CN116822380B - Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin - Google Patents

Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin Download PDF

Info

Publication number
CN116822380B
CN116822380B CN202311094435.2A CN202311094435A CN116822380B CN 116822380 B CN116822380 B CN 116822380B CN 202311094435 A CN202311094435 A CN 202311094435A CN 116822380 B CN116822380 B CN 116822380B
Authority
CN
China
Prior art keywords
data
parameters
model
unit
real
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202311094435.2A
Other languages
Chinese (zh)
Other versions
CN116822380A (en
Inventor
马军
张峻嘉
李祥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kunming University of Science and Technology
Original Assignee
Kunming University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kunming University of Science and Technology filed Critical Kunming University of Science and Technology
Priority to CN202311094435.2A priority Critical patent/CN116822380B/en
Publication of CN116822380A publication Critical patent/CN116822380A/en
Application granted granted Critical
Publication of CN116822380B publication Critical patent/CN116822380B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/30—Administration of product recycling or disposal
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Human Resources & Organizations (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Medical Informatics (AREA)
  • Geometry (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Hardware Design (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Sustainable Development (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention discloses a collaborative optimization method for tail gas recycling in a copper smelting process based on digital twinning, and belongs to the field of intelligent control of metal smelting tail gas recycling. Acquiring state parameters and process parameters of a sulfuric acid reaction tower to obtain complete data information of the whole production process, constructing a digital twin model of the desulfurization and acid production process according to the data, continuously iterating the constructed digital twin model of the production process to obtain optimal operation parameters through a BPNLP neural network, analyzing and comparing the optimal operation parameters with real-time data of entity equipment, uploading the acquired data to a man-machine interaction interface through a communication protocol, and realizing virtual-real linkage of the whole industrial production and a client; finally, the operation parameters of the actual production process are optimally controlled in real time through a process parameter control system according to the optimized parameter result, so that virtual control and real control are realized, and the desulfurization efficiency and the sulfuric acid preparation speed are maximized for a long time.

Description

Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin
Technical Field
The invention relates to the field of intelligent control of metal smelting tail gas recycling, in particular to a collaborative optimization method for tail gas recycling in a copper smelting process based on digital twinning.
Background
Currently, zujun, zhang Feng et al propose flue Gas heat exchange technology for MGG (Mitsubishi Gas-Gas Heater) with SO in the flue Gas 2 Is absorbed and oxidized into ammonium sulfate, and then evaporated, concentrated and crystallized to obtain an ammonium sulfate product, cai Bing et al propose a hydrogen peroxide flue gas desulfurization technology which has become the main means for solving the atmospheric pollution of sulfur dioxide in more factories at present, but the whole process flow involves the rectification operation of liquid phase separation of sulfur dioxide and hydrogen peroxide, which is typical of Multiple Input Multiple Output (MIMO), delay and severeThe problems of modeling, optimization and control in the complex process with heavy nonlinearity and strong parameter coupling are hot spot problems of many scholars. And a great deal of research is focused on the study of desulfurizing and acid making for pollution control instead of the conventional desulfurization. In the process of acid production, the input of relevant parameters such as liquid level, temperature, material ratio and the like can have great influence on the conversion rate and absorption rate of sulfur dioxide which are output parameters. Meanwhile, the research of the existing desulfurization and acid production is mostly focused on the aspect of environmental pollution evaluation, or only the recovery efficiency of sulfuric acid is concerned. The research on the collaborative optimization method for pollution control and resource recovery in the needle desulfurization and acid production process is less.
BP neural network is a model widely applied to industrial process optimization; whereas the gradient method adopted by BP neural network tends to be a local minimum of Fuls; if the fitting function is too complex, multiple fitting local minima may occur, which can present a significant hurdle to practical applications.
Disclosure of Invention
Aiming at the optimization problem in the acid making rectification operation, the invention aims to provide a collaborative optimization method based on digital twin in the recovery and utilization of tail gas in the copper smelting process, which can realize the real-time acquisition, analysis and decision making of state parameters in the acid making process, and establishes an interactive, visual and optimizable digital twin model of a flue gas acid making system by analyzing a sulfur dioxide method and coupling and distributing multiple parameters and a desulfurization process flow, thereby achieving the aims of 'virtual-real combination and virtual control-real', realizing the maximization of the production rate and larger economic value of sulfur dioxide acid making, and specifically comprising the following steps:
step1: the state parameters, the process parameters and the evaluation indexes of the desulfurization and acid production in the smelting process are called from the conventional database, and the data are preprocessed, predicted and mined to construct the conventional database.
Step2: real-time data, including state parameters and process parameters, are collected through sensors such as contact type sensors and electromagnetic type sensors of the field industry or by means of manual test records and the like, and the parameters together form a real-time database; the state parameters comprise temperature, pressure, liquid level and gas-liquid ratio values in the furnace body; the process parameters comprise the concentration and flow rate of the input sulfur dioxide and hydrogen peroxide, and the concentration and flow rate of the output sulfuric acid.
Step3: the prior database and the real-time database form a digital twin dynamic model database, the data intercommunication connection between the digital twin dynamic model database and a front-end interface formed by a server and a unit 3D is realized through a TCP communication protocol, and a visual graph of relevant data in the digital twin dynamic database is displayed on the front-end interface.
Step4: the digital twin dynamic model database obtained in Step3 is used as a digital twin model simulation signal to be introduced into a digital twin model for desulfurizing and producing acid, the model is simulated through data, the simulation model is built by adopting nonlinear optimization based on a two-stage BPNLP neural network and combining an approximate global optimization algorithm, and the simulation model is used for continuously optimizing process parameters corresponding to optimal parameters obtained through the digital twin model, and obtaining the most suitable production optimal engineering parameters according to the maximized parameters of sulfur dioxide desulfurization rate and sulfuric acid conversion rate.
Step5: carrying out data mining similar aggregation on the predicted process parameters in the state of maximum sulfur dioxide desulfurization rate and maximum sulfuric acid conversion rate and a conventional database, carrying out noise reduction and missing value compensation treatment, and then carrying out parameter comparison on the predicted process parameters and a real-time database to judge which parameters of the conventional real-time data need to be improved; and finally, obtaining an optimal parameter improvement scheme by optimizing a global algorithm under real-time data.
Step6: and visualizing the optimal parameter data improvement scheme through a TCP protocol, and displaying a visualized graph of the data on a human-computer interface of the unit 3D front end.
Step7: in the optimal parameter improvement scheme, an operator issues a control instruction according to the obtained parameter improvement scheme and manual experience.
Step8: the control instruction is connected with a simulation control model of the sulfur removal and acid production of the simulink (comprising a liquid level-flow cascade control system of gas and liquid inlet and outlet, a liquid level-pressure cascade control system in the tower, and a dual-input and dual-output hierarchical decoupling control system of the temperature in the tower) through a TCP protocol to form a control system, as shown in figure 3, the far end of the control system is connected with an actual industrial server, the optimal parameter value is compared with the real-time data value, the parameter control is carried out on the actual furnace body according to the difference value, and the valve and the regulator in production are controlled to realize guidance on the actual production.
Step9: continuously repeating the steps from Step2 to Step5, collecting real-time data, obtaining optimal control parameters through a simulated neural network digital twin model, comparing the two data, further feeding back and controlling furnace body process parameters, continuously repeating, and permanently guaranteeing the maximum desulfurization efficiency and the maximum sulfuric acid preparation speed.
Because the final sulfuric acid yield and the desulfurization conversion rate in the whole reaction process are jointly influenced by the coupling of multiple factors such as the temperature, the pressure, the catalyst and the hydrogen peroxide liquid level in the tower, a simulation model is built by adopting the nonlinear optimization combination approximate global optimization algorithm based on the two-stage BPNLP neural network as described in S4; the two-stage BPNLP network model is used for carrying out smelting parameter optimization process based on the fitted neural network model after the neural network in the first stage is preprocessed to form the fitting of the smelting process; according to the characteristics of the whole two-stage process, the corresponding condition constraint and the smoothing process of the BP neural network output function, an approximate global optimization algorithm is adopted to form a two-stage BPNLP network model with unified constraint, and the two-stage BPNLP network model specifically comprises the following components, relates to routes and operation flows as shown in figure 3:
S1: processing data in the data model base, and prescribing input samples transmitted from the data base; according to the production process of desulfurizing and producing acid in copper smelting, under the condition that an input sample obeys normal distribution, the mean value and standard deviation sigma in the sample are selected.
S2: the BP neural network model belongs to an error back propagation neural network, and needs to be constructed in three aspects: an input layer, a hidden layer and an output layer; the model contains neurons, weights, thresholds, layers, and activation functions; according to the invention, a BP neural network model is established according to real-time state parameters of the desulfurization and acid making process and the recycling conversion rate of pollutant sulfur.
S3: firstly, establishing parameters of an input layer, a hidden layer and an output layer, taking five state parameters in a database as input variables of a BP natural network model, namely pressure, temperature in a tower, liquid level, gas-liquid ratio value and flow of sulfur gas, introducing the parameters of the input layer, the hidden layer and the hidden layer into two layers on the basis of calculation of the hidden layer of the BP neural network model, setting nodes according to actual process, wherein the output layer is the conversion rate and the desulfurization efficiency of sulfuric acid.
S4: the transfer function for the two hidden layers can be defined as
Where p and q are the dimensions of the input and output variables and N is the number of samples; i is the hidden layer of the neuron, i=1 represents the input layer,i.e. as input variable, e.g.)>For the pressure of the reaction furnace in the smelting process, +.>Is the temperature in the tower>Is the liquid level in the tower>For feeding the gas-liquid ratio value and +.>Sulfur gas flow rate and the like, +.>Then the output variable parameter is represented and is for +.>The number of samples of each type of input variable in (1) is defined as +.>To calculate for each sample, the following is specific:
since i is the hidden layer of neurons, each sample needs to calculate its output valueThe method comprises the following steps:
it is assumed here that for the kth layer there are s neurons; l (l=1, 2, …… h) The number of layers that are hidden layers; wherein the method comprises the steps ofRepresenting the weight corresponding to each input variable, wherein θ is a bias value, and +.>The offset value of the 1 st hidden layer of the k layer is the offset value, and p is a fixed value to represent the number of samples; />Is the sample node of the input variable at the k-th layer.
Also, toAs input to the model, the output +.>：
To approximate the result to the target, outputThe error with the actual output is reversely transmitted from the output layer to the next layer of the network; />The bias value of the ith hidden layer in each transmission is represented, and the threshold value and the weight value of the neuron are adjusted once in each transmission; this recursive process is repeated until the kth hidden layer, so that the function value of the output layer can be obtained +. >：
S5: however, due to insufficient generalization capability of the BP neural network model, abnormal values may occur in the simulation result of the model, and therefore, secondary smoothing processing is required to be performed on the model output; the generalization precision of the model before non-smoothing is defined as the proportion of singular values (negative numbers) in all sample predicted values; so that the generalization accuracy can be set to an acceptable range;
according to the core of the invention: under the conditions of the conditions and the process parameter combination, the maximum conversion rate and the maximum desulfurization efficiency are obtained; and then, performing generalization precision optimization on the model, and performing prediction simulation learning of the two-stage BPNLP by adopting the BP neural network after parameter optimization.
S6: setting the output function of the BP neural network model as the maximum conversion rate and the maximum desulfurization efficiency of sulfuric acid in smelting flue gas, so as to perform better simulation prediction on the BP neural network model and solve the problem of minimization of the output function of the smelting flue gas desulfurization; the adopted pollution emission reduction optimized BPNLP neural network model is expressed as the following mathematical function:
j represents an index function for indicating a function index of an optimal desulfurization conversion rate and rate to be obtained,the function Min represents a mapping of the output of the two-stage BPNLP estimation result, where the input variable x is the desired index; wherein c is a linear transformation vector, which is a 2×1 vector of the conversion rate and desulfurization efficiency of sulfuric acid; Representing the estimated result of the BP model prediction,representing a linear transformation of the desired index from the output estimation result of the BPNLP model with input variables.
To minimize emissions problems, the nonlinear programming is as follows:
wherein b, L, U are constant column vectors, f is a nonlinear function, x is a decision variable vector which is a 5 x 1 vector comprising pressure, temperature in the tower, liquid level, gas-liquid ratio value and flow of sulfur dioxide; a is a constant matrix, and the linear cut-off function Ax represents data after passing through the normalized domain and after eliminating anomalies.
S7: by fitting the function N (x) above, x can generate multiple local minima, yielding multiple solutions and further affecting global optimization; selecting an initial value by adopting a random scattered point method so as to adapt to the equal step change of initial value optimization; confirming a minimum value of the model simulation selected from the optimization results in the optimal values; for the whole neural network model, the region of the variable to be optimized is gridded by using a grid and sample method, and a solving process in a framework is evolved to find a global optimal solution by combining the local searching skill.
S8: in addition, smoothing is carried out on the output function, so that the predicted value is meaningful; assuming that the output samples take a certain dimension of the output values, then a further expansion of the comparison of the model is required to model the output values, so that the output function takes values in the domain smoothly and filters out unreasonable values (e.g. under substantially the same conditions, the predicted output values differ significantly from their neighbors by too much or too little).
S9: and finally, solving the optimal parameters of the data twinning model, and sending the obtained result to a follow-up content for completing man-machine interaction in the unit 3D by carrying out data visualization in a specific step S6 of a collaborative optimization method for tail gas recycling in the copper smelting process based on digital twinning.
The invention further aims to provide a collaborative optimization system based on digital twin for recycling tail gas in a copper smelting process, which is shown in figure 2 and comprises a database module, a data analysis module, a visualization module, an equipment management control module and a communication service module.
The database module is used for acquiring real-time data in the operation process of desulfurizing and producing acid in the smelting process and acquiring data information of the tower body in the whole process flow; the system comprises a geometric parameter data unit, an operation state parameter unit, a real-time process parameter unit and a neural network model unit.
The data analysis module obtains state data from the database module, performs pretreatment and analysis on the data, and obtains an optimal parameter predicted value through a digital twin model, thereby achieving the aim of resource recovery collaborative optimization in the desulfurization and acid production process; the method comprises an algorithm library unit, a data preprocessing unit, a feature extraction unit and a data mining unit.
The visualization module is mainly used for graphically displaying the data of the two modules; the method comprises a fusion model unit, a result transmission unit, a data imaging unit and a graphic dynamic display unit.
The intelligent equipment management module is used for combining the module results, comprises the database module, realizes the full life cycle information acquisition of equipment, completes the real-time intelligent optimization operation and maintenance of the equipment, and carries out real-time and efficient control on the state parameters of actual production under the output of the control system according to the optimal operation parameters; the system comprises an evaluation index unit, a control module unit, an intelligent decision unit and a control instruction unit.
The communication service module mainly realizes the mutual association among the four modules, data in the database module is called to the data analysis module in real time, the result obtained by the data analysis module is sent to the visualization module, and finally the result is applied to the intelligent management module of the equipment according to the data visualization result to complete information calling and association among the modules; various types of messages including production data text, live images, etc.; the real-time communication function is provided to complete the image real-time updating function in the visualization module, ensure the correctness, stability and safety of communication, ensure the correct instruction of the intelligent operation and maintenance module of the equipment to be issued, and complete regulation and control; and records the running log of each module for recording the platform so as to be convenient for subsequent analysis and debugging.
Preferably, the geometric parameter data unit of the present invention contains actual field device data: the method comprises the steps of producing a furnace body, the relative sizes of an inlet pipeline and an outlet pipeline, gas-liquid flow, production parameters and the like, and storing and retrieving long-term operation data;
the operation state parameter unit contains operation data of production smelting, including relevant information such as temperature, pressure, liquid level and the like in a production furnace body, and describes the state and performance of a system or equipment in operation; and provides information about the health status, work efficiency, resource utilization, etc. of the whole system or equipment, and stores and retrieves long-term operation data.
The real-time process parameter unit comprises a pipeline input quantity and a production output quantity which are updated and enter the reaction in real time, is used for describing and evaluating the current state and performance of a system, equipment or a process, provides information about the health condition, the working efficiency, the resource utilization condition and the like of the whole system or the equipment, and stores and retrieves long-term operation data.
The neural network model unit records the operation log of each time in the neural network model in a long-term operation state, and records the weight of relevant parameters, the iteration times and the output-input ratio of each time.
The algorithm library unit is used for carrying out secondary smoothing on the output of the BP model, the prediction reliability is improved by utilizing the strong nonlinear mapping capability and the flexible network structure, large sample data such as a copper smelting desulfurization and acid making process tend to be better fitted, and the overall optimal search method is combined to finish the solving of the optimal parameters of the data twin model.
According to the data preprocessing unit, the optimal parameters in the historical data are changed, so that the data missing frequently occurring in the real-time data acquisition process and the large-deviation noise are preprocessed in advance, and the repeated and large-quantity identical optimization of equipment data is avoided.
According to the invention, the feature extraction unit predicts and trains data and verifies sample classification by taking five state parameters in a database as input variables of a BP natural network model through the BP neural network model, wherein the output layer is the conversion rate and the desulfurization efficiency of sulfuric acid, and the hidden layer is arranged.
The data mining unit discovers hidden modes, association rules and trends in long-term production data, performs advanced data analysis before feature extraction, and mines the overall features of the obtained production input information, so as to identify trends, abnormal modes and correlations in the data.
The fusion model unit model fusion combines the prediction results of a plurality of data analysis modules to obtain more accurate and reliable prediction; the performance of the whole system is improved by combining the prediction results obtained by different data and different dates in the model, so that higher accuracy and robustness are obtained.
The result transmission unit of the invention is used for carrying out timing retrieval on the optimal parameters obtained by the data analysis module, updating the obtained optimal parameters in real time, and transmitting the output results to the data analysis system at the rear end in a proper form so as to support decision making, provide insight, drive the subsequent visualization process and carry out further analysis.
The data imaging unit performs visual graphic display on the result optimal parameter data obtained by the result transmission unit, and uses a bar graph, a line graph, a scatter graph, a pie graph and the like to convey the relationship, trend and mode among data of output parameters, process parameters and the like.
The graphic dynamic display unit mainly drives the whole module to continuously call real-time data to the data analysis module and update the graphic of the data visualization, so that the graphic can be updated along with real-time or frequent change of the data, and the latest information and trend can be presented in time.
The evaluation index unit is mainly used for comparing the obtained optimal operation parameters with the real-time operation parameters, obtaining parameters which need to be weighted according to the evaluation indexes, and adjusting and controlling the parameters under different evaluation indexes with high weight, so that the accuracy, performance and reliability of the model are measured, and workers are helped to know the performance of the model and make decisions.
The control module unit of the invention is a liquid level-flow cascade control system for gas-liquid inlet and outlet according to the change value obtained by comparison, the liquid level-pressure cascade control system in the tower and the double-input and double-output graded decoupling control of the temperature in the tower, and the control from digital twin simulation to actual production is completed under the regulation of feedback.
The intelligent decision unit provided by the invention is used for identifying modes, trends and rules by utilizing the existing data and information of the data analysis module as a basis and analyzing, excavating and integrating the obtained optimal parameters and the on-site actual production data, so that a targeted and reliable production decision suggestion is provided.
The control instruction unit is a key component responsible for generating, verifying and sending the control instruction after intelligent decision after the intelligent decision unit, monitoring the system state and receiving feedback; and the operator sends a command to the system to regulate and control valves, regulators and controllers for actual production control parameters in actual production, and issues control instructions related to valve opening and controller opening adjustment to guide the final target copper smelting tail gas desulfurization and acid production resource recovery to be cooperatively optimized.
The communication service module provides the capability of connecting, transmitting and interacting the four modules; the communication module allows communication among different systems, devices or nodes through physical or virtual channels, supports transmission, reception and processing of information transmission of data from the database module, and transmission and reception of information of the data analysis module, including various types of information such as production data text, field images and the like; the real-time communication function is provided to complete the image real-time updating function in the visualization module, ensure the correctness, stability and safety of communication, ensure the correct instruction of the intelligent operation and maintenance module of the equipment to be issued, and complete regulation and control; and records the running log of each module for recording the platform so as to be convenient for subsequent analysis and debugging.
Compared with the prior art, the invention has at least the following beneficial effects:
(1) The digital twin platform is mainly used for establishing real-time data uploading interaction, establishing a digital twin model to perform real-time data analysis and obtain optimal operation parameters of desulfurization efficiency and sulfuric acid conversion rate prediction, and comparing and issuing real-time data and instructions according to a digital twin technology and a BPNLP neural network algorithm aiming at complex problems of modeling, optimizing and controlling problems in a typical complex process of multiple input and multiple output, large delay, serious nonlinearity and strong parameter coupling.
(2) According to the characteristics of the copper smelting desulfurization and acid production process in actual production, the optimal desulfurization parameters are obtained by adopting a BPNLP neural network based on the combination of the secondary smoothing treatment of the BP neural network and nonlinear optimization by using an approximate global optimization method. According to the invention, by setting up a simulation platform, the platform can acquire key parameter information such as pressure, temperature and liquid level height in a tower in real time in the flue gas acid making rectification process through TCP communication in real time, virtual-real linkage can be realized by data sharing of each unit, digital twin simulation is performed according to the conversion reaction of sulfur dioxide, and further, the optimal sulfur dioxide desulfurization rate and the optimal sulfuric acid equilibrium conversion rate in the flue gas acid making process and corresponding process parameters such as temperature, pressure and the like in the optimal equilibrium rate are predicted and obtained, and finally, the optimal desulfurization efficiency is achieved by controlling the actual production equipment controller through feedback.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for resource recovery collaborative optimization digital twin in copper smelting tail gas desulfurization and acid production based on a two-stage BPNLP network model;
FIG. 2 is a desulfurization and acid making process flow and a digital twin model constructed by the invention;
FIG. 3 is a flow chart of the operation of the two-stage BPNLP neural network model employed in the present invention;
FIG. 4 is a graph of the results of example sulfuric acid optimum parameter control parameters; (a) In the embodiment 1, the iterative optimization regulation and control process of the maximum sulfuric acid conversion rate under the BPNLP algorithm is carried out; (b) In the embodiment 1, the iterative optimization regulation and control process of the maximum desulfurization rate under the BPNLP algorithm is performed; (c) In the embodiment, the control process of the actual liquid level is assumed that the liquid level is 40 meters at the optimal output value.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
A collaborative optimization method for tail gas recycling in a copper smelting process based on digital twinning specifically comprises the following steps:
s1: the state parameters, the process parameters and the evaluation indexes of the desulfurization and acid production in the smelting process are called from the conventional database, and the data are preprocessed, predicted and mined to construct the conventional database which can be interacted, calculated and optimized.
The data content adopted in the prior database specifically comprises: the geometrical characteristics of the tower body equipment such as the height of the tower body, the diameter of the tower body, the size of the end enclosure of the tower body, the air supply quantity, the diesel quantity, the oxygen supply quantity and the like which are introduced by pipelines, and the state parameters including the temperature, the liquid level, the pressure, the ratio and the like of the tower body of the desulfurizing and acid-making tower; and process parameters such as an input end (related parameters such as flow, flow speed, concentration and the like of sulfur dioxide and hydrogen peroxide), an output end (related parameters such as sulfuric acid content representing sulfur dioxide conversion rate, water output representing desulfurization rate and the like); and carrying out relevant evaluation indexes such as tail gas treatment conversion rate, desulfurization efficiency, sulfuric acid conversion rate, sulfur content of output tail gas and the like on the output result.
S2: real-time data are acquired through contact type sensors, electromagnetic type sensors and the like of field industry or by means of manual test records and the like, the state parameters comprise state parameters of temperature, pressure, liquid level and gas-liquid ratio values in a furnace body, the process parameters comprise process parameters of concentration and flow rate of input sulfur dioxide and hydrogen peroxide and concentration and flow rate of output sulfuric acid, and a real-time database is formed by the parameters together.
S3: the traditional database and the real-time database form a digital twin dynamic model database, the data intercommunication connection between the digital twin dynamic model database and a front-end interface formed by a server and a unit 3D is realized through a TCP communication protocol, and a visual graph of relevant data in the digital twin dynamic database is displayed on the front-end interface.
S4: constructing a digital twin model of the copper smelting tail gas desulfurization and acid making process flow by using the collected past data and real-time production data: and (3) introducing the digital twin dynamic model database obtained in the step (S3) into a digital twin model for desulfurizing and producing acid as a digital twin model simulation signal, performing simulation on the model through data, and continuously optimizing process parameters corresponding to optimal parameters obtained through the digital twin model by adopting a method of combining nonlinear optimization and approximate global optimization based on a two-stage BPNLP neural network, thereby obtaining the most suitable production optimal engineering parameters according to the maximized parameters of sulfur dioxide desulfurization rate and sulfuric acid conversion rate.
S5: carrying out data mining similar aggregation on the predicted process parameters in the state of maximum sulfur dioxide desulfurization rate and maximum sulfuric acid conversion rate and a conventional database, carrying out noise reduction and missing value compensation treatment, and then carrying out parameter comparison on the predicted process parameters and a real-time database to judge which parameters of the conventional real-time data need to be improved; and finally, obtaining an optimal parameter improvement scheme by optimizing a global algorithm under real-time data.
S6: and the optimal parameter data improvement scheme is visualized through the TCP protocol, and a visualized graph of the data is displayed on a human-computer interface of the unit 3D front end.
S7: in the optimal parameter improvement scheme, an operator issues a control instruction according to the obtained parameter improvement scheme and manual experience.
The control instruction is connected with a simulation control model of the sulfur removal and acid production of the simulink (comprising a liquid level-flow cascade control system of gas and liquid inlet and outlet, a liquid level-pressure cascade control system in the tower and a dual-input and dual-output hierarchical decoupling control system of the temperature in the tower) through a TCP protocol to form a control system, as shown in figure 3, the far end of the control system is connected with an actual industrial server, the optimal parameter value is compared with the real-time data value, the parameter control is carried out on the actual furnace body according to the difference value, and the valve and the regulator in production are controlled to realize guidance on the actual production.
S8: and (3) continuously repeating the steps S2-S5, collecting real-time data, obtaining optimal control parameters through a simulated neural network digital twin model, comparing the two data, further feeding back and controlling furnace body process parameters, continuously repeating, and permanently ensuring the maximum desulfurization efficiency and the maximum sulfuric acid preparation speed.
As a further preferred embodiment of the present invention:
the invention relates to a whole copper smelting flue gas desulfurization acid-making desulfurization process, which comprises the steps of drying and washing flue gas discharged by a smelting furnace, wherein sulfur dioxide is catalyzed and oxidized into sulfur trioxide in a converter, and then the sulfur dioxide reacts with water in an absorption tower to generate sulfuric acid, and the chemical equation of the reaction process is as follows:
wherein Q is the heat evolved by the chemical reaction.
Because the final sulfuric acid yield and the desulfurization conversion rate in the whole reaction process are jointly influenced by the coupling of multiple factors such as the temperature, the pressure, the catalyst and the hydrogen peroxide liquid level in the tower, a simulation model is built by adopting the nonlinear optimization combination approximate global optimization algorithm based on the two-stage BPNLP neural network as described in S4; the two-stage BPNLP network model is used for carrying out smelting parameter optimization process based on the fitted neural network model after the neural network in the first stage is preprocessed to form the fitting of the smelting process; according to the characteristics of the whole two-stage process, the corresponding condition constraint and the smoothing process of the BP neural network output function, an approximate global optimization algorithm is adopted to form a two-stage BPNLP network model with unified constraint, and the two-stage BPNLP network model specifically comprises the following components, relates to routes and operation flows as shown in figure 3:
(1) Processing data in the data model base, and processing sample data before training the neural network; firstly, defining an input sample; according to the production process of desulfurization and acid production in copper smelting, assuming that an input sample is subjected to normal distribution, and selecting a mean value and a standard deviation sigma in the sample by utilizing a basic principle of statistical data processing; by expanding the actual interval output value of the model, the sample variance 2σ is the sum of squares of the differences between the sample data and the expected value; training and verifying sample classification, wherein the main content comprises the steps of dividing all experimental data into training samples and simulation verification samples to construct a BP natural network model; and comparing the corresponding model errors by calibrating the training set and the prediction set.
(2) The BP neural network model belongs to an error back propagation neural network, and needs to be constructed in three aspects: a parameter input layer, a hidden layer and a prediction result output layer in smelting production; the model contains neurons, weights, thresholds, layers, and activation functions; in the embodiment, a BP neural network model is established according to real-time state parameters of the desulfurization and acid making process and the recycling conversion rate of pollutant sulfur.
(3) Firstly, establishing parameters of an input layer, a hidden layer and an output layer, and taking the parameters as input variables of a BP natural network model according to five state parameters in a database, namely, pressure, temperature in a tower, liquid level, gas-liquid ratio value and flow of sulfur gas; the output layer is the conversion rate and desulfurization efficiency of sulfuric acid; based on the calculation of the hidden layer of the BP neural network model, the BP neural network model is a network composition structure with the dimension of 5-25-15-4, the hidden layer between the smelting parameters of the secondary desulfurization process and the sulfuric acid collaborative optimization is set as two layers, and nodes are set according to the actual process.
(4) The transfer function for the two hidden layers can be defined as
I.e. as input variable, e.g.)>For the pressure of the reaction furnace in the smelting process, +.>Is the temperature in the tower>Is the liquid level in the tower>For inputting gas-liquid ratio value of pipeline into tower、/>Is communicated with the flow of sulfur gas; />For output, ->Then the conversion of sulfuric acid at output is indicated, +.>Is the sulfuric acid preparation speed.
The recursive process of the optimal prediction iteration of the two phases of the BPNLP is repeated until the H-th hidden layer, so that the function values of the five-dimensional input x and the two-dimensional output y can be obtained：
(5) However, when the method is practically applied to the copper smelting industry, the single BP neural network model is adopted, so that the problem of insufficient generalization capability and inaccurate simulation result exists because the desulfurization process operation is a typical complex process with multiple inputs and multiple outputs, large delay and serious nonlinearity and strong parameter coupling; therefore, the model output needs to be subjected to secondary smoothing; the generalization precision of the model before non-smoothing is defined as the proportion of singular values (negative numbers) in all sample predicted values; so that the generalization accuracy can be set to an acceptable range.
(6) Setting the output function of the BP neural network model as the maximum conversion rate and the maximum desulfurization efficiency of sulfuric acid in smelting flue gas, so as to perform better simulation prediction on the BP neural network model and solve the problem of minimization of the output function of the smelting flue gas desulfurization; the adopted pollution emission reduction optimized BPNLP neural network model is expressed as the following mathematical function:
J represents an index function for representing a function index of an optimal desulfurization conversion rate and a rate to be obtained, and Min represents a predicted output to be obtained from a plurality of results estimated from the two-stage BPNLPWherein the input variable x (each input parameter) is the desired index; wherein c is a linear transformation vector, which is a 2×1 vector of the conversion rate and desulfurization efficiency of sulfuric acid;the estimation result representing the BP model prediction represents a linear transformation of a desired index from the output estimation result of the BPNLP model with input variables.
To minimize emissions problems, the nonlinear programming is as follows:
wherein b, L, U are constant column vectors, f is a nonlinear function, x is a decision variable vector which is a 5 x 1 vector comprising pressure, temperature in the tower, liquid level, gas-liquid ratio value and flow of sulfur dioxide; a is a constant matrix, and the linear cut-off function Ax represents data after passing through the normalized domain and after eliminating anomalies.
(7) By fitting the function N (x) above, x may generate multiple local minima, which may yield multiple solutions and further impact global optimization; selecting an initial value by adopting a random scattered point method so as to adapt to the equal step change of initial value optimization; and confirming the minimum value of the model simulation selected from the optimized results in the obtained optimal values in the desulfurization and acid production process.
(8) In addition, smoothing the output function to make the predicted value meaningful, assuming that the output sample takes a certain dimension value of the output value, then further extending the comparison of the model is needed to simulate the output value, so that the output function takes the value in the domain smoothly and filters out unreasonable values (e.g., negative values).
(9) And finally, solving the optimal parameters of the data twinning model, and sending the obtained result to a follow-up content for completing man-machine interaction in the unit 3D by carrying out data visualization in a specific step S6 of a collaborative optimization method for tail gas recycling in the copper smelting process based on digital twinning.
As a further preferred embodiment of the present invention:
s1: the traditional database is called, and meanwhile, the data of the desulfurization and acid making process in the copper smelting factory is taken as real-time production data for demonstration:
the physical parameters and state process parameters of this time are expressed as follows: the temperature in the reaction tower is set to be 1000-1200 ℃, the pressure is about 2MPa, the height of the reaction tower is 70 m, the diameter is 10 m, the flow rate of sulfuric acid is 500 cubic meters per hour, and the flow rate is 5 m per second.
S2: connecting a given device to a digital twin-based rolling bearing residual life online prediction apparatus, comprising: the system comprises a database module, a data analysis module, a visual graphic module, a communication service module and an equipment intelligent management module;
S3: the data acquired in the step S1 is assumed to be real-time data acquired in actual production, the real-time data is input into a digital twin simulation model, and a digital model taking the data as an operation state is constructed, so that dynamic interaction and virtual-real linkage with actual production equipment are realized through the data.
S4: according to the digital twin model constructed in the step S3, running the state parameter, the physical parameter and the process parameter related data under simulation production conditions through an algorithm library, and searching and predicting optimal running output parameters (the most suitable desulfurization efficiency and sulfuric acid conversion rate) according to five-dimensional input variables (reaction furnace pressure, tower internal temperature, tower internal liquid level, gas-liquid ratio value from an input pipeline to the tower and flow rate of sulfur gas) under continuous repeated simulation iteration of a main model in the algorithm library, namely a BPNLP neural network model; the optimal output simulated by the experiment is obtained according to the input: 94.23% desulfurization rate, 98.9% sulfuric acid conversion rate and 8% gas concentration. The desulfurization rate and sulfuric acid conversion rate results in the whole iteration process are shown in fig. 4 (a) and fig. 4 (b).
S5: according to the communication service module, the digital twin database and the simulation data are connected into a communication network, and the communication module specifically comprises: the system comprises a data packaging unit, a communication architecture unit, a data decoding unit and a transmission scheduling unit, wherein the data packaging unit, the communication architecture unit, the data decoding unit and the transmission scheduling unit are used for transmitting the data to a visualization platform, displaying a data visualization line graph and displaying optimal parameters.
S6: in actual production, an operator can issue a control instruction according to the comparison of the displayed optimal parameter and the current real-time operation parameter and the data weight, and under the simulation, the corresponding production indexes are found in a historical database according to the optimal output parameter desulfurization rate 94.23% and the sulfuric acid conversion rate 98.9%, the liquid level in the furnace is controlled to be about 40 m and is most suitable, and then the control instruction of the target liquid level of 40 m is issued.
S7: the control command is connected to a feedback control system built by the simulink, the control system completes process parameter regulation according to continuous feedback comparison of an input set value and a current value, and the whole liquid level regulation process is shown in fig. 4 (c) by taking a liquid level as an example.
S8: the communication service module is connected with a valve, a regulator and a controller under actual production working conditions to guide actual production; and according to the actual production needs, the real-time data acquisition, digital twin simulation, visual chart updating and virtual control are continuously repeated, and the desulfurization efficiency and the maximization of the sulfuric acid preparation speed are ensured for a long time.
Examples
A collaborative optimization system based on digital twin for recycling tail gas in copper smelting process is shown in figure 2, and comprises a database module, a data analysis module, a visualization module, an equipment management control module and a communication service module.
The database module is used for collecting real-time data in the operation process of desulfurizing and producing acid in the smelting process and acquiring data information of the tower body in the whole process flow; the system comprises a geometric parameter data unit, an operation state parameter unit, a real-time process parameter unit and a neural network model unit.
The geometric parameter data unit in this embodiment contains field device data of actual production: the method comprises the steps of producing relevant sizes, heights, production parameters and the like of a furnace body, and storing and retrieving long-term operation data;
the operation state parameter unit in the embodiment contains operation data of production smelting, including relevant information such as temperature, pressure, liquid level and the like in a production furnace body, and describes the state and performance of a system or equipment in operation; and provides information about the health status, work efficiency, resource utilization, etc. of the whole system or equipment, and stores and retrieves long-term operation data.
The real-time process parameter unit in this embodiment contains an input quantity and a production output quantity for updating the pipeline input quantity and the production output quantity for entering the reaction in real time, and is used for describing and evaluating the current state and performance of the system, the equipment or the process, providing information about the health condition, the working efficiency, the resource utilization condition and the like of the whole system or the equipment, and storing and retrieving long-term operation data.
In the neural network model unit, the log record of each operation in the neural network model is performed under the long-term operation state, and the weight of the relevant parameter, the iteration number and the output-input ratio of each operation are recorded.
The data analysis module in the embodiment obtains state data from the database module, performs pretreatment and analysis on the data, and obtains an optimal parameter predicted value through a digital twin model, thereby achieving the aim of resource recovery collaborative optimization in the desulfurization and acid production process; the method comprises an algorithm library unit, a data preprocessing unit, a feature extraction unit and a data mining unit.
The algorithm library unit is used for carrying out secondary smoothing on the output of the BP model, improving prediction reliability by utilizing the strong nonlinear mapping capability and flexible network structure, fitting large sample data such as copper smelting desulfurization and acid making process better, and solving the optimal parameters of the data twin model by combining a global optimal search method.
According to the data preprocessing unit, the optimal parameters in the historical data are changed, so that data missing frequently occurring in real-time data acquisition and large-deviation noise are preprocessed in advance, and repeated and large-scale identical optimization of equipment data is avoided.
According to the feature extraction unit, through a BP neural network model, according to five state parameters in a database as input variables of the BP natural network model, the conversion rate and the desulfurization efficiency of sulfuric acid are adopted as an output layer, and the hidden layer is set to conduct predictive training on data and verify sample classification.
The data mining unit in this embodiment finds hidden patterns, association rules and trends in long-term production data, performs advanced data analysis before feature extraction, and mines the overall features of the obtained production input information, so as to identify trends, abnormal patterns and correlations in the data.
The visualization module in this embodiment is mainly used for graphically displaying the data of the two modules; the method comprises a fusion model unit, a result transmission unit, a data imaging unit and a graphic dynamic display unit.
In the fusion model unit model fusion of the embodiment, the prediction results of a plurality of data analysis modules are combined to obtain more accurate and reliable prediction; the performance of the whole system is improved by combining the prediction results obtained by different data and different dates in the model, so that higher accuracy and robustness are obtained.
The result transmission unit in this embodiment is configured to perform timing retrieval on the optimal parameters obtained by the data analysis module, update the obtained optimal parameters in real time, and transmit the output results to the data analysis system at the back end in a suitable form, so as to support decision making, provide insight, drive a subsequent visualization process, and perform further analysis.
The data imaging unit in this embodiment performs visual graphic display on the result optimal parameter data obtained by the result transmission unit, and uses such as bar graphs, line graphs, scatter graphs, pie charts, and the like to convey relationships, trends, and modes between data of output parameters, process parameters, and the like.
The graphic dynamic display unit mainly drives the whole module to continuously call real-time data to the data analysis module and conduct graphic updating of data visualization, so that the graphic can be updated along with real-time or frequent changes of the data, and latest information and trends can be presented timely.
The intelligent management module of the device is used for combining the module results, and comprises a database module for realizing the full life cycle information acquisition of the device, completing the real-time intelligent optimization operation and maintenance of the device, and carrying out real-time and efficient control on the state parameters of actual production under the output of a control system according to the optimal operation parameters; the system comprises an evaluation index unit, a control module unit, an intelligent decision unit and a control instruction unit.
The evaluation index unit mainly compares the obtained optimal operation parameters with real-time operation parameters, obtains parameters which need to be weighted according to the evaluation indexes, and needs to have high weight for regulation and control under different evaluation indexes, so as to measure the accuracy, performance and reliability of the model, and help staff to know the performance of the model and make decisions.
The control module unit in the embodiment performs the liquid level-flow cascade control system of gas-liquid inlet and outlet according to the change value obtained by comparison, and performs the decoupling control of the liquid level-pressure cascade control system in the tower and the dual-input and dual-output grading of the temperature in the tower, and completes the control from digital twin simulation to actual production under the regulation of feedback.
The intelligent decision unit in this embodiment uses the existing data and information of the data analysis module as a basis, and identifies the mode, trend and rule by analyzing, mining and integrating the obtained optimal parameters and the on-site actual production data, thereby providing a targeted and reliable production decision suggestion.
The control instruction unit in this embodiment is a key component responsible for generating, verifying and sending the control instruction after intelligent decision after the intelligent decision unit, monitoring the system state and receiving feedback; and the operator sends a command to the system to regulate and control valves, regulators and controllers for actual production control parameters in actual production, and issues control instructions related to valve opening and controller opening adjustment to guide the final target copper smelting tail gas desulfurization and acid production resource recovery to be cooperatively optimized.
The communication service module mainly realizes the mutual association among the four modules, data in the database module is called to the data analysis module in real time, the result obtained by the data analysis module is sent to the visualization module, and finally the result is applied to the intelligent management module of the equipment according to the data visualization result to complete information calling and association among the modules; various types of messages including production data text, live images, etc.; the real-time communication function is provided to complete the image real-time updating function in the visualization module, ensure the correctness, stability and safety of communication, ensure the correct instruction of the intelligent operation and maintenance module of the equipment to be issued, and complete regulation and control; and records the running log of each module for recording the platform so as to be convenient for subsequent analysis and debugging.

Claims (5)

1. A collaborative optimization method for recycling tail gas in a copper smelting process based on digital twinning is characterized by comprising the following steps of: the method comprises the following steps:
step1: the method comprises the steps of calling state parameters, process parameters and evaluation indexes of desulfurization and acid production in a smelting process from a conventional database, preprocessing data, predicting the data and mining the data to construct the conventional database;
step2: collecting real-time data, including state parameters and process parameters, which together form a real-time database; the state parameters comprise temperature, pressure, liquid level and gas-liquid ratio values in the furnace body; the process parameters comprise the concentration and flow rate of input sulfur dioxide and hydrogen peroxide, and the concentration and flow rate of output sulfuric acid;
Step3: the method comprises the steps that a digital twin dynamic model database is formed by a traditional database and a real-time database, data intercommunication connection between the digital twin dynamic model database and a front-end interface formed by a server and a unit 3D is realized through a TCP communication protocol, and a visual graph of relevant data in the digital twin dynamic database is displayed on the front-end interface;
step4: introducing the obtained digital twin dynamic model database in Step3 as a digital twin model simulation signal into a digital twin model for desulfurizing and producing acid, wherein the model carries out simulation through data, and the simulation model is built by adopting nonlinear optimization based on a two-stage BPNLP neural network and combining an approximate global optimization algorithm and is used for continuously optimizing process parameters corresponding to optimal parameters obtained through the digital twin model, and obtaining the most suitable production optimal engineering parameters according to the maximized parameters of sulfur dioxide desulfurization rate and sulfuric acid conversion rate;
step5: carrying out data mining similar aggregation on the predicted process parameters in the state of maximum sulfur dioxide desulfurization rate and maximum sulfuric acid conversion rate and a conventional database, carrying out noise reduction and missing value compensation treatment, and then carrying out parameter comparison on the predicted process parameters and a real-time database to judge which parameters of the conventional real-time data need to be improved; finally, an optimal parameter improvement scheme is obtained through optimizing a global algorithm under real-time data;
Step6: the optimal parameter data improvement scheme is visualized through a TCP protocol, and a visualized graph of the data is displayed on a human-computer interface of the unit 3D front end;
step7: in the optimal parameter improvement scheme, an operator issues a control instruction according to the obtained parameter improvement scheme and manual experience;
step8: the control instruction is connected with a desulfurization and acidification simulation control model of the simulink through a TCP protocol to form a control system, the remote end of the control system is connected with an actual industrial server, an optimal parameter value is compared with a real-time data value, parameter control is carried out on an actual furnace body according to a difference value of the optimal parameter value, and a valve and a regulator in production are controlled to realize guidance on actual production;
step9: continuously repeating the steps from Step2 to Step5, collecting real-time data, obtaining optimal control parameters through a simulated neural network digital twin model, comparing the two data, further feeding back and controlling furnace body process parameters, continuously repeating, and permanently ensuring the maximization of desulfurization efficiency and sulfuric acid preparation speed;
the simulation model in Step4 is built by adopting nonlinear optimization based on a two-stage BPNLP neural network and combining an approximate global optimization algorithm, and specifically comprises the following steps:
S1: processing data in the data model base, and prescribing input samples transmitted from the data base; according to the production process of desulfurizing and preparing acid in copper smelting, under the condition that an input sample is subjected to normal distribution, selecting a mean value and a standard deviation sigma in the sample;
s2: the BP neural network model belongs to an error back propagation neural network, and needs to be constructed in three aspects: an input layer, a hidden layer and an output layer; the model contains neurons, weights, thresholds, layers, and activation functions; establishing a BP neural network model according to real-time state parameters of the desulfurization and acid making process and the recycling conversion rate of pollutant sulfur;
s3: firstly, establishing parameters of an input layer, a hidden layer and an output layer, taking five state parameters in a database as input variables of a BP natural network model, namely pressure, temperature in a tower, liquid level, gas-liquid ratio value and flow of sulfur gas, introducing the parameters of the input layer, the hidden layer and the hidden layer between the secondary desulfurization process smelting parameters and the sulfuric acid collaborative optimization into two layers on the basis of calculation of the hidden layer of the BP neural network model, and setting nodes according to an actual process;
s4: the transfer function for the two hidden layers can be defined as
x i ＝(x i 1 ，x i 2 ，...，x i p )，i＝1，…，N
y i ＝(y i 1 ，y i 2 ，...，y i q )，i＝1，…，N
Where p and q are the dimensions of the input and output variables and N is the number of samples; i is the hidden layer of the neuron, i=1 represents the input layer, x p I.e. as input variable, y p Then it represents the output variable parameter and is specific to x p Each type of input in (a)The number of samples of the variable is defined asEach sample was calculated as follows:
since i is the hidden layer of neurons, each sample needs to calculate its output valueThe method comprises the following steps:
assume that for the kth layer there are s neurons; l (l=1, 2,....h.) is the number of layers of the hidden layer; wherein the method comprises the steps ofRepresenting the weight corresponding to each input variable, wherein θ is a bias value, and +.>The offset value of the 1 st hidden layer of the k layer is the offset value, and p is a fixed value to represent the number of samples; />Sample nodes that are input variables at the k-th layer;
also, toAs input to the model, the output +.>
To approximate the result to the target, outputThe error with the actual output is reversely transmitted from the output layer to the next layer of the network;the bias value of the ith hidden layer in each transmission is represented, and the threshold value and the weight value of the neuron are adjusted once in each transmission; this recursive process is repeated until the kth hidden layer, so that the function value of the output layer can be obtained +. >
S5: performing secondary smoothing on the model output; the generalization precision of the model before non-smoothing is defined as the proportion of singular values in all sample predicted values; so that the generalization accuracy can be set to an acceptable range;
s6: setting the output function of the BP neural network model as the maximum conversion rate and the maximum desulfurization efficiency of sulfuric acid in smelting flue gas, and expressing the adopted pollution emission reduction optimized BPNLP neural network model as the following mathematical function:
j represents an index function for representing a function index of an optimal desulfurization conversion rate and rate to be obtained, and a function Min represents a map of an output from the two-stage BPNLP estimation result, wherein an input variable x is a desired index; wherein c is a linear transformation vector, which is a 2×1 vector of the conversion rate and desulfurization efficiency of sulfuric acid;representing the estimated result of BP model prediction, c T Representing a desired index from an output estimation result of a BPNLP model with input variablesLinear transformation;
to minimize emissions problems, the nonlinear programming is as follows:
Ax≤b；
f(x)≤0；
L≤x≤U
wherein b, L, U are constant column vectors, f is a nonlinear function, x is a decision variable vector which is a 5 x 1 vector comprising pressure, temperature in the tower, liquid level, gas-liquid ratio value and flow of sulfur dioxide; a is a constant matrix, and the linear cut-off function Ax represents data after passing through a normalized domain and eliminating anomalies;
S7: by fitting the function N (x), x can generate multiple local minima, yielding multiple solutions and further affecting global optimization; selecting an initial value by adopting a random scattered point method so as to adapt to the equal step change of initial value optimization; confirming a minimum value of the model simulation selected from the optimization results in the optimal values; for the whole neural network model, the region of the variable to be optimized is gridded by using a grid and sample method, and the global optimal solution is found by combining the local search skills and evolving the solution process in the framework;
s8: smoothing the output function; assuming that the output sample takes a certain dimension value of the output value, then the comparison of the model needs to be further expanded to simulate the output value, so that the output function smoothly takes the value in the domain, and unreasonable values are filtered out;
s9: and finally, completing the solution of the optimal parameters of the data twin model, and transmitting the obtained result to follow-up content of completing human-computer interaction in the unit 3D in a data visualization way.
2. The utility model provides a collaborative optimization system based on digit twin tail gas recycle in copper smelting process which characterized in that: the system comprises a database module, a data analysis module, a visualization module, a device management control module and a communication service module;
The database module is used for acquiring real-time data in the operation process of desulfurizing and producing acid in the smelting process and acquiring data information of the tower body in the whole process flow; the system comprises a geometric parameter data unit, an operation state parameter unit, a real-time process parameter unit and a neural network model unit;
the data analysis module obtains state data from the database module, performs pretreatment and analysis on the data, and obtains an optimal parameter predicted value through a digital twin model, thereby achieving the aim of resource recovery collaborative optimization in the desulfurization and acid production process; the device comprises an algorithm library unit, a data preprocessing unit, a feature extraction unit and a data mining unit;
the algorithm library unit is used for carrying out secondary smoothing on the output of the BP model, improving prediction reliability by utilizing the nonlinear mapping capability and the network structure of the algorithm library unit, fitting large sample data such as copper smelting desulfurization and acid making technology better, and completing the solution of the optimal parameters of the data twin model by combining a global optimal searching method;
the data preprocessing unit preprocesses data missing and large-deviation noise frequently occurring in real-time data acquisition in advance through changing optimal parameters in historical data, so that repeated and large-scale identical optimization of equipment data is avoided;
The feature extraction unit predicts and trains data and verifies sample classification through a BP neural network model according to five state parameters in a database as input variables of the BP natural network model, wherein an output layer is the conversion rate and desulfurization efficiency of sulfuric acid, and a hidden layer is arranged;
the data mining unit finds hidden modes, association rules and trends in long-term production data, performs advanced data analysis before feature extraction, and mines the overall features of the obtained production input information, so as to identify trends, abnormal modes and correlations in the data;
the visualization module is mainly used for graphically displaying the data of the two modules; the system comprises a fusion model unit, a result transmission unit, a data graphical unit and a graphical dynamic display unit;
the intelligent equipment management module is used for combining the module results, comprises the database module, realizes the full life cycle information acquisition of equipment, completes the real-time intelligent optimization operation and maintenance of the equipment, and carries out real-time and efficient control on the state parameters of actual production under the output of the control system according to the optimal operation parameters; the system comprises an evaluation index unit, a control module unit, an intelligent decision unit and a control instruction unit;
The communication service module mainly realizes the mutual association among the four modules, data in the database module is called to the data analysis module in real time, the result obtained by the data analysis module is sent to the visualization module, and finally, the communication service module is applied to the intelligent management module of the equipment according to the result of the data visualization to complete information calling and association among the modules.
3. The collaborative optimization system based on digital twin in tail gas recycling in copper smelting process according to claim 2, wherein the collaborative optimization system is characterized in that:
the geometric parameter data unit contains actual production field device data: the method comprises the steps of producing relevant size, height and production parameters of a furnace body, and storing and retrieving long-term operation data;
the operation state parameter unit contains operation data of production smelting, including temperature, pressure and liquid level related information in the production furnace body, and describes the state and performance of the system or equipment in operation; information about the health condition, working efficiency and resource utilization condition of the whole system or equipment is provided, and long-term operation data are stored and retrieved;
the real-time process parameter unit contains pipeline addition input quantity and production output quantity which are updated and enter the reaction in real time, is used for describing and evaluating the current state and performance of a system, equipment or a process, provides information about the health condition, working efficiency and resource utilization condition of the whole system or the equipment, and stores and retrieves long-term operation data;
And the neural network model unit records the operation log of each time in the neural network model under the long-term operation state, and records the related parameter weight, the iteration number and the output/input ratio of each time.
4. The collaborative optimization system based on digital twin in tail gas recycling in copper smelting process according to claim 2, wherein the collaborative optimization system is characterized in that:
the fusion model unit model fusion combines the prediction results of a plurality of data analysis modules to obtain more accurate and reliable prediction; the performance of the whole system is improved by combining the prediction results obtained by different data and different dates in the model, so that higher accuracy and robustness are obtained;
the result transmission unit is used for carrying out timing retrieval on the optimal parameters obtained by the data analysis module, updating the obtained optimal parameters in real time, and transmitting the output results to a data analysis system at the rear end in a proper form so as to support decision making, provide insight, drive a subsequent visual flow and carry out further analysis;
the data imaging unit is used for carrying out visual graph display on the result optimal parameter data obtained by the result transmission unit, and utilizing the bar graph, the line graph, the scatter graph and the pie graph to convey the relation, the trend and the mode among the data of the output parameters and the process parameters;
The graphic dynamic display unit mainly drives the whole module to continuously call real-time data to the data analysis module and conduct graphic updating of data visualization, so that the graphic can be updated along with real-time or frequent changes of the data, and latest information and trends can be timely presented.
5. The collaborative optimization system based on digital twin in tail gas recycling in copper smelting process according to claim 2, wherein the collaborative optimization system is characterized in that:
the evaluation index unit is mainly used for comparing the obtained optimal operation parameters with real-time operation parameters, obtaining parameters which need to be weighted according to the evaluation indexes, and adjusting and controlling the parameters under different evaluation indexes with high weight, so as to measure the accuracy, performance and reliability of the model, and help staff to know the performance of the model and make decisions;
the control module unit is used for carrying out decoupling control of the liquid level-flow cascade control system of gas-liquid inlet and outlet, the liquid level-pressure cascade control system in the tower and the dual-input and dual-output grading of the temperature in the tower according to the change values obtained by comparison, and completing control from digital twin simulation to actual production under the regulation of feedback;
the intelligent decision unit is used for identifying modes, trends and rules by utilizing the existing data and information of the data analysis module as a basis and analyzing, excavating and integrating the obtained optimal parameters and the on-site actual production data, so that targeted and reliable production decision suggestions are provided;
The control instruction unit is responsible for generating, verifying and sending the control instruction after intelligent decision after the intelligent decision unit, and monitoring the system state and receiving the key components of feedback; and the operator sends a command to the system to regulate and control valves, regulators and controllers for actual production control parameters in actual production, and issues control instructions related to valve opening and controller opening adjustment to guide the final target copper smelting tail gas desulfurization and acid production resource recovery to be cooperatively optimized.
CN202311094435.2A 2023-08-29 2023-08-29 Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin Active CN116822380B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311094435.2A CN116822380B (en) 2023-08-29 2023-08-29 Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311094435.2A CN116822380B (en) 2023-08-29 2023-08-29 Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin

Publications (2)

Publication Number Publication Date
CN116822380A CN116822380A (en) 2023-09-29
CN116822380B true CN116822380B (en) 2023-11-14

Family

ID=88117025

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311094435.2A Active CN116822380B (en) 2023-08-29 2023-08-29 Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin

Country Status (1)

Country Link
CN (1) CN116822380B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117150819B (en) * 2023-10-30 2024-01-30 天津市东瑞钢铁铸造股份有限公司 Molten steel filtering method and system for steel casting
CN117787507B (en) * 2024-02-23 2024-05-03 宝鸡核力材料科技有限公司 Full chain optimizing method and device for tape rolling process

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110045608A (en) * 2019-04-02 2019-07-23 太原理工大学 Based on the twin mechanical equipment component structural dynamic state of parameters optimization method of number
CN112427624A (en) * 2020-11-04 2021-03-02 中国科学院重庆绿色智能技术研究院 Casting and forging double-control forming system based on digital twinning and parameter optimization method
CN112926765A (en) * 2021-01-22 2021-06-08 湖南大唐先一科技有限公司 Desulfurization system operation optimization method and information physical fusion system
CN113192568A (en) * 2021-03-15 2021-07-30 山东钢铁股份有限公司 Refining furnace desulfurization terminal point forecasting method and system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110045608A (en) * 2019-04-02 2019-07-23 太原理工大学 Based on the twin mechanical equipment component structural dynamic state of parameters optimization method of number
CN112427624A (en) * 2020-11-04 2021-03-02 中国科学院重庆绿色智能技术研究院 Casting and forging double-control forming system based on digital twinning and parameter optimization method
CN112926765A (en) * 2021-01-22 2021-06-08 湖南大唐先一科技有限公司 Desulfurization system operation optimization method and information physical fusion system
CN113192568A (en) * 2021-03-15 2021-07-30 山东钢铁股份有限公司 Refining furnace desulfurization terminal point forecasting method and system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Synergic optimization of pollution prevention and resource recovery of secondary lead smelting industry based on two-stage BPNLP network model;Yanping Li et al.;Journal of Cleaner Production;1-10 *

Also Published As

Publication number Publication date
CN116822380A (en) 2023-09-29

Similar Documents

Publication Publication Date Title
CN116822380B (en) Collaborative optimization method for tail gas recycling in copper smelting process based on digital twin
Xu et al. Mid-term prediction of electrical energy consumption for crude oil pipelines using a hybrid algorithm of support vector machine and genetic algorithm
CN105302096B (en) Intelligent factory scheduling method
CN101630376A (en) Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process
CN111651931A (en) Blast furnace fault diagnosis rule derivation method based on deep neural network
US20220365521A1 (en) Virtual simulation manufacturing platform based on automatic control
CN102231057A (en) Method for carrying out soft-sensing on lysine fermenting process on basis of chaos particle swarm optimization (CPSO)
CN110410819B (en) Coal-fired boiler combustion power field rapid calculation and real-time display system
CN103942422B (en) Granular-computation-based long-term prediction method for converter gas holder positions in metallurgy industry
CN102830625A (en) Process control system and method based on neural network predictive control
CN104651559A (en) Blast furnace liquid iron quality online forecasting system and method based on multivariable online sequential extreme learning machine
CN105425583A (en) Control method of penicillin production process based on cooperative training local weighted partial least squares (LWPLS)
Hu et al. Multi-model ensemble prediction model for carbon efficiency with application to iron ore sintering process
Wu et al. Integrated soft sensing of coke-oven temperature
CN103778466B (en) Converter steel-making temperature modeling forecasting method based on vector error and system thereof
CN103279030A (en) Bayesian framework-based dynamic soft measurement modeling method and device
CN116755409B (en) Coal-fired power generation system coordination control method based on value distribution DDPG algorithm
CN117034808A (en) Natural gas pipe network pressure estimation method based on graph attention network
CN102385706A (en) Soft sensing method based on model learning and used in pure terephthalic acid (PTA) production
Chen et al. Modeling of goethite iron precipitation process based on time-delay fuzzy gray cognitive network
Wang et al. Multi-layer encoding genetic algorithm-based granular fuzzy inference for blast furnace gas scheduling
CN111650894A (en) Bayesian network complex industrial process soft measurement method based on hidden variables
CN117930786B (en) Intelligent digital twin simulation system for steel production process
Zied A Contribution to Chemical Process Operation Support: New Machine Learning and Surrogate Models-Based Approaches for Process Optimization, Supervision and Control
Meng et al. Automatic control method of automobile steering-by-wire based on fuzzy PID

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant