An industrial process control method and system based on expected setting error neural network

Abstract

The present invention discloses an industrial process control method and system based on an expected setting error neural network, comprising: establishing a control output network model and a target prediction network model, determining a specific control target and a corresponding control object in the industrial process, selecting a control variable according to the control target, and setting a corresponding expected value; collecting time series data in the industrial process, and performing data processing on the collected time series data; training a target prediction network using the time series data after data processing, obtaining a prediction result, and calculating and integrating a plurality of error types to optimize the target prediction network; training a control output network according to the prediction result and time series data provided by the target prediction network, and calculating and integrating a plurality of error types to optimize the control output network; applying the trained network model to industrial process control, and continuously performing additional training according to environmental changes. The present invention can improve the control effect of the industrial process.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial process control, and in particular to an industrial process control method and system based on an expected setting error neural network. Background Art

[0002] As an important part of the modern industrial system, process manufacturing covers many fields such as chemical industry, petroleum, pharmaceuticals, and food. In the process of process manufacturing, there are a large number of equipment working together through automatic control, and the quality of automatic control directly determines the quality of the products produced by the enterprise. Especially with the intensification of globalization and market competition, the process manufacturing industry faces multiple challenges such as improving production efficiency, reducing production costs, ensuring product quality and environmental protection. In this context, the introduction and application of intelligent control technology has become the key to the transformation and upgrading of the process manufacturing industry.

[0003] At present, most of the automatic control of industrial production processes still uses traditional PID control. With the introduction of the concept of intelligent manufacturing in recent years and the improvement of machine learning technology, more and more companies have begun to use machine learning algorithms to achieve automatic control of production processes. However, the currently available machine learning control schemes are generally divided into two types. One is to fit historical control data with relatively good performance through various algorithms such as neural networks for subsequent production control outputs. The other is to make regular compensation adjustments to traditional controllers by predicting the trend of future control variables.

[0004] The shortcoming of the first solution is that the control factors in the production process are changeable, so the use of excellent historical data experience cannot cope with abnormal scenarios; the shortcoming of the second solution is that the influencing factors of the control target are diverse, and it is a very complicated process. It is one-sided to only correct traditional controllers such as PID controllers through future target predictions. Summary of the invention

[0005] Technical purpose: In view of the shortcomings of existing industrial process control technology, the present invention discloses an industrial process control method and system based on an expected set error neural network, which can improve the control effect.

[0006] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:

[0007] An industrial process control method based on an expected setting error neural network, comprising:

[0008] Establish a control output network model and a target prediction network model, determine the specific control target and the corresponding control object in the industrial process, select the control variable according to the control target, and set the corresponding expected value, wherein the control output network includes the state timing network of the control output network, the expected target neural network and the integrated neural network of the control output network, and the target prediction network includes the state timing network of the target prediction network, the next step output neural network and the integrated neural network of the target prediction network;

[0009] Collect time series data in industrial processes and process the collected time series data;

[0010] The target prediction network is trained using the processed time series data to obtain the prediction results and calculate and integrate various error types to optimize the target prediction network;

[0011] The control output network is trained based on the prediction results and time series data provided by the target prediction network, and multiple error types are calculated and integrated to optimize the control output network;

[0012] The trained control output network and target prediction network are applied to industrial process control, and continuous additional training is performed according to environmental changes.

[0013] The present invention also provides an industrial process control system based on a desired set error neural network, which is used to implement the above-mentioned industrial process control method based on a desired set error neural network, comprising:

[0014] A control output network is used to output subsequent control actions according to the current control state and the desired target. The control output network includes a state timing network of the control output network, a desired target neural network and an integrated neural network of the control output network;

[0015] The state timing network of the control output network is used to learn and extract features from the timing data of the current control state;

[0016] A desired target neural network, used to calculate the characteristics of the desired control target;

[0017] The integrated neural network of the control output network is used to integrate the state timing network of the control output network and the output of the desired target neural network to generate the next control action;

[0018] A target prediction network is used to predict the control target according to the current control state and the next control action. The target prediction network includes a state timing network of the target prediction network, a next output neural network and an integrated neural network of the target prediction network.

[0019] The state time series network of the target prediction network is used to learn and extract features from the time series data of the current control state;

[0020] The next step is to output the neural network, which is used to calculate the features of the next control action;

[0021] The integrated neural network of the target prediction network is used to integrate the state timing network of the target prediction network with the output of the next step output neural network to generate a prediction target.

[0022] Beneficial effects: The industrial process control method and system based on the expected setting error neural network provided by the present invention have the following beneficial effects:

[0023] The present invention evaluates and optimizes system performance at multiple levels by introducing factual data errors, adversarial errors and empirical errors. In particular, adversarial errors can enhance the system's adaptability to abnormal and extreme situations. The accuracy of control output is improved by integrating the state timing network and the desired target neural network. The integrated neural network can effectively fuse the output characteristics of different networks to ensure that the generated control instructions are more accurate and efficient.

[0024] By continuously collecting new data and conducting real-time incremental learning and model retraining, the system can dynamically adapt to changes in the production environment and maintain long-term stable and efficient operation. In actual production, the network is continuously supplemented with training to ensure that the system can adapt to changes in the production environment and improve the robustness and adaptability of the control system.

[0025] Compared with the PID control in the prior art, the present invention can more accurately predict and adjust the control variables and improve production efficiency and product quality by combining the control output network and the target prediction network; and the present invention can process multiple control variables at the same time and optimize them, making complex industrial control processes more efficient and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0027] Figure 1 It is a flow chart of the industrial process control method based on the expected setting error neural network of the present invention;

[0028] Figure 2 It is a structural framework diagram of an industrial process control system based on a desired setting error neural network of the present invention;

[0029] Figure 3 It is a schematic diagram of the temperature error control result in the practical application of the present invention.

[0030] In the figure: 100, control output network; 101, state timing network of control output network; 102, expected target neural network; 103, integrated neural network of control output network; 110, target prediction network; 111, state timing network of target prediction network; 112, next step output neural network; 113, integrated neural network of target prediction network. DETAILED DESCRIPTION

[0031] The present invention will be described more clearly and completely below by way of a preferred embodiment in conjunction with the accompanying drawings, but the present invention is not limited to the scope of the embodiment.

[0032] In one embodiment, Figure 1 As shown, an industrial process control method based on an expected setting error neural network is provided, comprising the following steps:

[0033] S101. Establish a control output network model and a target prediction network model, determine the specific control target and the corresponding control object in the industrial process, select the control variable according to the control target, and set the corresponding expected value, wherein the control output network includes the state timing network of the control output network, the expected target neural network and the integrated neural network of the control output network, and the target prediction network includes the state timing network of the target prediction network, the next step output neural network and the integrated neural network of the target prediction network.

[0034] In one embodiment, the control target refers to the process parameters that need to be controlled in the industrial process, such as temperature, pressure, etc., and the control object refers to the equipment or certain processes associated with the process parameters.

[0035] S102: Collect time series data in the industrial process and perform data processing on the collected time series data.

[0036] In one embodiment, the time series data includes historical operation data and real-time monitoring data to ensure the comprehensiveness and accuracy of the data, covering a variety of operating states and abnormal situations.

[0037] In one embodiment, data processing of the collected time series data includes feature expansion, feature selection and normalization. The purpose of feature expansion of the data is to introduce derived features to enhance the predictive ability of the system model; the purpose of feature selection is to screen the most representative data features and propose redundant information to simplify the system model; the purpose of normalization of the data is to unify the size and reduce the impact of data differences on system model training.

[0038] S103, using the processed time series data to train the target prediction network, obtain the prediction results and calculate and integrate various error types to optimize the target prediction network.

[0039] The purpose of training the target prediction network is to predict future control targets and provide a basis for the system's automated decision-making. The specific steps include:

[0040] S301. Input the timing data corresponding to the Kth timing node into the state timing network of the target prediction network, input the next step output C calculated by the control output network corresponding to the timing data of the K+1th timing node into the next step output neural network, and output the predicted target P1 of the K+Nth timing node after calculation by the integrated neural network of the target prediction network.

[0041] The predicted target P1 is the future target predicted by the target prediction network under normal conditions based on the current time series data and control output. The specific process is: the current time series data such as temperature, pressure and other control states are input into the state timing network 111 of the target prediction network. The state timing network 111 of the target prediction network learns the timing characteristics of the current state and outputs a feature vector. The current control output is used as input and input into the next step output neural network 112. The integrated neural network 113 of the target prediction network integrates the outputs of the state timing network 111 of the target prediction network and the next step output neural network 112 to generate the predicted target P1.

[0042] S302: Calculate the actual data training error E1 based on the actual output Q of K+N time series nodes and the predicted target P1. The calculation formula is as follows:

[0043]

[0044] in, represents the mean square error, represents the S-curve, Indicates absolute value.

[0045] The actual data training error E1 is used to evaluate the difference between the predicted target and the actual output. The mean square error (MSE) and sigmoid function are applied to make the error between 0 and 1.

[0046] S303, input the next step output C2 calculated by the control output network corresponding to the timing data of K+1 timing nodes into the next step output neural network, and output the predicted target P2 of the K+Nth timing node after calculation by the integrated neural network of the target prediction network.

[0047] The predicted target P2 is the next control action predicted by the control output network under normal conditions based on the current state, and the predicted future control target. The specific process is as follows: the desired target such as ideal temperature or pressure is input into the desired target neural network 102, and the current timing data is input into the state timing network 101 of the control output network. The integrated neural network 103 of the control output network integrates the outputs of the state timing network 101 of the control output network and the desired target neural network 102 to generate a control action C2, and C2 is input into the next output neural network 112 in the target prediction network. The predicted target P2 is calculated again through the state timing network 111 of the target prediction network, the next output neural network 112 and the integrated neural network 113 of the target prediction network.

[0048] S304. Calculate the adversarial error E2 of the control output network according to the actual output Q of the K+N time series nodes and the predicted target P2. The adversarial error E2 is used to evaluate the robustness of the system under abnormal conditions. The calculation formula is as follows:

[0049]

[0050] The calculation formula uses adversarial examples to evaluate the system's ability to cope with extreme conditions.

[0051] S305. Select a value C3 in the interval [C, x1] as the next step output corresponding to the time series data of K+1 time series nodes, and input it into the next step output neural network. After calculation by the integrated neural network of the target prediction network, the predicted target P3 of the K+Nth time series node is output, where x1 represents the upper limit of C3. Generally, after normalization, the value of x1 is 1.

[0052] The prediction target P3 is obtained by randomly selecting a value within a reasonable range based on the current control output, and then calculating it through the target prediction network. The specific process is as follows: based on the current control output, a random value within a reasonable range is selected, and the selected random value is input into the target prediction network as the new control output. The target prediction network obtains the prediction target P3 through the joint action of the state timing network 111 of the target prediction network, the next step output neural network 112 and the integrated neural network 113 of the target prediction network.

[0053] S306. Calculate the empirical error E3 of the positive correlation between the control variable and the expected target based on the actual output Q of K+N time series nodes and the predicted target P3. The calculation formula is as follows:

[0054]

[0055] If the actual system control target is positively correlated with the control input, the theoretical P3 must be larger than the actual value Q. If P3>Q, the loss is 0, otherwise the loss E3>0.

[0056] S307. Select a value C4 in the interval [y1, C] and input it into the next step output neural network. The prediction target P4 of the K+Nth time series node is calculated through the integrated neural network of the target prediction network, where y1 represents the lower limit of C4. After normalization, the value of y1 is -1.

[0057] The predicted target P4 is another value randomly selected within a reasonable range based on the current control output, and then calculated through the target prediction network. The specific process is as follows: based on the current control output, a random value within a reasonable range is selected, and the selected random value is input into the target prediction network as a new control output. The target prediction network obtains the predicted target P4 through the joint action of the state timing network 111 of the target prediction network, the next step output neural network 112 and the integrated neural network 113 of the target prediction network.

[0058] S308. Calculate the empirical error E4 of the anti-correlation between the control variable and the expected target based on the actual output Q of the K+N time series nodes and the predicted target P4. The calculation formula is as follows:

[0059]

[0060] The empirical error E3 of the positive correlation between the control variable and the expected target and the empirical error E4 of the negative correlation between the control variable and the expected target are used to evaluate the system's handling of the positive and negative correlations between the control variables. By adjusting the input variables and observing the relationship between the prediction and the actual, it is ensured that the system responds to changes reasonably.

[0061] S309, taking E1+E2+E3+E4 as the total error to adjust the weights of the target prediction network, the back propagation algorithm is used here to minimize the total error, and the network weights are updated by the gradient descent method to improve the prediction ability of the system.

[0062] S104: training the control output network according to the prediction results and time series data provided by the target prediction network, and optimizing the control output network by calculating and integrating various error types. The purpose of training the control output network is to ensure the accuracy and reliability of the system operation, which specifically includes the following steps:

[0063] S401, input the actual output Q of K+N timing nodes into the state timing network of the control input network, and calculate the next step output C2 through the integrated neural network of the control output network.

[0064] By utilizing a state-series network with the same state-series network structure as the target prediction network and focusing on the training of the desired target neural network, the computational burden can be reduced and the accuracy of the control output can be ensured.

[0065] S402: Calculate the adversarial error E5 of the control output network according to the actual output Q of the K+N time series nodes and the predicted target P2. The adversarial error E5 is used to evaluate the difference between the output and the actual demand. The calculation formula is as follows:

[0066]

[0067] S403, input 0 into the desired target neural network of the control output network, calculate the next output C6 through the integrated neural network of the control output network, then input the next output C6 into the next output neural network of the target prediction network, and calculate the target prediction P6 through the integrated neural network of the target prediction network.

[0068] The predicted target P6 is the future control target predicted by the target prediction network on the basis of inputting the expected target value 0 into the expected target neural network 102 of the control output network, and specifically includes the following steps: inputting the expected target value 0 into the expected target neural network 102, and the expected target neural network 102 processes the expected target value to generate an expected target feature vector; inputting the current control state data, such as temperature, pressure, etc. into the state timing network 101 of the control output network, extracting the state feature vector, and the integrated neural network 103 of the control output network integrates the feature vectors of the state timing network 101 of the control output network and the expected target neural network 102 to generate a control action C6; and then inputting the control action C6 into the target prediction network to calculate the target prediction P6.

[0069] S404, calculate the ideal expected error E6 according to the target prediction P6. The ideal expected error E6 is used to ensure that the control output achieves the expected target. The calculation formula is as follows:

[0070]

[0071] S405. Select a value Q1 in the interval [Q, x2] and input it into the desired target neural network. The next output C7 is calculated through the integrated neural network of the control output network. x2 represents the upper limit of Q1. The value of x2 after normalization is generally 1.

[0072] S406, calculating the empirical error E7 of the negative correlation between the control variable and the next step output according to the next step output C and the next step output C7, the calculation formula is as follows:

[0073]

[0074] S407. Select a value Q2 in the interval [y2, Q] and input it into the desired target neural network. The next output C8 is calculated through the integrated neural network of the control output network. After normalization, the value of Y2 is generally -1.

[0075] S408, calculating the empirical error E8 of the negative correlation between the control variable and the next step output according to the next step output C and the next step output C8, the calculation formula is as follows:

[0076]

[0077] The empirical error E7 of the positive correlation between the control variable and the next output and the empirical error E8 of the negative correlation between the control variable and the next output are used to ensure that the control output achieves the expected goal, adjust the control output, observe the relationship between the period and the actual system state, and adjust the error to optimize the control effect.

[0078] S409, taking E5+E6+E7+E8 as the total error to adjust the weight of the control output network. Here, the network parameters are adjusted by back propagation to improve the control effect and ensure the decision-making ability of the control output network.

[0079] S105, applying the trained control output network and target prediction network to industrial process control, and continuously performing additional training according to environmental changes to ensure the stability and adaptability of the system in a real environment, specifically including the following steps:

[0080] S501, applying the trained network to actual production, generating control instructions in real time, and adjusting and optimizing the control strategy according to feedback from the production environment;

[0081] S502. Continuously collect new operating data, update training data sets, regularly retrain the system, adapt to new data and environmental changes, and maintain continuous optimization of system performance;

[0082] S503. Adjust the control strategy in a timely manner according to the real-time feedback in the production process, use the incremental learning mechanism to continuously update and optimize the model, and improve the robustness and stability of the system.

[0083] like Figure 2 As shown, the present invention also provides an industrial process control system based on a desired set error neural network, which is used to implement the above-mentioned industrial process control method based on a desired set error neural network, including:

[0084] The control output network 100 is used to output subsequent control actions according to the current control state and the expected target to ensure the accuracy and reliability of the system operation. The control output network 100 includes a state timing network 101 of the control output network, a desired target neural network 102 and an integrated neural network 103 of the control output network.

[0085] The state timing network 101 of the control output network is used to learn and extract features of the timing data of the current control state; a long short-term memory network (LSTM) or a gated recurrent unit (GRU) is used to process the timing data, and a Tanh activation function is used. The number of input and output neurons is determined according to actual industrial control requirements, and the number of hidden layer neurons is determined by an automatic parameter adjustment algorithm, such as Bayesian hyperparameter tuning.

[0086] The desired target neural network 102 is used to calculate the characteristics of the desired control target; a fully connected neural network (FCNN) is used, and a Tanh activation function is used. The number of input and output neurons is consistent with the number of control targets. If the temperature and pressure need to be controlled simultaneously, the number of input and output neurons is 2.

[0087] The integrated neural network 103 of the control output network is used to integrate the output of the state timing network of the control output network and the desired target neural network to generate the next control action; FCNN is adopted, and the Tanh activation function is used. The number of input neurons is the same as the number of output neurons of the state timing network 101 of the control output network; the number of hidden layer neurons is finally determined according to the overall network automatic parameter adjustment algorithm, such as Bayesian hyperparameter tuning; the number of output neurons is determined according to the control object, such as controlling the temperature and pressure by controlling 1 cold water valve and 1 high-pressure gas valve, then the output neuron is 2.

[0088] The target prediction network 110 is used to predict the control target according to the current control state and the next control action. The target prediction network includes a state timing network 111 of the target prediction network, a next output neural network 112 and an integrated neural network 113 of the target prediction network.

[0089] The state timing network 111 of the target prediction network is used to learn and extract features from the timing data of the current control state; it shares the architecture with the state timing network 101 of the control output network, adopts LSTM or GRU, uses the Tanh activation function, the number of input and output neurons is determined according to actual needs, and the number of hidden layer neurons is determined by an automatic parameter adjustment algorithm.

[0090] The next step is to output the neural network 112, which is used to calculate the characteristics of the next control action; FCNN is used, and the Tanh activation function is adopted. The number of input and output neurons is consistent with the number of controlled objects.

[0091] The integrated neural network 113 of the target prediction network is used to integrate the state timing network of the target prediction network with the output of the next output neural network to generate a prediction target. FCNN is adopted, and the Tanh activation function is used. The number of input neurons is the same as the number of output neurons of the state timing network 111 of the target prediction network. The number of hidden layer neurons is finally determined according to the overall network automatic parameter adjustment algorithm, such as Bayesian hyperparameter tuning. The number of output neurons is determined according to the control target. If temperature and pressure need to be controlled simultaneously, the number of input and output neurons is 2.

[0092] In specific applications, if a project needs to control the temperature during the polymerization reaction by controlling the cold water flow rate, the timing state characteristics of this system are selected such as liquid level, water flow, and temperature, and the control output is the opening of the cold water valve.

[0093] After the data is normalized, the time series data corresponding to the first 100 time series nodes are used as the input of the time series state network, and the temperature of 130°C is used as the training data for the expected target and the predicted target. The definition of the expected target is the ideal state that the system wants to achieve, which guides the output of the control network during the training process. The cold water valve opening is set to 101% and used as the training data for the next step of outputting the neural network. Through this data, the system can learn how to adjust the opening of the cold water valve under a specific temperature target, and use the industrial process control method based on the expected setting error neural network provided by the present invention for training and control.

[0094] The target prediction network is trained to simulate the complex multivariable control process in actual production. By learning the relationship between different state features, the model can predict future temperature changes.

[0095] The goal of the control output network is to generate corresponding control instructions (such as the opening adjustment of the next cold water valve) based on the expected future temperature changes (for example, the set value error is 0). The output of this network is the control instruction directly used in production.

[0096] After training is completed, the final target prediction network can reflect the complex model of actual production multivariable control, while the control output network can output the control operation that should be performed in the next step based on the expected target after 30 steps, such as the set value error of 0.

[0097] In the actual production process, only the control output network is used. However, since the system will gradually change with time and environment during the production process, incremental training must be performed continuously. The purpose of incremental training is to enable the system to continuously adapt to new production conditions and ensure its stability and effectiveness in actual production. In incremental training, the two networks need to be updated and trained together.

[0098] like Figure 3 As shown, the horizontal axis in the figure represents the subscript of the time series, and the vertical axis represents the deviation of the control result. By using historical data simulation, the control method of the present invention can well control the temperature error to near 0. In the figure, the blue is the temperature error of the production PID control, and the red is the temperature error of the control of the present invention.

[0099] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An industrial process control method based on expected setting error neural network, characterized in that: include: Establish a control output network model and a target prediction network model, determine the specific control target and the corresponding control object in the industrial process, select the control variable according to the control target, and set the corresponding expected value, wherein the control output network includes the state timing network of the control output network, the expected target neural network and the integrated neural network of the control output network, and the target prediction network includes the state timing network of the target prediction network, the next step output neural network and the integrated neural network of the target prediction network; Collect time series data in industrial processes and process the collected time series data; The target prediction network is trained using the processed time series data to obtain the prediction results and calculate and integrate various error types to optimize the target prediction network; The control output network is trained based on the prediction results and time series data provided by the target prediction network, and multiple error types are calculated and integrated to optimize the control output network; Apply the trained control output network and target prediction network to industrial process control, and continuously perform additional training according to environmental changes; Use time series data to train the target prediction network, obtain the prediction results, and calculate and integrate various error types to optimize the target prediction network, including: The time series data corresponding to the Kth time series node is input into the state time series network of the target prediction network, and the next step output C calculated by the control output network corresponding to the time series data of the K+1th time series node is input into the next step output neural network. After calculation by the integrated neural network of the target prediction network, the predicted target P1 of the K+Nth time series node is output; The actual data training error E1 is calculated based on the actual output Q of K+N time series nodes and the predicted target P1. The calculation formula is as follows: ； in, represents the mean square error, represents the S-curve, Indicates absolute value; The next step output C2 calculated by the control output network corresponding to the time series data of the K+1 time series nodes is input into the next step output neural network, and the prediction target P2 of the K+Nth time series node is output after calculation by the integrated neural network of the target prediction network; The adversarial error E2 of the control output network is calculated based on the actual output Q of K+N time series nodes and the predicted target P2. The calculation formula is as follows: ； A value C3 is selected in the interval [C, x1] as the next step output corresponding to the time series data of the K+1 time series nodes, and is input into the next step output neural network. After calculation by the integrated neural network of the target prediction network, the prediction target P3 of the K+Nth time series node is output, where x1 represents the upper limit of the value of C3; According to the actual output Q of K+N time series nodes and the predicted target P3, the empirical error E3 of the positive correlation between the control variable and the expected target is calculated. The calculation formula is as follows: ； Select a value C4 in the interval [y1, C] and input it into the next step output neural network. The prediction target P4 of the K+Nth time series node is calculated through the integrated neural network of the target prediction network, where y1 represents the lower limit of C4. According to the actual output Q of K+N time series nodes and the predicted target P4, the empirical error E4 of the anti-correlation between the control variable and the expected target is calculated. The calculation formula is as follows: ； Take E1+E2+E3+E4 as the total error to adjust the weights of the target prediction network.

2. The industrial process control method based on the expected setting error neural network according to claim 1 is characterized in that: The control output network is trained based on the prediction results and time series data provided by the target prediction network, and multiple error types are calculated and integrated to optimize the control output network, including: Set the state timing network of the control output network to be the same as the state timing network of the target prediction network, input the actual output Q of the K+N timing nodes into the state timing network of the control input network, and calculate the next output C2 through the integrated neural network of the control output network; The adversarial error E5 of the control output network is calculated based on the actual output Q of K+N time series nodes and the predicted target P2. The calculation formula is as follows: ； Input 0 into the desired target neural network of the control output network, calculate the next output C6 through the integrated neural network of the control output network, then input the next output C6 into the next output neural network of the target prediction network, and calculate the target prediction P6 through the integrated neural network of the target prediction network; The ideal expected error E6 is calculated based on the target prediction P6. The calculation formula is as follows: ； Select a value Q1 in the interval [Q, x2] and input it into the desired target neural network. The next output C7 is calculated by the integrated neural network of the control output network, where x2 represents the upper limit of Q1. According to the next step output C and the next step output C7, the empirical error E7 of the positive correlation between the control variable and the next step output is calculated. The calculation formula is as follows: ； Select a value Q2 in the interval [y2, Q] and input it into the desired target neural network. The next output C8 is calculated by the integrated neural network of the control output network, where y2 represents the upper limit of Q2. According to the next step output C and the next step output C8, the empirical error E8 of the negative correlation between the control variable and the next step output is calculated. The calculation formula is as follows: ； Take E5+E6+E7+E8 as the total error to adjust the weights of the control output network.

3. The industrial process control method based on the expected setting error neural network according to claim 1 is characterized in that: Time series data includes historical data and real-time data. Data processing of the collected time series data includes feature expansion, feature selection and normalization.

4. An industrial process control system based on a desired set error neural network, used to implement an industrial process control method based on a desired set error neural network as described in any one of claims 1 to 3, characterized in that: include: A control output network is used to output subsequent control actions according to the current control state and the desired target. The control output network includes a state timing network of the control output network, a desired target neural network and an integrated neural network of the control output network. The state timing network of the control output network is used to learn and extract features from the timing data of the current control state; A desired target neural network, used to calculate the characteristics of the desired control target; The integrated neural network of the control output network is used to integrate the state timing network of the control output network and the output of the desired target neural network to generate the next control action; A target prediction network is used to predict the control target according to the current control state and the next control action. The target prediction network includes a state timing network of the target prediction network, a next output neural network and an integrated neural network of the target prediction network. The state time series network of the target prediction network is used to learn and extract features from the time series data of the current control state; The next step is to output the neural network, which is used to calculate the features of the next control action; The integrated neural network of the target prediction network is used to integrate the state timing network of the target prediction network with the output of the next step output neural network to generate a prediction target.

Images