, a Random Forest (Random Forest) model, a Light Gradient Boosting Machine (LightGBM) model of a lightweight frame, an Adaboost regression model, a decision tree model, a K-nearest neighbor regression model, a thieson regression (TheilSen regression) model, a minimum angle regression model, a Lasso minimum angle regression model, a support vector Machine model, a Passive attack (Passive attack regression) model, a Huber regression model, and the like. The above listed models are only examples, and may be added or deleted according to the requirement in practical application, and are not limited thereto.

Wherein, the random sampling consistency model can estimate the parameters of the mathematical model from a group of observation data sets containing 'local points' in an iterative mode; the linear regression model is a statistical analysis model which determines the interdependent quantitative relationship between two or more variables by using regression analysis in mathematical statistics; the ridge regression is proposed to solve the problem of unstable numerical solution in the ordinary linear regression, which adds a penalty term to the original objective function of the model in order to limit the numerical size of the model parameters, and the process is called Regularization (Regularization); the Bayesian ridge regression is different from the ordinary ridge regression in that the Bayesian gradual updating prior strategy is adopted; LASSO regression is a compression estimation, a more refined model is obtained by constructing a penalty function, so that some regression coefficients are compressed, namely the sum of absolute values of the forcing coefficients is smaller than a certain fixed value; setting some regression coefficients to be zero; the idea of elastic network regression is to combine an L1 norm regularization term and an L2 norm regularization term on the basis of least squares; compared with a Random Forest (RF), the extreme random tree is characterized in that each tree adopts all training samples, namely the sample set of each tree is the same, and the extreme random tree directly selects the bifurcation characteristics at random; in each iteration, firstly calculating the negative gradient of the current model on all samples, then training a new weak classifier by taking the value as a target to perform fitting and calculating the weight of the weak classifier, and finally updating the model; in a related way, the general idea of the XGBoost is to generate a series of CART regression trees in series, each regression tree is to fit the residual between the previous tree and the target value, and the process is repeated until the preset number of trees is reached or the convergence is reached; the Catboost is a GBDT framework which is realized by taking a symmetric decision tree (Obevious trees) as a base learner, has few parameters, supports the class-type variables and has high accuracy, and mainly solves the problem that the pain point is to efficiently and reasonably process the class-type characteristics; the orthogonal matching pursuit selects the columns of the dictionary in a greedy iteration method, so that the selected columns are maximally related to the current redundant vector in the process of each iteration, a related part is subtracted from an original signal vector, and iteration is repeated until the iteration number reaches the required sparsity K; a random forest is a classifier comprising a plurality of decision trees and the class of its output is dependent on the mode of the class output by the individual trees; the LightGBM is used for solving the problems of the gradient lifting decision tree in mass data and enabling the gradient lifting decision tree to be better and faster used for industrial practice; adaboost is an iterative algorithm whose core idea is to train different classifiers (weak classifiers) for the same training set and then to assemble these weak classifiers to form a stronger final classifier (strong classifier; decision tree model is a simple easy-to-use nonparametric classifier; K-neighbor idea is to give a training data set and to find K instances (i.e. K neighbors mentioned above) in the training data set which are nearest to the instances, most of the K instances belonging to a certain class, then to classify the input instances into this class; Talsen regression is a parametric median estimator which is suitable for generalization median and to estimate multidimensional data, thus having strong robustness to the singular points of multiple dimensions; minimal solution regression is an efficient solution of LASSO regression, accelerating the parameter calculation process; support vector machine is to perform binary classification on data in a class-supervised learning (latent learning) manner The generalized linear classifier of (1); the passive attack model is a classifier suitable for large-scale learning.

In some embodiments, the sub-models may be fused into a fusion model by a model fusion method. Optionally, the model fusion method may adopt average fusion, weighted fusion, voting fusion, or the like, or may also adopt more complicated Blending or Stacking, where the Blending and Stacking may set each base learner (i.e., the sub-model referred to in this specification) in a multi-layer architecture.

Taking Stacking as an example, the process is to create a plurality of sub-models of the first layer, which may be homogeneous or heterogeneous. Randomly equally dividing a training data set into K parts (for example, 5 parts) which are respectively Fold 1-Fold 5, training each first layer of submodel Mi on { Fold2, Fold3, Fold4 and Fold5} to obtain a learner Mi _1, and predicting Fold1 to obtain a new feature value NewFeaturei _1 on Fold 1; mi _1 is trained through { Fold1, Fold3, Fold4 and Fold5} to obtain a learner Mi _2, and then the Fold2 is predicted to obtain a new Featurei _2 of the new characteristic on the Fold 2; by analogy, NewFeaturei _ 3-NewFeaturei _5 can be obtained in sequence, and finally NewFeaturei _ 1-NewFeaturei _5 are combined together to obtain the NewFeaturei corresponding to the new characteristic Mi. The N sub-models generate N newfeatures, thereby forming a new training data set newtrain = { NewFeature1, NewFeature 2.., NewFeature en }, as a training data set for the sub-model of the next layer.

Optionally, a training data set is generated according to the historical period data of the hot blast stove, and the training data set participates in training of each sub-model in the fusion model to obtain a trained model. Each piece of training data in the training data set includes input data and corresponding reference data. Taking training of a furnace condition prediction model, a total gas consumption prediction model and an air supply temperature prediction model as an example, input data for each piece of training data of the furnace condition prediction model includes: a time period of furnace condition status, the reference data comprising: the furnace condition status at a time after the time period; the input data for each piece of training data of the supply air temperature prediction model includes: the reference data of the furnace condition state at one moment comprises: the air supply temperature at the same time; the input data for each piece of training data of the total gas consumption prediction model includes: a time period of furnace condition status, the reference data comprising: total gas consumption at a time after the period.

For example, a piece of training data includes, as input data, the opening degrees of the gas regulating valve and the air regulating valve of the hot blast stove from the x minute to the x +5 minute on a certain day, the gas and air flow rates, the dome temperature and the flue temperature, the gas pipe pressure, and the like, and the dome temperature and the flue temperature corresponding to the x +6 minute, and the like. The furnace condition prediction model is trained by the training data set, so that each sub-model is trained to accurately learn the relationship between the input data and the output data, and the second furnace condition state at the second time can be predicted according to the first furnace condition state in the past time period as in step S201.

In some embodiments, the model evaluation index can be used to evaluate the merits of the sub-models, and the better part can be selected for use. The model evaluation index may include: for example, regression evaluation indexes for predicted results include some or all of the following: mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), coefficient of determination R-Square, mean absolute percentage error (MSPE), and the like. Further, an index relating to the model operation time may be included.

Of course, in another example, all of the above listed submodels (and possibly more) may be used for fusion for better stability without being limited to perform well in some known scenarios and perform abruptly in other scenarios.

In some embodiments, in order to improve the overall performance of the fusion model (for example, the furnace condition prediction model, the total gas consumption prediction model, the air supply temperature prediction model, and other prediction models) while maintaining the stability of the fusion model in which multiple submodels are fused, prediction weight parameters representing the contribution of the submodels to the prediction result of the fusion model may be assigned to each submodel.

The prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments. Specifically, the prediction model may automatically update the weight parameters of each sub-model according to the following formula.

Equation (1) below illustrates an example of the calculation of the output error for each submodel, which illustrates the manner in which the standard deviation of the different prediction weights for the j to n minutes is found:

（1）

wherein σ_inGlobal error for the prediction result n minutes before the ith model, w_ijPredicted weight for ith model at jth minute, P_ijPrediction of jth minute for ith model, R_ijActual values for the ith model versus the jth minute.

It should be noted that, in other examples, other types of error calculation may be used instead, for example, the above-mentioned Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), coefficient of determination R-Square, mean absolute percentage error (MSPE), etc., and the formula (1) is not limited thereto.

Further, as shown in formula (2), the prediction weight parameter of the ith model at the n +1 th minute is determined according to the overall prediction error of the model at the previous n times and the corresponding prediction weight parameter in formula (1):

（2）

exp is an exponential function with e as a base, lambda is a model adjustable parameter, and the larger lambda is, the more error is required for sub-model selection.

In some embodiments, the weighted sum of the prediction outputs of the sub-models and their respective prediction weights may be used as the prediction result of the fusion model, and by assigning the prediction weight parameters updated by the overall prediction error (e.g., the above-mentioned error σ) to each sub-model over time, the sub-model with smaller error can take a higher weight in the prediction result of the whole fusion model, so as to improve the prediction accuracy, i.e., the performance of the prediction model.

Under the condition that a prediction model (such as a furnace condition prediction model, a total gas consumption prediction model, an air supply temperature prediction model and other prediction models) is realized by adopting the multi-submodel fusion model, the stability and the prediction accuracy are better, so that the prediction accuracy can be ensured under the condition of less input data. For example, a first state of furnace conditions for a period of time that is part of a time in a cycle in which the stove is used, at least one of a second state of furnace conditions, a predicted supply air temperature and a predicted total gas consumption at the remaining times in the overall cycle may be predicted.

In some embodiments, the predicted data for future times may be predicted from individual actual data over a period of time, a combination of actual and predicted data over a period of time, or individual predicted furnace condition data over a period of time to obtain data for each time in a cycle of operation of the stove. The actual data can be at least one of actual furnace condition state, actual air supply temperature data and actual total gas consumption, and the predicted data can be at least one of predicted furnace condition state, predicted air supply temperature data and predicted total gas consumption.

In some embodiments, for example, the furnace condition prediction model predicts the second furnace condition state, the second furnace condition state may be predicted according to the actual furnace condition state, the combination of the actual furnace condition state and the predicted furnace condition state, and the predicted furnace condition states of the corresponding time period after each step by stepping according to a time sequence in a fixed or unfixed length of time period (for example, fixed 5 minutes and the like) by a predetermined step width (for example, 1 minute).

By way of further example, it is assumed that all actual furnace conditions (T1, P1.), (T2, P2.), (T5, P5..) are present, depending on the first furnace conditions of the first 5 minutes of the furnace cycle, T and P may be, for example, the dome temperature and the flue temperature, respectively. Predicting a predicted furnace condition state at the 6 th minute from an actual furnace condition state at the first 5 minutes (T6, p 6.); and forming a group of input according to the actual furnace condition data (T2, P2.), (T5, P5..) of the 2 th to 5 th minutes and the predicted furnace condition state (T6, P6.) of the 6 th minute, and predicting the predicted furnace condition state (T7, P7.) of the 7 th minute. By analogy, starting with (T6, p 6.), (T7, p 7.), (T10, p 10.) predictions (T11, p 11.), the input data are all predicted furnace condition states until a second furnace condition state for the full duty cycle is predicted. Through practical verification, the difference between the predicted second furnace condition state and the actual furnace condition state at the same moment is very small, and the prediction is very accurate.

Returning again to fig. 1, step S103 is shown: and determining a target air coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption.

In some embodiments, the air-to-coal control strategy corresponds to an air-to-coal ratio that needs to be achieved, or to only a gas consumption amount that needs to be achieved, and so on. The control quantity of the gas regulating valve and the air regulating valve (only the gas regulating valve or the air regulating valve is possible) can be obtained according to the air-coal control strategy. The air-coal control strategy adopted at the first moment can influence the air supply temperature and the total gas consumption at the second moment. The first time and the second time may be, for example, front and back times, such as the 5 th minute and the 6 th minute, and the empty coal control strategy adopted in the 5 th minute may affect the state of the furnace condition in the 6 th minute, that is, the blast air temperature and the total gas consumption at the 6 th minute. Therefore, a target air-coal control strategy needs to be selected so as to increase the air supply temperature and reduce the total gas consumption as much as possible, so as to meet the optimization trend.

An example of obtaining the target empty coal control strategy is given below by way of example, and it should be noted that the actual implementation manner is not only unique, and may be changed according to actual needs, and the following example is not limited thereto.

As shown in fig. 3, a schematic flow chart of obtaining a target empty coal control strategy in the embodiment of the present application is shown.

The process comprises the following steps:

step S301: and calculating corresponding comprehensive evaluation results based on various situations of the predicted air supply temperature and the predicted total gas consumption under the second furnace condition state which is changed into the second time by adopting each alternative air-coal control strategy under the first furnace condition state.

Wherein the comprehensive evaluation result comprises: the aggregate result of the gas consumption ratings generated by the air and coal control strategy at the first time (i.e., the instantaneous losses generated at the first time using the air and coal control strategy) and the expected values based on the predicted supply air temperature at the second time and the predicted ratings of the predicted total gas consumption (i.e., the future predicted revenue that would be generated at the second time using the air and coal control strategy).

Step S302: and determining a target empty coal control strategy according to the optimal one in each comprehensive evaluation result.

It can be understood that the target empty coal control strategy at the first moment needs to be found, so that the comprehensive evaluation result obtained at the first moment is optimal, namely the profit is maximum.

In some embodiments, the second time may be considered as the identity of the "predicted time", and the first time may be the time prior to the second time as the identity of the "policy enforcement time".

For the second time of the "predicted time", the predicted supply air temperature and the predicted total gas consumption amount in various second furnace condition states possible at the second time can be predicted according to the previous steps S101 and S102. In order to predict the air supply temperature and the total gas consumption, the air-coal control strategy performed at the second moment is not considered due to the identity of the predicted moment; only the instant benefit at the second moment, namely the prediction evaluation obtained based on the predicted air supply temperature and the predicted total gas consumption at the second moment, needs to be considered.

In some embodiments, the predictive assessment based on the predicted supply air temperature and the predicted total gas consumption comprises: a difference between a first representation relating to an error between the predicted supply air temperature and the requested supply air temperature and a second representation relating to a total gas consumption. Wherein the required supply air temperature is a supply air temperature that is predetermined to be actually achieved at the second time.

Exemplarily, the first time is the nth time, the second time is the n +1 th time, and K second furnace condition states are S^n+1_k(ii) a The predicted blowing air temperature of the kth second furnace condition state at the predicted (n + 1) th time is set to M_kThe required supply air temperature at the second moment is Nⁿ⁺¹Predicting total gas consumption as Q_k(ii) a Corresponding predictive evaluation V^n+1_kCan be shown as the following formula (3):

V^n+1_k=（M_k-Nⁿ⁺¹）*a-Q_k*b；（3）

wherein a and b are coefficients such that M_k、Nⁿ⁺¹、Q_kConversion to a consistent magnitude was performed for comparison.

The first time as the "strategy implementation time" needs to be based on the furnace condition state at the first time, and an optimal comprehensive evaluation result in the furnace condition state is obtained to determine the target empty coal control strategy, which can be expressed by the following formula (4):

（4）

wherein the content of the first and second substances,

(i.e., S)ⁿ) The state of the furnace condition at the nth moment,

as a result of the comprehensive evaluation at the nth time,

the control strategy of the empty coal is shown,

a target empty coal control strategy which can be adopted at the nth moment;

the gain (loss reduction also can be obtained) obtained when x is adopted at the nth time, and in a specific example, the gain can be calculated according to x in a preset mode;

the predicted furnace condition state at the n +1 th moment,

to represent

The prediction evaluation of the n +1 th moment under the condition; the open E represents the expectation value for

In other words, the expected value may be

Change into

Probability and corresponding

The probability of state change can be learned in the previous training of the furnace condition state prediction model, and can be obtained in the prediction model by a Monte Carlo random simulation method for example; γ represents a discount coefficient for the future time prediction evaluation, and a smaller γ represents a smaller importance of the future time prediction evaluation. The maximum max value is determined here to be

And (5) obtaining an optimal comprehensive evaluation result to obtain a target empty coal control strategy.

To provide an intuitive understanding of the above principles, a simplified example of a calculation is provided below.

Assume that at the 80 th minute, the furnace condition status S⁸⁰The middle crown temperature is 1290 degrees, the flue temperature is 360 degrees, and there are two alternative air coal control strategies in total, which are simplified to only control the gas flow, and are expressed as 31000 and 30000 cubic meters per hour.

Based on the results of the furnace state prediction, if 31000 is selected, there is a probability of 0.7 changing to the furnace state S at the 81 th minute^81_1Vault temperature of 1291 ℃ and flue temperature of 362 ℃; there is a 0.3 probability of changing to the furnace condition state S at the 81 th minute^81_2Vault temperature 1290 degrees, flue temperature 361 degrees.

If 30000 is selected, there is a probability of 0.7 changing to furnace condition state S at the 81 th minute^81_3Vault temperature 1290 degrees, flue temperature 362 degrees; there is a probability of 0.3 of changing to the furnace condition state S at the 81 st minute^81_4Vault temperature 1289 degrees, flue temperature 361 degrees.

Four furnace condition states can be obtained for the predicted 81 th minute:

S^81_1: vault temperature 1291 degrees flue temperature 362 degrees;

S^81_2: vault temperature 1290 ℃, flue temperature 361 ℃;

S^81_3: vaultThe temperature is 1290 ℃, and the flue temperature is 362 ℃;

S^81_4: the temperature of the vault is 1289 ℃, and the flue temperature is 361 ℃;

the required air supply temperature in the 81 th minute is 1210 ℃, and the predicted air supply temperature and the predicted total gas consumption of 4 furnace condition states are respectively obtained.

Is provided with a pair S^81_1Predicting to obtain the predicted air supply temperature of 1220 ℃, the total gas consumption of 58000 cubic meters, and calculating to obtain V (S) by setting a to be 1000 and b to be 60/110 in the formula (3)^81_1）=（1220-1210）*1000-58000*60/110=10000-31636 = -21636

Similarly, for S^81_2V (S) is obtained by calculation^81_2) Is-22000;

to S^81_3V (S) is obtained by calculation^81_3) Is-21800;

to S^81_4V (S) is obtained by calculation^81_4) Is-22200.

In the case that the selectable number of the air coal control strategies x is not large, the comprehensive evaluation result obtained by traversing each type of x in the 80 th-minute furnace condition state can be directly calculated through the formula (4), and then the optimal one is compared to select the target air coal control strategy.

The results of comprehensive evaluation using different x at 80 th minute were calculated according to the following formula (4), and γ was taken to be 0.3, and C was represented by a value of x/60, since it was found that:

when the gas consumption is 31000, the expected benefits are calculated as follows:

C（S⁸⁰，31000）+γE（V（S⁸¹) I S⁸⁰，31000）=-31000/60-0.3* （0.7*21636 + 0.3*22000）=-7040；

When the gas consumption is 30000, the expected benefits are calculated as follows: -7076.

As can be seen, since the yield is higher when the gas amount is 31000, the target empty coal control strategy at 80 th minute is selected to be 31000 cubic meters per hour.

To improve the calculation efficiency, the pre-association between each furnace condition status and the corresponding prediction evaluation at each time in a time period can be stored in advance as pre-stored data, such as a tableIn the form, the prediction evaluation of each furnace condition state at the 'predicted moment' can be found by looking up a table, so that the expected value can be quickly calculated, the optimal comprehensive evaluation result can be quickly calculated, and the target empty coal control strategy can be obtained. For example, in the above example, S^81_1Store-21636 can be associated with, S^81_2And-22000 may be stored in association.

As shown in fig. 4, a schematic flow chart of calculating a target empty coal control strategy by querying pre-stored data in the embodiment of the present application is shown.

The process specifically comprises the following steps:

step S401: and inquiring the predicted evaluation related to various second furnace condition states in the prestored data according to the furnace condition state at the first moment.

Step S402: calculating expected values according to the predicted associated prediction evaluation of the various second furnace condition states, and calculating various comprehensive evaluation results by combining various alternative air coal control strategies at the first moment;

step S403: and calculating to obtain the target empty coal control strategy based on the optimal comprehensive evaluation result.

The principle is explained below by specific examples.

In one example, the table form of the pre-stored data may be presented in the form of:

as can be seen from the table, there are k kinds of furnace condition states at the T-th moment, S^T_1To S^T-_kRespectively associate V (S)^T_1) To V (S)^T_k) (ii) a At the T-1 th moment, there are g kinds of furnace condition states, which are respectively related to V (S)^T-1_1) To V (S)^T-1_g). When the Tth time is taken as the predicted time, V (S) is inquired and extracted according to a certain current furnace condition state of the T-1^T1) to V (S)^TK) from the current state of the furnace condition of the T-1 th to k states of the furnace condition, to quickly calculate the expected value. Similarly, when the time T-1 is taken as the 'predicted time', the current furnace condition state at the time T-2 is determinedQuery and extract V (S)^T-1_1) To V (S)^T-1_g) To quickly calculate the expected value.

Further, the expected value is substituted into the formula (4) to respectively calculate the comprehensive evaluation result under each alternative air coal control strategy x, and the optimal value is obtained to obtain the corresponding target air coal control strategy.

In some embodiments, the calculation may be performed using equation (3) when building the table. With the Tth time as the 'predicted time', k predicted furnace condition states S exist^T_kThe corresponding prediction evaluation V (S) can be calculated by equation (3)^T_k) (ii) a The furnace condition state S is predicted according to g types of the time T-1 by taking the time T-1 as the predicted time^T-1The corresponding prediction evaluation V (S) can be calculated by equation (3)^T-1_g). By analogy, the prediction evaluation of each predicted furnace condition state at each moment can be obtained.

In some alternative embodiments, the expected value is calculated by including random results caused by random fluctuation parameters, since the furnace condition state may have furnace condition fluctuation, such as coal gas amount fluctuation, furnace temperature fluctuation, and the like. Illustratively, the expectation value on the right side of the equality sign of equation (4) may be further expanded to equation (5) below:

（5）

the formula (5) expands the calculation expected by mathematics, and facilitates accumulation and solution. Optionally, a random fluctuation parameter θ is also introduced into the formula, and is related to, for example, fluctuation of coal gas amount, fluctuation of temperature in the furnace, and the like, and if there are K cases, it is correspondingly expressed as θ_kRespectively probability of

Resulting in a random result at time n +1 of

(stochastic factors such as gas mains fluctuations associated with fluctuations in gas volume);

i.e. is shown in

Take a value of

And theta_kThe prediction evaluation of the furnace condition state predicted at the n +1 th moment is obtained under the condition;

denotes the introduction of x and theta_kResult in random results under the conditions

Is composed of

L denotes that there are L kinds of random results

(ii) a In the second step of formula (5), θ is further introduced_k、Wⁿ⁺¹The latter mathematics expect further transformation unwrapping,

probability corresponding to j cases;

and

can be obtained through data statistics;

is that

And theta_kUnder the condition of

The probability density function of (2) can be obtained by superposing a mixed Gaussian error on the basis of a prediction model (such as a furnace condition prediction model). Equation (5) may be substituted into equation (4) to solve the largest target empty coal control strategy that satisfies the right side of equation (4).

It can be understood that the calculation manner for obtaining the target empty coal control strategy in the above embodiments is only an example, and may be changed according to the actual situation; for example, the consideration of fluctuation is omitted depending on the actual situation, the random fluctuation parameter and the random result may not be introduced, and the calculation of the expected value is simplified, and the like, so the above example is not limited.

Fig. 5 is a schematic flow chart showing a training method of a predictive model for a hot blast stove in the embodiment of the present application.

The training method comprises the following steps:

step S501: obtaining a training data set based on the furnace condition state of the hot blast furnace in the historical time period; each piece of training data in the training data set comprises input data and corresponding reference data;

step S502: at least one predictive model is trained based on the training data set for enabling the at least one predictive model to predict at least a predicted supply air temperature and a predicted total gas consumption at a future time based on a furnace condition state over a period of time.

In some embodiments, the predictive models are multiple, including:

The furnace condition prediction model, the air supply temperature prediction model and the total gas consumption prediction model can be trained through corresponding training data, the error between the output of the prediction model and reference data is calculated, model parameters are adjusted, and the training is completed after iteration until the error is converged.

In some embodiments, there is at least one of:

In a further alternative embodiment, the plurality of submodels of the fusion model in at least one of case 1 to case 3 are respectively assigned with prediction weight parameters representing the magnitude of contribution of the submodels to the prediction result of the fusion model; the prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments. The features of the fusion model have been described in the foregoing embodiments, and are not repeated here.

Fig. 6 shows a block diagram of a control device of a hot blast stove in the embodiment of the present application. The hot blast stove control device 600 in this embodiment corresponds to the embodiment of the hot blast stove control method in the embodiment of fig. 1, and reference may be made to technical features thereof, which are not repeated in this embodiment.

The hot blast stove control device 600 includes:

the acquisition module 601 is used for acquiring a first furnace condition state of the hot blast stove in a period of time; the first furnace condition state includes: empty coal data and furnace condition temperature data;

the prediction module 602 is used for predicting the predicted air supply temperature and the predicted total gas consumption at a second future moment at the first moment according to the first furnace condition state;

and the strategy determining module 603 is used for determining a target air-coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption.

Fig. 7 is a block diagram showing a training device for a prediction model of a hot blast stove in the embodiment of the present application. The training apparatus 700 in this embodiment corresponds to the training method embodiment in fig. 5, and reference may be made to technical features thereof, which are not repeated in this embodiment.

The training apparatus 700 includes:

a training data acquisition module 701, configured to obtain a training data set based on a furnace condition state of the hot blast furnace at a historical time period; each piece of training data in the training data set comprises input data and corresponding reference data;

a training module 702 configured to train at least one prediction model based on the training data set, so as to enable the at least one prediction model to predict at least a predicted supply air temperature and a predicted total gas consumption at a future time according to a furnace condition state of a period of time.

The electronic device 800 comprises at least one processor 801, at least one memory 802 and at least one communication interface 803. The processor 801, the memory 802, and the communication interface 803 are connected by a communication bus and communicate with each other.

A communication interface 803 for communicating with other devices or communication networks, such as ethernet, RAN, WLAN, etc.

The memory 802 may be, but is not limited to, a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, an EEPROM, a CD-ROM or other optical disk storage, optical disk storage (including compact disk, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory may be self-contained and coupled to the processor via a bus. The memory may also be integral to the processor.

The memory 802 is used for storing program instructions for executing the above scheme, and is controlled by the processor 801 to execute. The processor 801 is configured to execute program instructions stored in the memory 802.

The memory 802 stores code that may perform the steps in the hot blast stove control method provided in the above embodiments (e.g., FIG. 1), or sub-steps of the steps thereof developed in other embodiments, etc.; alternatively, the stored code may perform steps in the training method provided in the above embodiments (e.g., fig. 5), or substeps of expansion of steps thereof in other embodiments, and so on.

The processor 801 may also employ or one or more integrated circuits for executing associated programs to implement the steps in the above-described methods.

The processor 801 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the traffic congestion analysis method of the present application may be implemented by integrated logic circuits of hardware in the processor 801 or instructions in the form of software. In implementation, the steps of the congestion identification, identity awareness and identity correlation method of the present application may be implemented by an integrated logic circuit of hardware in the processor 801 or instructions in the form of software. The processor 801 described above may also be a general purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component. The various methods, steps and block diagrams of modules disclosed in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in the memory 802, and the processor 801 reads information in the memory 802, and completes the object recognition method or the model training method according to the embodiment of the present application in combination with hardware thereof.

The communication interface 803 enables communication with other devices or communication networks using transceiver means, such as, but not limited to, transceivers. For example, traffic data collected by the data side can be acquired through the communication interface 803, and a traffic travel policy and the like can be pushed to the terminal.

A bus may include a pathway to transfer information between various components of the device (e.g., memory 802, processor 801, communication interface 803).

As shown in fig. 9, a schematic structural diagram of a hot blast stove system provided in an embodiment of the present application is shown.

The hot blast stove system comprises:

at least one hot blast stove 901 (3-4 in the figure, one example is shown in the figure) which is provided with a gas regulating valve 902 and an air regulating valve 903;

the hot blast stove control system 904 is in communication connection with and controls the gas regulating valve 902 and the air regulating valve 903, and collects the state of the stove condition from the hot blast stove;

the electronic device 905 (structure can refer to fig. 8) for implementing the hot blast stove control method in the foregoing embodiment by executing program instructions may be communicatively connected to the hot blast stove control system 904 through a communication interface, and is configured to acquire the collected furnace condition status and output the target air-coal control strategy to the hot blast stove control system.

In some embodiments, the electronic device 905 may be a server, a server group, a computer, a smart phone, a tablet computer, or the like. The electronic device 905 can be remotely communicated with a hot blast stove control system through a remote network to interact. Or, the electronic device 905 is integrated with the hot blast stove control system, for example, the electronic device may also be located at one side of the hot blast stove control system, and is in communication with the control device for controlling the gas regulating valve 902 and the air regulating valve 903; or, for example, when the electronic device 905 can normally operate in a high-temperature severe environment of the hot blast stove, the electronic device can also be integrated in the gas regulating valve 902 and the air regulating valve 903 to realize intelligent valves, and the opening degree of the electronic device can be intelligently self-regulated according to a target air-coal control strategy.

In some embodiments, the electronic device can implement technical means such as exception capture, memory detection, thread management and the like by running a program, so as to ensure the stability of the execution of the hot blast stove control method. And when abnormal conditions occur, the corresponding log is displayed and stored, so that problems can be conveniently found and analyzed.

In some embodiments, the electronic device may perform upper and lower limit determination on the input and output key parameters of the hot blast stove according to a preset range, and if the determination parameter exceeds the preset range, the global state parameter may be set to 0, and the operation is switched to manual operation or the last output result is maintained. The upper and lower limits of the key parameters of the hot blast stove can be preset as shown below:

KQLL _ MAX =29000 # upper air limit KQLL _ MIN =6000# lower air limit

KQFMKD _ MAX =80# air valve opening upper limit KQFMKD _ MIN =50# air valve opening lower limit

MQLL _ MAX =33000 # gas upper limit MQLL _ MIN =25000 # gas lower limit

MQFMKD _ MAX =60# gas valve opening upper limit MQFMKD _ MIN =35# gas valve opening lower limit

MKB _ MAX =1.3# coal-to-air ratio maximum MKB _ MIN =1.1# coal-to-air ratio minimum

YDWD _ MAX =410# highest temperature at last stage of furnace burning YDWD _ MIN =390 # lowest temperature at last stage of furnace burning

GDWD _ MAX =1350# vault temperature upper limit GDWD _ MIN =1000# vault temperature lower limit

MQF _ SQ = 1# gas valve dead zone KQF _ SQ = 1# air valve dead zone

As shown in fig. 10, a schematic diagram of control signal flow of the control system of the hot blast stove in the embodiment of the present application is shown.

The hot blast stove control system is provided with a master control loop taking the stove temperature as a control object, specifically, the left side 'stove temperature set value' is input, an error quantity is compared with a fed back 'stove temperature measured value', the error quantity is input into a stove temperature Generalized Predictive Control (GPC) model and is correspondingly output to a first sub-control loop taking the gas flow as the control object, the first sub-control loop comprises a gas flow PID controller, a gas flow regulating valve and a gas flow object in sequence, the output signal value of the gas flow object is filtered and then fed back to the input end of the first sub-control loop to calculate the output deviation of the stove temperature generalized predictive control model, and closed-loop regulation is performed; and the result of the output filtered signal value of the gas flow object is multiplied by a proportionality coefficient K and then input into a second sub-control loop taking the air flow as a control object, the difference is compared with the signal value of the air flow object self-fed back by the second sub-control loop, and the second sub-control loop is made to self-regulate according to the difference. The signal value about the gas flow object output by the first sub-control loop and the signal value about the gas flow object output by the second first sub-control loop are output to control the heating furnace temperature, and the measured value of the heated furnace temperature is detected and continuously fed back to be compared with the set value of the furnace temperature, so that the air flow, the gas flow and the related furnace temperature heating are automatically regulated. The air heater control system can be used for setting an air flow regulating valve and a coal gas flow regulating valve according to a target air-coal control strategy.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.

In the foregoing embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In several embodiments provided in the present application, for example, the embodiments of fig. 6 and 7, it can be understood that the apparatus disclosed therein can be implemented in various ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, where a division of modules is merely a logical division, an actual implementation may have another division, for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or modules through some interfaces, and may be in an electrical or other form.

The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.

In addition, functional modules in the embodiments of the present invention may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode.

The integrated processing module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned memory comprises: various media capable of storing program codes, such as a U disk, a ROM, a RAM, a removable hard disk, a magnetic disk, or an optical disk.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable memory, which may include: flash disk, ROM, RAM, magnetic or optical disk, and the like.

The above embodiments are merely illustrative of the principles and utilities of the present application and are not intended to limit the application. Any person skilled in the art can modify or change the above-described embodiments without departing from the spirit and scope of the present application. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical concepts disclosed in the present application shall be covered by the claims of the present application.

Claims

1. A method of controlling a hot blast stove, comprising:

acquiring each first furnace condition state of the hot blast furnace at a time period; the first furnace condition state includes: empty coal data and furnace condition temperature data; the empty coal data comprises at least one of: control parameters of a gas regulating valve and an air regulating valve of the hot blast stove; gas and air flow;

predicting the predicted air supply temperature and the predicted total gas consumption at a second time in the future of the first time according to each first furnace condition state of the time period, wherein the predicting comprises the following steps: predicting a second furnace condition state and total gas consumption at a second moment according to each first furnace condition state of the time interval; and predicting according to the second furnace condition state to obtain the predicted air supply temperature; the second furnace condition state and the total gas consumption at the second moment are predicted according to the first furnace condition state, and the prediction is completed through a furnace condition prediction model; the predicted air supply temperature is obtained according to the second furnace condition state prediction and is completed through an air supply temperature prediction model; the total gas consumption is predicted according to the second furnace condition state, and the total gas consumption is predicted through a total gas consumption prediction model;

determining a target empty coal control strategy of the hot blast stove at the first moment according to the predicted air supply temperature and the predicted optimization trend of the total gas consumption, wherein the control strategy comprises the following steps: calculating corresponding comprehensive evaluation results based on various situations of the predicted air supply temperature and the predicted total gas consumption under the second furnace condition state which is changed into the second time by adopting each alternative air-coal control strategy under the first furnace condition state; wherein the comprehensive evaluation result comprises: the comprehensive result between the gas consumption evaluation generated by the air coal control strategy at the first moment and the expected value of the prediction evaluation based on the predicted air supply temperature and the predicted total gas consumption at the second moment is obtained; and determining a target empty coal control strategy according to the optimal one in the comprehensive evaluation results.

2. The stove control method of claim 1, wherein the furnace condition temperature data comprises at least one of: the dome temperature of the hot blast stove; flue temperature.

3. The stove control method according to claim 1, characterized in that the first condition state further comprises: gas pipeline pressure.

4. The stove control method of claim 1, wherein the second furnace condition status comprises predicted furnace condition temperature data; the predicted furnace condition temperature data includes at least one of: predicting the vault temperature; and predicting the flue temperature.

5. The hot blast stove control method according to claim 1, characterized in that at least one of the following situations exists:

6. The hot-blast stove control method according to claim 5, characterized in that a plurality of submodels of the fusion model in at least one of cases 1 to 3 are respectively assigned with a prediction weight parameter indicating a magnitude of contribution of the submodels to the prediction result of the fusion model; the prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments.

7. The hot blast stove control method according to claim 1, comprising: predicting the predicted data of the future time according to the actual data of the time interval, the combination of the actual data and the predicted data of the time interval or the predicted data of the time interval so as to obtain the data of each time in one working cycle of the hot blast stove; the actual data and the predicted data correspond to at least one of a furnace condition state, air supply temperature data and total gas consumption.

8. The hot blast stove control method according to claim 1, wherein the predictive evaluation based on the predicted supply air temperature and the predicted total gas consumption amount comprises: a difference between a first representation relating to an error between the predicted supply air temperature and the requested supply air temperature and a second representation relating to a total gas consumption.

9. The hot blast stove control method according to claim 1, characterized in that the pre-association between each state of furnace conditions at each moment in a time period and the corresponding predictive evaluation is stored as pre-stored data; determining a target empty coal control strategy according to the optimal one of the comprehensive evaluation results, wherein the method comprises the following steps:

according to the furnace condition state at the first moment, inquiring the predicted associated prediction evaluation of various second furnace condition states in the prestored data;

calculating expected values according to the predicted associated prediction evaluation of the various second furnace condition states, and calculating various comprehensive evaluation results by combining various alternative air coal control strategies at the first moment;

and calculating to obtain the target empty coal control strategy based on the optimal comprehensive evaluation result.

10. The stove control method according to claim 1, characterised in that the calculation of the desired value comprises a random result caused by a random fluctuation parameter.

11. A training method for a predictive model of a hot blast stove is characterized by comprising the following steps:

obtaining a training data set based on the furnace condition state of the hot blast furnace in the historical time period; each piece of training data in the training data set comprises input data and corresponding reference data;

at least one predictive model is trained based on the training data set for enabling the at least one predictive model to predict at least a predicted supply air temperature and a predicted total gas consumption at a future time based on a furnace condition state over a period of time.

12. The training method of claim 11, wherein the plurality of predictive models comprises:

13. Training method according to claim 12, characterized in that at least one of the following situations exists:

14. The training method according to claim 13, wherein a plurality of submodels of the fusion model in at least one of cases 1 to 3 are each assigned with a prediction weight parameter indicating a magnitude of a contribution of the submodel to a prediction result of the fusion model; the prediction weight parameters are updated at each moment, and the updating result of the prediction weight parameters at each moment is related to the overall prediction error performance of the sub-model at a plurality of previous moments.

15. A hot blast stove control device, comprising:

the acquisition module is used for acquiring a first furnace condition state of the hot blast stove in a period of time; the first furnace condition state includes: empty coal data and furnace condition temperature data; the empty coal data comprises at least one of: control parameters of a gas regulating valve and an air regulating valve of the hot blast stove; gas and air flow;

the prediction module is used for predicting the predicted air supply temperature and the predicted total gas consumption at the future second moment at the first moment according to the first furnace condition state, and comprises: predicting a second furnace condition state and total gas consumption at a second moment according to each first furnace condition state of the time interval; and predicting according to the second furnace condition state to obtain the predicted air supply temperature; the second furnace condition state and the total gas consumption at the second moment are predicted according to the first furnace condition state, and the prediction is completed through a furnace condition prediction model; the predicted air supply temperature is obtained according to the second furnace condition state prediction and is completed through an air supply temperature prediction model; the total gas consumption is predicted according to the second furnace condition state, and the total gas consumption is predicted through a total gas consumption prediction model;

the strategy determination module is used for determining a target air-coal control strategy of the hot blast stove at the first moment according to the optimization trends of the predicted air supply temperature and the predicted total gas consumption, and comprises the following steps:

calculating corresponding comprehensive evaluation results based on various situations of the predicted air supply temperature and the predicted total gas consumption under the second furnace condition state which is changed into the second time by adopting each alternative air-coal control strategy under the first furnace condition state; wherein the comprehensive evaluation result comprises: the comprehensive result between the gas consumption evaluation generated by the air coal control strategy at the first moment and the expected value of the prediction evaluation based on the predicted air supply temperature and the predicted total gas consumption at the second moment is obtained; and determining a target empty coal control strategy according to the optimal one in the comprehensive evaluation results.

16. Training device for a predictive model of a hot blast stove, characterized in that it comprises:

the training data acquisition module is used for acquiring a training data set based on the furnace condition state of the hot blast furnace in the historical time period; each piece of training data in the training data set comprises input data and corresponding reference data;

and the training module is used for training at least one prediction model based on the training data set so as to enable the at least one prediction model to at least predict the predicted air supply temperature and the predicted total gas consumption at the future moment according to the furnace condition state of a period of time.

17. An electronic device comprising a processor and a memory;

the memory to store computer program instructions;

the processor executing computer program instructions that invoke the memory performs the stove control method of any of claims 1 to 10 or the training method of any of claims 11 to 14.

18. A computer-readable storage medium, characterized in that it has stored therein program instructions which, when executed by a processor, carry out the hot blast stove control method according to any one of claims 1 to 10 or the training method according to any one of claims 11 to 14.

19. A hot blast stove system, comprising:

at least one hot blast stove with a gas regulating valve and an air regulating valve;

the hot blast stove control system is in communication connection with and controls the gas regulating valve and the air regulating valve and collects the state of the stove condition from the hot blast stove;

an electronic device comprising a processor, a memory, and a communication interface;

the memory to store computer program instructions;

the processor executing computer program instructions that invoke the memory to perform the hot blast stove control method of any one of claims 1 to 10; the communication interface is in communication connection with or integrated with the hot blast stove control system and is used for acquiring the acquired furnace condition state and outputting the target air coal control strategy to the hot blast stove control system.

20. The stove system according to claim 19, wherein the stove control system has an overall control loop with stove temperature as a control target, the overall control loop comprising: a sub-control loop taking the gas flow as a control object; and a sub-control circuit for controlling the air flow rate.