CN111160518B - Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium - Google Patents
Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium Download PDFInfo
- Publication number
- CN111160518B CN111160518B CN201811327591.8A CN201811327591A CN111160518B CN 111160518 B CN111160518 B CN 111160518B CN 201811327591 A CN201811327591 A CN 201811327591A CN 111160518 B CN111160518 B CN 111160518B
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- layer
- cod
- node
- 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
Links
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2219/00—Indexing scheme relating to application aspects of data processing equipment or methods
- G06F2219/10—Environmental application, e.g. waste reduction, pollution control, compliance with environmental legislation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The embodiment of the invention provides a soft measurement method and device for COD (chemical oxygen demand) of refining sewage and a machine-readable storage medium, belonging to the technical field of environmental protection. The soft measurement method comprises the following steps: obtaining measured values and COD sample data for auxiliary variables for training, wherein the auxiliary variables include the following parameters: volatile organic compounds, temperature, pH, and turbidity; acquiring a measured value of an auxiliary variable for measurement; training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data; and determining a soft measurement result of the COD of the sewage based on the neural network model after training and the obtained measurement value of the auxiliary variable for measurement. Therefore, soft measurement of the COD of the sewage is realized, real-time estimation of the COD of the sewage is realized, and the prediction effect of the COD of the sewage is improved.
Description
Technical Field
The invention relates to the technical field of environmental protection, in particular to a soft measurement method and device for COD (chemical oxygen demand) of refining sewage and a machine-readable storage medium.
Background
Chemical Oxygen Demand (COD) is an important index for measuring the water quality of sewage, and can comprehensively characterize the content of organic matters in the sewage. The current method for measuring COD in national standard GB11914-89 is a potassium dichromate reflux method: the method comprises the steps of masking chloride ions by utilizing mercury sulfate, taking potassium dichromate as an oxidant and silver sulfate as a catalyst in the environment of acidic medium sulfuric acid, carrying out high-temperature digestion, oxidizing reducing substances in water by the potassium dichromate, determining the residual potassium dichromate by utilizing a titration method or a colorimetric method, and finally calculating the COD value based on a differential method. However, this method has problems of labor input, time and effort consuming and high reagent consumption. With the rise of automation technology, a series of automatic water quality analysis instruments are developed for municipal sewage by a plurality of domestic and foreign companies, and the principle of the potassium dichromate method is also based.
Compared with municipal sewage, the sewage treatment water has the characteristics of complex sewage source, multiple pollutant types, complex components, high toxicity, serious harm and the like, and the incoming water is often required to be analyzed so as to adjust the process parameters later. However, when analyzing the COD of the incoming water of the refining wastewater, the following difficulties are often encountered: the problems of high labor cost, longer time, reagent consumption and secondary pollution exist when the national standard method is used. When water quality analysis instruments are used for measuring inflow water, the inflow water is often influenced by dirty oil and other corrosive pollutants, so that the instruments are rapidly damaged and aged, and therefore, on-line COD measurement instruments are rarely installed at the inflow water position in oil refining chemical enterprises.
In recent years, soft measurement methods are often used in municipal sewage to solve the problem of real-time estimation of COD. The basic core is that a data-driven mode is adopted, the quantity easy to measure is selected, and the quantity difficult to measure in real time is estimated through a construction algorithm. Two core problems of soft measurement are the selection of auxiliary variables and the construction of algorithms. The invention relates to a sewage COD soft measurement method based on an output observer (the authorized bulletin number is CN 103399134B), which selects TOC, DO, ORP, pH, T, HRT and r as auxiliary variables. Wherein TOC is difficult to measure in real time and ORP and COD are less relevant for refinery wastewater. The soft measurement model in the patent is designed based on the observation variables in the linear control system model, the model of the type needs to have linear assumption conditions, and the relationship between COD and other observation variables in reality is extremely complex and has stronger nonlinear characteristics, so the model does not accord with the actual situation of COD soft measurement; in addition, the model can optimize the quantity of the observers to be selected to be too small, so that the optimized model has poor measurement effect.
Disclosure of Invention
The object of the present invention is to provide a soft measurement method and device for COD of sewage and a machine readable storage medium, which solve or at least partially solve the above problems.
In order to achieve the above object, an aspect of the present invention provides a soft measurement method of COD of sewage from refining, the soft measurement method comprising: obtaining measured values and COD sample data for auxiliary variables for training, wherein the auxiliary variables include the following parameters: volatile organic compounds, temperature, pH, and turbidity; acquiring a measured value of an auxiliary variable for measurement; training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data; and determining a soft measurement result of the COD of the sewage based on the neural network model after training and the obtained measurement value of the auxiliary variable for measurement.
Optionally, the training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data comprises: determining an output value of the neural network model based on the measured value of the auxiliary variable for training and the neural network model, wherein in the neural network model, an input value of each node is a linear combination of output values of all nodes of a layer above a layer on which the node is located, and the output value of each node is determined based on the input value of the node and an excitation function; and training the neural network model in the following way under the condition that the output value of the neural network model and the COD sample data do not meet the preset condition: constructing a cost function based on the output value of the neural network model and the COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network based on the cost function, and adjusting the coefficients according to the value of the function form after deriving; and/or adjusting the number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model to train the neural network model.
Optionally, the linear combination is: is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for the j-th node of the first layer of the neural network model, +.>Is the input of the j-th node of the first layer of the neural network model.
Optionally, the cost function is:wherein y is the COD sample data, a L Is the output value of the neural network model.
Optionally, the adjusting the coefficient includes: reducing the coefficient when the value of the function form after derivation is larger than 0; in the case that the value of the derived functional form is equal to 0, the coefficient is not changed; and increasing the coefficient in the case that the value of the derived functional form is less than 0.
Optionally, the measurement of volatile organic compounds includes a measurement of a photoionization detector and a measurement of a hydrogen flame ionization detector.
Optionally, the neural network model after training includes four layers, which are an input layer, two hidden layers, and an output layer, wherein the input layer includes 5 nodes, and the 5 nodes are a measured value of the photoionization detector, a measured value of the hydrogen flame ionization detector, a measured value of the temperature, a measured value of the pH, and a measured value of the turbidity, respectively, each of the two hidden layers includes 50 nodes, and the output layer includes 1 node.
Optionally, each node in the neural network model after training obtains an output of the node based on the sigmoid function and an input of the node.
Accordingly, another aspect of the present invention provides a soft measurement device for chemical wastewater COD, the soft measurement device comprising: a first acquisition module for acquiring measured values and COD sample data of auxiliary variables for training, wherein the auxiliary variables include the following parameters: volatile organic compounds, temperature, pH, and turbidity; a second acquisition module for acquiring a measured value of an auxiliary variable for measurement; a training module for training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data; and the determining module is used for determining a soft measurement result of the COD of the sewage based on the neural network model after training and the acquired measurement value of the auxiliary variable for measurement.
Optionally, the training module training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data comprises: determining an output value of the neural network model based on the measured value of the auxiliary variable for training and the neural network model, wherein in the neural network model, an input value of each node is a linear combination of output values of all nodes of a layer above a layer on which the node is located, and the output value of each node is determined based on the input value of the node and an excitation function; training the neural network model under the condition that the output value of the neural network model and the COD sample data do not meet the preset condition by the following steps: constructing a cost function based on the output value of the neural network model and the COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network based on the cost function, and adjusting the coefficients according to the value of the function form after deriving; and/or adjusting the number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model to train the neural network model.
Optionally, the linear combination is: is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for the j-th node of the first layer of the neural network model, +.>Is the input of the j-th node of the first layer of the neural network model.
Optionally, the cost function is:wherein y is the COD sample data, a L Is the output value of the neural network model.
Optionally, the adjusting the coefficient includes: reducing the coefficient when the value of the function form after derivation is larger than 0; in the case that the value of the derived functional form is equal to 0, the coefficient is not changed; and increasing the coefficient in the case that the value of the derived functional form is less than 0.
Optionally, the measurement of volatile organic compounds includes a measurement of a photoionization detector and a measurement of a hydrogen flame ionization detector.
Optionally, the neural network model after training includes four layers, which are an input layer, two hidden layers, and an output layer, wherein the input layer includes 5 nodes, and the 5 nodes are a measured value of the photoionization detector, a measured value of the hydrogen flame ionization detector, a measured value of the temperature, a measured value of the pH, and a measured value of the turbidity, respectively, each of the two hidden layers includes 50 nodes, and the output layer includes 1 node.
Optionally, each node in the neural network model after training obtains an output of the node based on the sigmoid function and an input of the node.
Still another aspect of the present invention provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform the above-described method.
Through the technical scheme, the neural network model is trained, and the soft measurement result of the COD of the sewage is determined based on the trained neural network model, so that the soft measurement of the COD of the sewage is realized, and the real-time estimation of the COD of the sewage is realized. In addition, the auxiliary variable is selected from volatile organic compounds, temperature, pH and turbidity, and has high correlation with the COD of the sewage, so that the prediction effect of the COD of the sewage is improved. And the selected auxiliary variable can be measured in real time, so that the soft measurement of the COD of the sewage is facilitated. In addition, a neural network model is selected, and the neural network model has little requirement on the correlation between the auxiliary variable and the COD of the sewage, so that the soft measurement method for the COD of the sewage better accords with the actual condition of the soft measurement of the COD of the sewage, and has wider application.
Additional features and advantages of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings are included to provide a further understanding of embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain, without limitation, the embodiments of the invention. In the drawings:
FIG. 1 is a flow chart of a soft measurement method of COD in sewage of refining provided by an embodiment of the invention;
FIG. 2 is a schematic diagram of data acquisition and processing according to another embodiment of the present invention;
FIG. 3 is a flow chart of training a neural network model provided by another embodiment of the present invention;
FIG. 4 is a schematic diagram of the structure of a neural network model;
FIG. 5 is a schematic diagram of the computation of inputs and outputs of nodes in a neural network model according to another embodiment of the present invention;
FIG. 6 is a flow chart for adjusting coefficients in a linear combination according to another embodiment of the present invention;
FIG. 7 is a schematic diagram of an error iterative evolution result according to another embodiment of the present invention;
FIG. 8 is a bar graph of error versus time provided by another embodiment of the present invention;
FIGS. 9 a-9 d are graphs showing the result of linear fitting of predicted values to measured values according to another embodiment of the present invention; and
Fig. 10 is a schematic structural diagram of a soft measurement device for COD of sewage for refining according to another embodiment of the present invention.
Description of the reference numerals
1. First acquisition Module 2 second acquisition Module
3. Training module 4 determination module
Detailed Description
The following describes the detailed implementation of the embodiments of the present invention with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
One aspect of the embodiment of the invention provides a soft measurement method for COD of sewage after refining. Fig. 1 is a flow chart of a soft measurement method for COD of sewage from refining according to an embodiment of the present invention. As shown in fig. 1, the method includes the following steps.
In step S10, measured values of auxiliary variables and COD sample data for training are acquired, wherein the auxiliary variables include the following parameters: volatile organics, temperature, pH, and turbidity.
Based on research, the content of Volatile Organic Compounds (VOC) in the sewage is internally related to COD, and the VOC can indirectly reflect the content of COD in the sewage to a certain extent. Compared with parameters such as oil content, ammonia nitrogen and the like, the VOC measurement is rapid and is not influenced by the pollution degree of polluted water. The VOC content is selected as one of the auxiliary variables. In addition, T and pH affect the volatility of the VOC and thus can be used as auxiliary variables. Turbidity reflects the oil and suspended matter content of the wastewater and is also selected as an auxiliary variable.
In step S11, a measured value of an auxiliary variable for measurement is acquired.
In step S12, a neural network model is trained based on the measured values of the auxiliary variables for training and the COD sample data.
In step S13, a soft measurement result of the chemical wastewater COD is determined based on the neural network model after training and the acquired measurement values of the auxiliary variables for measurement. That is, a neural network model with trained parameters is imported into a real-time monitoring system, auxiliary variables collected in real time are imported into the model as input, and predicted values of the model are calculated as soft measurement results.
Training a neural network model, and determining a soft measurement result of the COD of the sewage after training based on the neural network model, so that the soft measurement of the COD of the sewage is realized, and the real-time estimation of the COD of the sewage is realized. In addition, in the embodiment of the invention, the auxiliary variable is selected from volatile organic compounds, temperature, pH and turbidity, and has relatively high correlation with the COD of the sewage, so that the prediction effect of the COD of the sewage is improved. In addition, the auxiliary variable selected in the embodiment of the invention can be measured in real time, so that the soft measurement of COD of the refining sewage can be realized conveniently. In addition, a neural network model is selected in the embodiment of the invention, and the neural network model has little requirement on the correlation between the auxiliary variable and the COD of the sewage, so that the soft measurement method of the COD of the sewage, which is provided by the embodiment of the invention, is more suitable for the actual situation of the soft measurement of the COD of the sewage, and has wider application.
Optionally, in the embodiment of the present invention, a real-time testing device is set for the auxiliary variable, and the auxiliary variable collected by the real-time testing device is transmitted to the server through the data acquisition system. In addition, the collected data may be pre-processed before using the collected auxiliary variables for soft measurement of the COD of the sewage, the pre-processing comprising: and carrying out standard normalization processing on the acquired original data, checking whether the acquired original data is defective, if so, adopting variables corresponding to the defective data to acquire the data acquired in the upper and lower time periods of the time of the defective data, and processing the data as defective processing by utilizing the established judgment data if the data is wrong.
Specifically, the process of data acquisition and preprocessing can be seen in fig. 2. Setting up a real-time data acquisition platform, acquiring real-time data of selected auxiliary variables from a control system of the device or an enterprise real-time database by using an OPC DA server framework realized by VC++ programming according to a data acquisition interface (the sampling frequency in the real-time database can reach 1 time/minute or 2 times/minute, and the data frequency acquired from a DCS system can reach 1 time/second by using an OPC DA specification) of the refining device, refreshing a real-time data table of the server, and simultaneously storing the real-time data in a time sequence into a historical data table of each parameter. In addition, since the soft measurement method of the COD in the sewage is a supervised machine learning method, the training data is needed to measure the COD (for example, by manual actual sampling measurement), and the measurement value and the measurement time are recorded. Setting the sampling frequency of COD to be the same as that of the auxiliary variable, and aligning the COD sampling data at the same moment with the auxiliary variable data, thereby obtaining training data with labels, and correcting the constructed neural network model. Subsequently, data preprocessing is performed: and carrying out standard normalization processing on the collected original operation data, checking whether the data is missing, if so, adopting data in the upper and lower time periods of the time corresponding to the missing data by utilizing the variable corresponding to the missing data to carry out filtering processing, and judging whether the data is wrong or not by utilizing a set rule (for example, through manual input), if so, processing the data as the missing. The preprocessed data are grouped according to time, each group of data comprises auxiliary variables and COD measurement data at the same moment, and the data are randomly classified into 3 types and are respectively used for model training, model verification and model test. The data for model training is the measured value of the auxiliary variable for training in the embodiment of the invention, and the data for model testing is the measured value of the auxiliary variable for measuring in the embodiment of the invention. In the process of training the neural network model, the condition that the mean square error (the error between the output value of the neural network model and COD sample data) continuously oscillates is avoided, and a part of the data sample is reserved for model verification. And (3) optimizing the training adjustment stride through model verification, and finally optimizing the prediction capacity of the model. In addition, the verification data can be used for calculating the error between the output value of the neural network model and the COD sample data and performing linear fitting.
Fig. 3 is a flowchart of training a neural network model according to another embodiment of the present invention. As shown in fig. 3, training the neural network model based on the measured values of the auxiliary variables for training and COD sample data includes the following.
In step S30, an output value of the neural network model is determined based on the measured value of the auxiliary variable for training and the neural network model. Wherein the output value is the predicted value of COD of the smelting sewage. Furthermore, in the neural network model, the input value of each node is a linear combination of the output values of all nodes of a layer above the layer in which the node is located, and the output value of each node is determined based on the input value of the node and the excitation function.
The schematic structural diagram of the neural network model may be referred to as fig. 4, and it should be noted that the neural network shown in fig. 4 is only an example of the neural network, and is not intended to limit the present invention. In the neural network model shown in fig. 4, the neural network model includes 4 layers in total, wherein the first layer is an input layer, the fourth layer is an output layer, and the second layer and the third layer are hidden layers. For a node in each hidden layer, its input comes from the output of all nodes in the previous layer, and its output will become the input of all nodes in the next layer. In the embodiment of the invention, the number of layers of the neural network model and the number of nodes contained in each layer respectively can be set according to actual conditions.
Specifically, in the embodiment of the present invention, the neural network model may be pre-built according to the actual situation, and the auxiliary variables are set as the input variables, and in the embodiment of the present invention, there are at least four measured values of the auxiliary variables, so that the input layer of the pre-built neural network model includes at least 4 nodes. In addition, the output value of the neural network model represents the predicted value of the neural network model for the COD of the sewage, so that the nodes of the output layer in the pre-constructed neural network model at least comprise one node. Optionally, in the embodiment of the present invention, a node of an output layer of the pre-constructed neural network model includes one node, and an output of the node is a prediction result of COD. In addition, in the embodiment of the invention, the node number of each layer in the hidden layer of the pre-constructed neural network model can be the same or different. The total number of network layers and/or the number of nodes of each layer in the hidden layer in the pre-constructed neural network model can be used as adjustable parameters, and the prediction effect of the neural network model is optimized when the neural network model is trained. Optionally, in the embodiment of the present invention, the number of nodes in each layer in the hidden layer is n, and the total layer number and/or n of the neural network model is used as an adjustable parameter, and in the process of training the neural network model, the total layer number and/or n of the neural network model is adjusted to optimize the prediction effect of the neural network model.
Optionally, in an embodiment of the present invention, in the neural network model, the input value of each node is a linear combination of the output values of all nodes of a layer above the layer where the node is located, and the output value of each node is determined based on the input to and the excitation function of the node. Wherein, for the nodes of the input layer, the inputs of the nodes are obtained from the measured values of the auxiliary variables for training, and the measured values of the auxiliary variables for training are corresponding to the inputs of the nodes in the input layer according to the node conditions of the input layer of the constructed neural network model. The output of a node in the input layer is determined by the stimulus function.
Alternatively, in an embodiment of the present invention, the linear combination may be: is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for j-th node of first layer of neural network model, +.>Is the input of the j-th node of the first layer of the neural network model, as shown in fig. 5. Furthermore, the excitation function-> Is the output of the j-th node of the first layer of the neural network, as shown in fig. 5. Optionally, in an embodiment of the present invention, the excitation function is a sigmoid function.
In step S31, in the case where the output value of the neural network model and the COD sample data do not meet the preset condition, the neural network model is trained by: constructing a cost function based on the output value of the neural network model and COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network by using the cost function as a basis through error back propagation, and adjusting the coefficients according to the value of the function form after deriving; and/or adjusting the number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model to train the neural network model. Optionally, in an embodiment of the present invention, the preset condition includes at least one of: the mean square error between the output value of the neural network model and the COD sample data is smaller than or equal to a preset mean square error, the error between the output value of the neural network model and the COD sample data is smaller than or equal to a preset error, and the slope of the linear fit between the output value of the neural network model and the COD sample data is within a preset slope range. Optionally, the preset mean square error is 5. Alternatively, the preset error is 0.01. Optionally, the preset slope is in the range of 0.95-1. In addition, in the embodiment of the invention, the training process is ended when the training frequency of the neural network model reaches the preset frequency.
Alternatively, in the embodiment of the present invention, the quadratic cost function is constructed by comparing the output value of the neural network model (predicted value of COD of the sewage) with the COD sample data (measured value of COD of the sewage)Wherein y is COD sample data, a L Is the output value of the neural network model.
Alternatively, in an embodiment of the present invention, the adjustment coefficient may include the following. Reducing the coefficient under the condition that the value of the function form after derivation is larger than 0; under the condition that the value of the function form after derivation is equal to 0, the coefficient is not changed; when the value of the derived functional form is smaller than 0, the coefficient is increased. It should be noted that the content of the adjustment coefficient described herein is suitable for adjusting the coefficient in the linear combination corresponding to each node.
In addition, after the coefficients in the linear combination are adjusted, if the errors between the output value of the neural network model and the COD sample data still do not meet the preset conditions through verification of the neural network model, the number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model are adjusted. And then repeating the adjustment of the coefficients in the linear combination, the layer numbers in the neural network model and/or the nodes in at least one layer in the neural network model, and finally enabling the neural network model to meet preset conditions. In the embodiment of the present invention, when the coefficients in the linear combination, the number of layers of the neural network model, and/or the nodes in at least one layer of the neural network model are adjusted, there is no sequence, and the adjustment can be performed according to the actual situation, which is not intended to limit the present invention.
In addition, in the embodiment of the invention, the function form of the cost function after deriving the coefficients in the linear combination is determined layer by the following contents: in one layer of the neural network model, the coefficient component vector in the linear combination corresponding to all nodes of the layer and the output component vector in the upper layer of the layer are used for representing the input of the layer by adopting the vector composed of the coefficients and the vector composed of the output component, and the output vector of the layer is determined based on an excitation function; deriving the output vector in each layer by layer from the last layer (output layer), namely deriving the output vector of the last layer by the cost function, deriving the output vector in the previous layer of the last layer by using the function derived from the output vector of the last layer, and so on until the first layer (input layer); in one layer of the neural network model, coefficients in the linear combination corresponding to each node in the layer are respectively derived by using the function corresponding to the layer after deriving the output vector, the function after deriving is set to be 0, and a function form of the cost function after deriving the linear coefficients corresponding to each point in the layer is determined.
The following describes a function form of the layer-by-layer determination cost function after deriving coefficients in the linear combination with reference to fig. 6, and the method for training the neural network model used in the embodiment of the present invention is error back propagation. In the initial stage of model construction, coefficients in the linear combination are randomly set, in the process of training a neural network model, the coefficients are derived layer by layer from back to front in an error back propagation mode, the parameter sensitivity of a cost function to the coefficients in the linear combination is calculated, the coefficients are adjusted, the coefficients are continuously corrected, and the prediction function of the neural network model is trained. Wherein, in the embodiment shown in FIG. 6, the cost function employed is Wherein y is COD sample data, a L For the output values of the neural network model, the linear combination is: /> Is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for j-th node of first layer of neural network model, +.>Is the input of the j-th node of the first layer of the neural network model, as shown in fig. 5. Furthermore, the excitation function-> Is the output of the j-th node of the first layer of the neural network. In the constructed neural network model, besides the network layer number, the node number of each layer and the connection between the nodes, two network parameters of the input error of the nodes and the weight of the connection (i.e. the coefficient in the linear combination) also influence the network structure, thereby influencing the prediction function of the network. In the training process of the neural network model, the network prediction data is compared with the actual measurement data, and the errors are conducted reversely so as to adjust network parameters, so that the prediction capacity of the network is gradually improved, and the errors are reduced. A specific error feedback adjustment procedure is shown in fig. 6.
Initializing w l And b l L=1, 2, … … L, L is the total number of layers of the pre-built neural network model, w l In the first layerVectors of composition b l Is +.>Vector of components, inputA sample x is entered, which represents the measured value of the auxiliary variable and COD sample data. In the neural network model, based on the relationship between input and output set above, input and output of each layer are represented by vectors, z l =w l a l-1 +b l ,a l =σ(z l ) Wherein z is l And a l Representing the input and output in the first layer, respectively. After determining the output of the neural network model, a cost function is determined. At the last layer of the neural network model, the cost function is matched with the output vector z in the output layer L Deriving, obtaining a first value, delta, as shown in FIG. 6 L The first value is respectively +.>And->Deriving to obtain cost function of 0 for L layer>And->Substituting the sample data x into the derived functional form, judging the value of the derived functional form, and reducing the coefficient when the value of the derived functional form is greater than 0; when the value of the function form after derivation is equal to 0, the coefficient is not changed; when the value of the derived function is smaller than 0, the coefficient is reduced. In the case of obtaining layer L +.>And->After adjustment, the obtained first value is compared with the output vector z in the L-1 layer L-1 Deriving a second value and looking at layer L aboveAnd->The coefficients of the corresponding linear combinations of each node in the L-1 layer are adjusted in such a way that the second values are respectively corresponding to the +_ of the L-1 layer>And->Deriving, wherein the derived function is 0 to obtain the +.f cost function to the L-1 layer respectively>And->Deriving a functional form, carrying in sample data to obtain values of the functional form, and then respectively carrying out the comparison of the values of the functional form>And->And (5) adjusting. And then, the coefficients in the linear combinations corresponding to the nodes in the L-2 layer are adjusted by using the second value, and the coefficients are derived layer by layer until the layer is input. For example, taking layer l as an example, the output vector z to layer l l The derivative is in the form of a function delta as shown in figure 6 l Will delta l Coefficients in linear combinations corresponding to respective nodes in layer l (++>And->) Conducting derivation to make the derivative be 0, obtaining the cost functionRespectively pair->And->The derivative is in functional form as shown in fig. 6. When->Then decrease->When->Increase->When->Does not change->For input error->With the same operation, when->Then decrease->When->Increase->When->Does not change->
Alternatively, in an embodiment of the present invention, the measurement of the volatile organic compound includes a measurement of a photo ionization detector and a measurement of a hydrogen flame ionization detector. Currently, the on-line detector of VOC mainly includes a Photo Ionization Detector (PID) and a hydrogen Flame Ionization Detector (FID), and both of the detectors can perform rapid and sensitive detection. But the two detectors respond differently to organics. For a PID, the response values are ordered as follows: aromatic compounds and iodides >Paraffin, ketone, ether, amine, sulfide>Esters, aldehydes, alcohols, fats>Halogenated fat and ethane. For FID, the response values are ordered as follows: aromatic compounds and long-chain compounds>Short chain compounds (methane, etc)>Chlorine, bromine and iodine and their compounds. Different petrochemical sewage pollutants have different compositions, so that COD components are different, and thus, the volatile organic compounds have different components. Therefore, the two detectors can be used together to perform mutual verification in the VOC measurement of different petrochemical wastewater, and can be used as auxiliary variable for COD soft measurement of the refinery wastewater, and can be recorded as VOC P And VOC (volatile organic Compounds) F . Optionally, in an embodiment of the present invention, the input layer of the neural network model may be constructed to include 5 nodes, where the 5 nodes correspond to the measurement value of the photoionization detector, the measurement value of the hydrogen flame ionization detector, the measurement value of the temperature, the measurement value of the pH, and the measurement value of the turbidity, respectively.
Optionally, in the embodiment of the present invention, the neural network model after training includes four layers, which are an input layer, two hidden layers, and an output layer, wherein the input layer includes 5 nodes, and the 5 nodes are a measured value of the photoionization detector, a measured value of the hydrogen flame ionization detector, a measured value of the temperature, a measured value of the pH, and a measured value of the turbidity, respectively, each of the two hidden layers includes 50 nodes, and the output layer includes 1 node.
Optionally, in an embodiment of the present invention, each node in the neural network model after training obtains an output of the node based on the sigmoid function and an input of the node.
The soft measurement method of COD of the sewage is described below by combining an embodiment. In the embodiment, a neural network prediction model is established according to the data of monitoring and manually collecting the sewage treatment pool of a certain refining enterprise, and the model is trained and tested, so that the accuracy rate of the model is within an acceptable range.
Data sources: the input variables (auxiliary variables) include VOC of the sewage treatment tank P 、VOC F And T, pH and turbidity, data acquisition is carried out on the input variable, the sampling frequency is 2 times/hour, and the output data is COD content data of the sewage treatment tank. COD sample data for training and testing the neural network is obtained by manual sampling, the sampling frequency is 2 times/hour, and the time stamp of the COD sample data is aligned with the time stamp of the input data. The input data and the output data were collected for a total of 30 days, for a total of 4320 samples.
Model construction: the input layer of the neural network model is 5 nodes, corresponding to 5 input variables respectively, the middle layer is 2 layers, the number of nodes in each layer is 50, the output layer is 1 node, and corresponding to the output variables. The training mode of the network is a classical BP neural network, 70% of data samples are used for training a model, 15% of data samples are used for testing the model, and 15% of data samples are used for verifying the model.
Training results: after 101 iterations, the mean square error between the output value of the neural network model and the COD sample data always shows a descending trend, no oscillation condition occurs, and finally the neural network model is stabilized in an acceptable range, and the minimum mean square error in model verification can reach 4.19 (shown in fig. 7). The error concentration is distributed around the 0 value (see fig. 8). In fig. 9a-9d, the predicted value (output value of the neural network model) and the actual measured value (COD sample data) of the model are linearly fitted, the slope of the obtained straight line is between 0.994 and 0.996, which shows that the slope is approximately a straight line with 1, the fitting effect is good, and the prediction capability of the model is strong.
Accordingly, another aspect of the embodiments of the present invention provides a soft measurement device for COD of sewage. Fig. 10 is a schematic diagram showing a soft measurement device for COD of sewage from refining according to another embodiment of the present invention. As shown in fig. 10, the soft measuring device includes a first acquisition module 1, a second acquisition module 2, a training module 3, and a determination module 4. Wherein, the first acquisition module 1 is used for acquiring the measured value and COD sample data of the auxiliary variable used for training, wherein the auxiliary variable comprises the following parameters: volatile organic compounds, temperature, pH, and turbidity; the second acquisition module 2 is used for acquiring the measured value of the auxiliary variable for measurement; the training module 3 is used for training a neural network model based on the measured value of the auxiliary variable used for training and the COD sample data; the determining module 4 is used for determining soft measurement results of the COD of the sewage after training based on the neural network model and the obtained measurement values of the auxiliary variables for measurement.
Training a neural network model, and determining a soft measurement result of the COD of the sewage after training based on the neural network model, so that the soft measurement of the COD of the sewage is realized, and the real-time estimation of the COD of the sewage is realized. In addition, in the embodiment of the invention, the auxiliary variable is selected from volatile organic compounds, temperature, pH and turbidity, and has relatively high correlation with the COD of the sewage, so that the prediction effect of the COD of the sewage is improved. In addition, the auxiliary variable selected in the embodiment of the invention can be measured in real time, so that the soft measurement of COD of the refining sewage can be realized conveniently. In addition, a neural network model is selected in the embodiment of the invention, and the neural network model has little requirement on the correlation between the auxiliary variable and the COD of the sewage, so that the soft measurement method of the COD of the sewage, which is provided by the embodiment of the invention, is more suitable for the actual situation of the soft measurement of the COD of the sewage, and has wider application.
Optionally, in an embodiment of the present invention, the training module training the neural network model based on the measured values of the auxiliary variables for training and the COD sample data includes: determining an output value of a neural network model based on the measured value of the auxiliary variable for training and the neural network model, wherein in the neural network model, an input value of each node is a linear combination of output values of all nodes of a layer above the layer in which the node is located, and the output value of each node is determined based on the input value of the node and an excitation function; under the condition that the output value of the neural network model and the COD sample data do not meet the preset conditions, training the neural network model by the following modes: constructing a cost function based on the output value of the neural network model and COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network based on the cost function, and adjusting the coefficients according to the value of the function form after deriving; and/or adjusting the number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model to train the neural network model.
Alternatively, in an embodiment of the present invention, the linear combination is: is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for j-th node of first layer of neural network model, +.>Is the input of the j-th node of the first layer of the neural network model.
Optionally, in an embodiment of the present invention, the cost function is:wherein y is COD sample data, a L Is the output value of the neural network model.
Optionally, in an embodiment of the present invention, adjusting the coefficient includes: reducing the coefficient under the condition that the value of the function form after derivation is larger than 0; under the condition that the value of the function form after derivation is equal to 0, the coefficient is not changed; and increasing the coefficient when the value of the derived functional form is less than 0.
Alternatively, in an embodiment of the present invention, the measurement of the volatile organic compound includes a measurement of a photo ionization detector and a measurement of a hydrogen flame ionization detector.
Optionally, in an embodiment of the present invention, the neural network model after training includes four layers, which are an input layer, two hidden layers, and an output layer, wherein the input layer includes 5 nodes, and the 5 nodes are a measured value of the photoionization detector, a measured value of the hydrogen flame ionization detector, a measured value of the temperature, a measured value of the pH, and a measured value of the turbidity, respectively, each of the two hidden layers includes 50 nodes, and the output layer includes 1 node.
Optionally, in an embodiment of the present invention, each node in the neural network model after training obtains an output of the node based on the sigmoid function and an input of the node.
The specific working principle and benefits of the soft measurement device for COD in the sewage are similar to those of the soft measurement method for COD in the sewage, and will not be described here again.
Still another aspect of an embodiment of the present invention provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform the above-described method.
In summary, the neural network model is trained, and the soft measurement result of the COD of the sewage is determined based on the trained neural network model, so that the soft measurement of the COD of the sewage is realized, and the real-time estimation of the COD of the sewage is realized. In addition, in the embodiment of the invention, the auxiliary variable is selected from volatile organic compounds, temperature, pH and turbidity, and has relatively high correlation with the COD of the sewage, so that the prediction effect of the COD of the sewage is improved. In addition, the auxiliary variable selected in the embodiment of the invention can be measured in real time, so that the soft measurement of COD of the refining sewage can be realized conveniently. In addition, a neural network model is selected in the embodiment of the invention, and the neural network model has little requirement on the correlation between the auxiliary variable and the COD of the sewage, so that the soft measurement method of the COD of the sewage, which is provided by the embodiment of the invention, is more suitable for the actual situation of the soft measurement of the COD of the sewage, and has wider application.
The foregoing details of the optional implementation of the embodiment of the present application have been described in detail with reference to the accompanying drawings, but the embodiment of the present application is not limited to the specific details of the foregoing implementation, and various simple modifications may be made to the technical solution of the embodiment of the present application within the scope of the technical concept of the embodiment of the present application, and these simple modifications all fall within the protection scope of the embodiment of the present application.
In addition, the specific features described in the above embodiments may be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, various possible combinations of embodiments of the present application are not described in detail.
Those skilled in the art will appreciate that all or part of the steps in implementing the methods of the embodiments described above may be implemented by a program stored in a storage medium, including instructions for causing a single-chip microcomputer, chip or processor (processor) to perform all or part of the steps of the methods of the embodiments described herein. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In addition, any combination of various embodiments of the present invention may be performed, so long as the concept of the embodiments of the present invention is not violated, and the disclosure of the embodiments of the present invention should also be considered.
Claims (15)
1. A soft measurement method for COD of sewage after refining is characterized by comprising the following steps:
obtaining measured values and COD sample data for auxiliary variables for training, wherein the auxiliary variables include the following parameters: volatile organic compounds, temperature, pH, and turbidity;
acquiring a measured value of an auxiliary variable for measurement;
training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data; and
determining a soft measurement result of the COD of the sewage based on the trained neural network model and the obtained measurement value of the auxiliary variable for measurement;
wherein said training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data comprises:
determining an output value of the neural network model based on the measured value of the auxiliary variable for training and the neural network model, wherein in the neural network model, an input value of each node is a linear combination of output values of all nodes of a layer above a layer on which the node is located, and the output value of each node is determined based on the input value of the node and an excitation function; and
Training the neural network model under the condition that the output value of the neural network model and the COD sample data do not meet the preset condition by the following steps:
constructing a cost function based on the output value of the neural network model and the COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network based on the cost function, and adjusting the coefficients according to the value of the function form after deriving; and/or
The number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model are adjusted to train the neural network model.
2. The soft measurement method of claim 1, wherein the linear combination is: to be from and to the godWeights from kth node to jth node of l-1 layer of the network model, +.>Input error for the j-th node of the first layer of the neural network model, +.>Input for the j-th node of the first layer of the neural network model, +.>Is the output of the kth node of the first-1 layer of the neural network model.
3. The soft measurement method of claim 1, wherein the cost function is: Wherein y is the COD sample data, a L Is the output value of the neural network model.
4. The soft measurement method of claim 1, wherein the adjusting the coefficients comprises:
reducing the coefficient when the value of the function form after derivation is larger than 0;
in the case that the value of the derived functional form is equal to 0, the coefficient is not changed; and
and when the value of the function form after derivation is smaller than 0, the coefficient is increased.
5. The soft measurement method according to any one of claims 1 to 4, wherein the measurement value of the volatile organic compound includes a measurement value of a photoionization detector and a measurement value of a hydrogen flame ionization detector.
6. The soft measurement method of claim 5, wherein the trained neural network model comprises four layers, an input layer, two hidden layers, and an output layer, wherein the input layer comprises 5 nodes, the 5 nodes being the measurement of the photoionization detector, the measurement of the hydrogen flame ionization detector, the measurement of the temperature, the measurement of the pH, and the measurement of the turbidity, respectively, each of the two hidden layers comprising 50 nodes, and the output layer comprising 1 node.
7. The soft measurement method of claim 5, wherein each node in the neural network model after training obtains an output of the node based on a sigmoid function and an input of the node.
8. A soft measurement device for the COD of a sewage, the soft measurement device comprising:
a first acquisition module for acquiring measured values and COD sample data of auxiliary variables for training, wherein the auxiliary variables include the following parameters: volatile organic compounds, temperature, pH, and turbidity;
a second acquisition module for acquiring a measured value of an auxiliary variable for measurement;
a training module for training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data; and
the determining module is used for determining a soft measurement result of the COD of the refining sewage based on the neural network model after training and the acquired measurement value of the auxiliary variable for measurement;
wherein the training module training a neural network model based on the measured values of the auxiliary variables for training and the COD sample data comprises:
determining an output value of the neural network model based on the measured value of the auxiliary variable for training and the neural network model, wherein in the neural network model, an input value of each node is a linear combination of output values of all nodes of a layer above a layer on which the node is located, and the output value of each node is determined based on the input value of the node and an excitation function; and
Training the neural network model under the condition that the output value of the neural network model and the COD sample data do not meet the preset condition by the following steps:
constructing a cost function based on the output value of the neural network model and the COD sample data, determining a function form of the cost function after deriving coefficients in the linear combination layer by layer from the last layer of the neural network based on the cost function, and adjusting the coefficients according to the value of the function form after deriving; and/or
The number of layers of the neural network model and/or the number of nodes in at least one layer of the neural network model are adjusted to train the neural network model.
9. The soft measurement device of claim 8, wherein the linear combination is: is the weight from the kth node to the jth node of the l-1 layer of the neural network model, +.>Input error for the j-th node of the first layer of the neural network model, +.>Input for the j-th node of the first layer of the neural network model, +.>Is the output of the kth node of the first-1 layer of the neural network model.
10. The soft measurement device of claim 8, wherein the cost function is: Wherein y is the COD sample data, a I Is the output value of the neural network model.
11. The soft measurement device of claim 8, wherein the adjusting the coefficients comprises:
reducing the coefficient when the value of the function form after derivation is larger than 0;
in the case that the value of the derived functional form is equal to 0, the coefficient is not changed; and
and when the value of the function form after derivation is smaller than 0, the coefficient is increased.
12. The soft measurement device of any one of claims 8-11, wherein the measurement of volatile organic compounds comprises a measurement of a photoionization detector and a measurement of a hydrogen flame ionization detector.
13. The soft measurement device of claim 12, wherein the trained neural network model comprises four layers, an input layer, two hidden layers, and an output layer, wherein the input layer comprises 5 nodes, the 5 nodes being the measurement of the photoionization detector, the measurement of the hydrogen flame ionization detector, the measurement of the temperature, the measurement of the pH, and the measurement of the turbidity, respectively, each of the two hidden layers comprising 50 nodes, and the output layer comprising 1 node.
14. The soft measurement device of claim 12, wherein each node in the neural network model after training obtains an output of the node based on a sigmoid function and an input of the node.
15. A machine-readable storage medium having stored thereon instructions for causing a machine to perform the method of any of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811327591.8A CN111160518B (en) | 2018-11-08 | 2018-11-08 | Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811327591.8A CN111160518B (en) | 2018-11-08 | 2018-11-08 | Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111160518A CN111160518A (en) | 2020-05-15 |
CN111160518B true CN111160518B (en) | 2023-08-15 |
Family
ID=70555170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811327591.8A Active CN111160518B (en) | 2018-11-08 | 2018-11-08 | Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111160518B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10316381A1 (en) * | 2003-04-10 | 2004-10-28 | Bayer Technology Services Gmbh | Procedure for training neural networks |
CN102313796A (en) * | 2011-05-27 | 2012-01-11 | 北京工业大学 | Soft measuring method of biochemical oxygen demand in sewage treatment |
CN107132325A (en) * | 2017-04-14 | 2017-09-05 | 华南理工大学 | A kind of flexible measurement method of the Anaerobic Waste Treatment System water outlet volatile fatty acid based on particle cluster algorithm and SVMs |
-
2018
- 2018-11-08 CN CN201811327591.8A patent/CN111160518B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10316381A1 (en) * | 2003-04-10 | 2004-10-28 | Bayer Technology Services Gmbh | Procedure for training neural networks |
CN102313796A (en) * | 2011-05-27 | 2012-01-11 | 北京工业大学 | Soft measuring method of biochemical oxygen demand in sewage treatment |
CN107132325A (en) * | 2017-04-14 | 2017-09-05 | 华南理工大学 | A kind of flexible measurement method of the Anaerobic Waste Treatment System water outlet volatile fatty acid based on particle cluster algorithm and SVMs |
Non-Patent Citations (1)
Title |
---|
污水处理出水COD软测量预测建模方法的研究;曹为阳;《中国优秀硕士论文电子期刊网》;20180215;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN111160518A (en) | 2020-05-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Langergraber et al. | A guideline for simulation studies of wastewater treatment plants | |
Rieger et al. | Data Reconciliation for Wastewater Treatment Plant Simulation Studies—Planning for High‐Quality Data and Typical Sources of Errors | |
Corominas et al. | Performance evaluation of fault detection methods for wastewater treatment processes | |
US11346831B2 (en) | Intelligent detection method for biochemical oxygen demand based on a self-organizing recurrent RBF neural network | |
Rieger et al. | Progress in sensor technology-progress in process control? Part I: Sensor property investigation and classification | |
CN109828089B (en) | DBN-BP-based water quality parameter nitrous acid nitrogen online prediction method | |
CN110824914B (en) | Intelligent wastewater treatment monitoring method based on PCA-LSTM network | |
Torres et al. | Partial Least Squares local calibration of a UV–visible spectrometer used for in situ measurements of COD and TSS concentrations in urban drainage systems | |
CN109085805B (en) | Industrial process fault detection method based on multi-sampling-rate factor analysis model | |
Joseph et al. | Methodology for Bayesian belief network development to facilitate compliance with water quality regulations | |
CN107665288A (en) | A kind of water quality hard measurement Forecasting Methodology of COD | |
CN105550492B (en) | System and method for predicting the parameter of wastewater treatment process | |
Byun et al. | Data-driven optimization of reliability using buffered failure probability | |
CN111160518B (en) | Soft measurement method and device for COD of sewage refining and Chemical Oxygen Demand (COD) and machine-readable storage medium | |
CN109019717A (en) | A kind of heat-engine plant desulfurized waste water intelligent processing method and system | |
Park et al. | Application of transfer function ARIMA modeling for the sedimentation process on water treatment plant | |
Hansen et al. | Exploring data quality and seasonal variations of N2O in wastewater treatment: a modeling perspective | |
CN117233340A (en) | Method and system for detecting concentration of organic pollutants | |
Benedetti et al. | The incorporation of variability and uncertainty evaluations in WWTP design by means of stochastic dynamic modeling: the case of the Eindhoven WWTP upgrade | |
Bhutiani et al. | DO-BOD modeling of River Ganga from Devprayag to Roorkee, India using BMKB model | |
Samuelsson | Sensor Fault Detection and Process Monitoring in Water Resource Recovery Facilities | |
Volf et al. | Use of Machine Learning in Processes Optimization for Drinking Water Treatment Plant Butoniga (Istria, Croatia) | |
CN117192063B (en) | Water quality prediction method and system based on coupled Kalman filtering data assimilation | |
Kwater et al. | Comprehensive online estimation of object signals for a control system with an adaptive approach and incomplete measurements | |
CN116628121A (en) | River environmental factor and microbial community structure association measurement method |
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 | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20230705 Address after: 100728 No. 22 North Main Street, Chaoyang District, Beijing, Chaoyangmen Applicant after: CHINA PETROLEUM & CHEMICAL Corp. Applicant after: Sinopec Safety Engineering Research Institute Co.,Ltd. Address before: 100728 No. 22 North Main Street, Chaoyang District, Beijing, Chaoyangmen Applicant before: CHINA PETROLEUM & CHEMICAL Corp. Applicant before: SINOPEC Research Institute OF SAFETY ENGINEERING |
|
TA01 | Transfer of patent application right | ||
GR01 | Patent grant | ||
GR01 | Patent grant |