CN-121189552-B - Intelligent prediction control method and system for full-process energy consumption in lithium battery diaphragm production

CN121189552BCN 121189552 BCN121189552 BCN 121189552BCN-121189552-B

Abstract

The application provides a full-process energy consumption intelligent prediction control method and system for lithium battery diaphragm production, which belong to the technical field of lithium battery production, and the method comprises the steps of obtaining multi-source data in the lithium battery diaphragm production process of a prediction target period; the method comprises the steps of inputting multi-source data into a preset graph neural network model to obtain a predicted energy consumption value of a predicted target period, obtaining an actual energy consumption measured value of the predicted target period, calculating an energy consumption deviation value between the predicted energy consumption value and the actual energy consumption measured value, judging whether control optimization is needed or not based on comparison between an absolute value of the energy consumption deviation value and a preset deviation threshold value, screening from the multi-source data to obtain a target state variable set if optimization is needed, combining the energy consumption deviation value, the preset energy consumption target value and the target state variable set into state information, and inputting the state information into a pre-trained reinforcement learning model to obtain a target control strategy. The application can improve the timeliness and the accuracy of energy consumption control.

Inventors

WANG HONGBING
WANG CONG
JIANG HUI
ZHAO YUAN
BIAN GUANGYU
DU FEIYUE

Assignees

武汉惠强新能源材料科技有限公司

Dates

Publication Date: 20260508
Application Date: 20250910

Claims (8)

1. The intelligent prediction control method for the full-process energy consumption of the lithium battery diaphragm production is characterized by comprising the following steps of: Acquiring multi-source data in the production process of the lithium battery diaphragm in a predicted target period, wherein the multi-source data comprises process parameters, equipment operation parameters, environment parameters and quality detection parameters; The construction process of the graph neural network model comprises the steps of taking a production process as a node and taking a material transmission relation and an energy interaction relation among the processes as edges to construct a graph structure, adopting an adjacent matrix to represent the connection strength among the nodes, and enabling a node characteristic vector to be composed of multi-source data of the corresponding process, wherein the graph neural network model comprises an input layer, a graph convolution layer, a pooling layer and an output layer, and the output layer outputs the predicted energy consumption value by adopting a linear activation function; After the prediction target period is finished, acquiring an actual energy consumption value of the prediction target period, and calculating an energy consumption deviation value between the prediction energy consumption value and the actual energy consumption value; Judging whether control optimization is needed or not based on the comparison of the absolute value of the energy consumption deviation value and a preset deviation threshold value; Comparing the absolute value of the energy consumption deviation value with a preset deviation threshold value, and extracting the state variable set based on the comparison result to obtain a target state variable set, wherein the state variable set is used as the target state variable set if the absolute value of the energy consumption deviation value is smaller than or equal to the preset deviation threshold value, and the state variable in the state variable set is extracted as the target state variable set based on the feature importance evaluation result if the absolute value of the energy consumption deviation value is larger than the preset deviation threshold value; Combining the energy consumption deviation value and the target state variable set into state information; Inputting the state information into a pre-trained reinforcement learning model to obtain a target control strategy, wherein the target control strategy comprises a process parameter adjustment amount and/or an equipment operation parameter adjustment amount; Acquiring an actual energy consumption value and corresponding evaluation multi-source data in a preset evaluation period after executing a target control strategy; Calculating a reward signal through a reward function based on the actual energy consumption value, the energy consumption target value in a preset evaluation period, the predicted quality index, the preset quality index target value and the quality qualification threshold value; Constructing an experience tuple by using the initial state information of the evaluation period, the executed target control strategy, the reward signal and the state information of the evaluation period at the end of the evaluation period; inputting the evaluation multisource data into the graph neural network model to obtain a predicted energy consumption value, calculating a prediction error between the energy consumption value and an actual energy consumption value, and optimizing parameters of the graph neural network model based on the prediction error.
2. The intelligent predictive control method for full-process energy consumption in lithium battery diaphragm production according to claim 1, further comprising: In the execution process of the target control strategy, calculating an adjustment step length of a process parameter adjustment amount and/or an equipment operation parameter adjustment amount based on an absolute value of the energy consumption deviation value and a history adjustment record, wherein the history adjustment record comprises adjustment amounts in a plurality of past control periods and corresponding absolute values of the energy consumption deviation; executing the adjustment amount of the target control strategy in the adjustment step size range; and if the adjustment quantity required by the target control strategy is larger than the adjustment step length, executing the adjustment quantity of the target control strategy based on the adjustment step length, and generating an alarm signal.
3. The intelligent predictive control method for full-process energy consumption in lithium battery diaphragm production according to claim 2, further comprising: After the alarm signal is generated, if the absolute values of the energy consumption deviation values in the continuous N control periods are all larger than a preset deviation threshold value, suspending the optimization process of lithium battery diaphragm production control; And positioning potential fault equipment or an abnormal process link based on the correlation analysis of the historical multi-source data and the energy consumption deviation value.
4. The intelligent predictive control method for full-process energy consumption in lithium battery diaphragm production according to claim 1, further comprising: When the predicted quality index is smaller than the quality qualification threshold, inputting the estimated multi-source data into a quality constraint proxy model to generate a process parameter safe operation interval; And regenerating a target control strategy through a reinforcement learning model under the constraint of the safe operation interval.
5. The intelligent predictive control method for full-process energy consumption in lithium battery separator production according to claim 1, wherein calculating a reward signal by a reward function based on an actual energy consumption value, a preset evaluation period energy consumption target value, a predicted quality index, a preset quality index target value, and a quality qualification threshold value comprises: if the predicted quality index is smaller than the quality qualification threshold, outputting a negative punishment signal as a reward signal; If the predicted quality index is greater than or equal to a quality qualification threshold, calculating an energy consumption rewarding component according to the deviation between the actual energy consumption value and an energy consumption target value in a preset evaluation period; Calculating a quality rewarding component according to the deviation between the predicted quality index and a preset quality index target value; And carrying out weighted fusion on the energy consumption rewarding component and the quality rewarding component to obtain a rewarding signal.
6. The intelligent predictive control method for full-process energy consumption in lithium battery diaphragm production according to claim 1, wherein the extracting the state variable in the state variable set as the target state variable set based on the feature importance evaluation result comprises: Based on a pre-trained feature importance evaluation model, evaluating all state variables in the state variable set to obtain an importance score of each state variable; ranking from high to low based on importance scores corresponding to all state variables in the state variable set; Determining a target cumulative importance duty cycle threshold based on the absolute value of the energy consumption bias value; and sequentially selecting the state variables based on the sorting until the sum of the accumulated importance scores of the selected state variables reaches or is larger than the product of the target accumulated importance duty ratio threshold and the sum of the total importance scores, and taking the selected state variables as a target state variable set.
7. The intelligent predictive control method for full-process energy consumption in lithium battery separator production according to claim 6, wherein the determining a target cumulative importance duty cycle threshold based on the absolute value of the energy consumption deviation value comprises: If the absolute value of the energy consumption deviation value is smaller than a first threshold value, the target accumulated importance ratio threshold value is a first preset proportion; if the absolute value of the energy consumption deviation value is larger than or equal to a first threshold value and smaller than or equal to a second threshold value, the target accumulated importance duty ratio threshold value is a second preset proportion; if the absolute value of the energy consumption deviation value is larger than a second threshold value, the target accumulated importance ratio threshold value is a third preset proportion; Wherein the first predetermined ratio is less than the second predetermined ratio, which is less than the third predetermined ratio.
8. The utility model provides a lithium cell diaphragm production whole flow energy consumption intelligent prediction control system which characterized in that includes: the data acquisition module is used for acquiring multi-source data in the lithium battery diaphragm production process of a predicted target period, wherein the multi-source data comprises process parameters, equipment operation parameters, environment parameters and quality detection parameters; The data processing module is used for inputting the multi-source data into a preset graph neural network model to obtain a predicted energy consumption value of a predicted target period, wherein the construction process of the graph neural network model comprises the steps of taking a production process as a node, taking a material transmission relation and an energy interaction relation among the processes as edges to construct a graph structure, adopting an adjacent matrix to represent the connection strength among the nodes, and enabling a node characteristic vector to be composed of multi-source data of a corresponding process; the data calculation module is used for obtaining the actual energy consumption value of the predicted target period and calculating an energy consumption deviation value between the predicted energy consumption value and the actual energy consumption value; The data judging module is used for judging whether control optimization is needed or not based on the comparison of the absolute value of the energy consumption deviation value and a preset deviation threshold value; The control strategy module is used for extracting the multi-source data to obtain a state variable set if optimization is needed, comparing the absolute value of the energy consumption deviation value with a preset deviation threshold value, and extracting the state variable set based on a comparison result to obtain a target state variable set, wherein the state variable set is used as the target state variable set if the absolute value of the energy consumption deviation value is smaller than or equal to the preset deviation threshold value, and the state variable in the state variable set is extracted as the target state variable set based on a feature importance evaluation result if the absolute value of the energy consumption deviation value is larger than the preset deviation threshold value; the energy consumption deviation value, the preset energy consumption target value and the target state variable set are combined to form state information, the state information is input into a pre-trained reinforcement learning model to obtain a target control strategy, and the target control strategy comprises a process parameter adjustment amount and/or an equipment operation parameter adjustment amount; Acquiring an actual energy consumption value and corresponding evaluation multi-source data in a preset evaluation period after executing a target control strategy; Calculating a reward signal through a reward function based on the actual energy consumption value, the energy consumption target value in a preset evaluation period, the predicted quality index, the preset quality index target value and the quality qualification threshold value; Constructing an experience tuple by using the initial state information of the evaluation period, the executed target control strategy, the reward signal and the state information of the evaluation period at the end of the evaluation period; inputting the evaluation multisource data into the graph neural network model to obtain a predicted energy consumption value, calculating a prediction error between the energy consumption value and an actual energy consumption value, and optimizing parameters of the graph neural network model based on the prediction error.

Description

Intelligent prediction control method and system for full-process energy consumption in lithium battery diaphragm production Technical Field The application relates to the technical field of lithium battery production, in particular to an intelligent predictive control method and system for full-process energy consumption in lithium battery diaphragm production. Background In the field of lithium battery diaphragm production, accurate control of full-process energy consumption is a core difficult problem of green manufacturing. In the current production process, multi-source data such as process parameters, equipment states, environmental conditions, product quality and the like have strong coupling characteristics, and the problem of complex energy consumption fluctuation is difficult to be solved by a single-factor adjustment mode. In the prior art, the traditional energy consumption prediction model depends on a single parameter fitting or simple machine learning method, and internal correlation among technological parameters, equipment states and environmental factors cannot be effectively excavated, so that prediction accuracy is low, and reliable support cannot be provided for real-time regulation and control. Meanwhile, the existing control mode often falls into the dilemma of passive intervention after the energy consumption exceeds the standard, and the problems of serious energy waste, low production benefit and the like caused by the difficulty in realizing the dynamic optimization of the whole process are caused. In addition, the parameter setting depending on manual experience is easy to cause the conditions of high energy consumption or unstable product quality, and the existing control method lacks the dynamic integration capability of multi-source data, so that the whole-flow energy consumption optimization target is difficult to achieve. Along with the green transformation acceleration of manufacturing industry, the timeliness and the accuracy requirements of enterprises on energy consumption management and control are continuously improved, so that a full-process intelligent energy consumption prediction control method and system for lithium battery diaphragm production are needed to achieve the improvement of the timeliness and the accuracy requirements on energy consumption management and control. Disclosure of Invention In order to solve the technical problems, the application provides a full-process energy consumption intelligent prediction control method and system for lithium battery diaphragm production, so as to improve the timeliness and accuracy requirements of energy consumption management and control. In a first aspect of the embodiment of the application, an intelligent prediction control method for full-process energy consumption in lithium battery diaphragm production is provided, which comprises the following steps: Acquiring multi-source data in the production process of the lithium battery diaphragm in a predicted target period, wherein the multi-source data comprises process parameters, equipment operation parameters, environment parameters and quality detection parameters; Inputting the multi-source data into a preset graph neural network model to obtain a predicted energy consumption value of a predicted target period; after the prediction target period is finished, acquiring an actual energy consumption measured value of the prediction target period, and calculating an energy consumption deviation value between the prediction energy consumption value and the actual energy consumption measured value; Judging whether control optimization is needed or not based on the comparison of the absolute value of the energy consumption deviation value and a preset deviation threshold value; Screening from the multi-source data to obtain a target state variable set if optimization is required; Combining the energy consumption deviation value and the target state variable set into state information; and inputting the state information into a pre-trained reinforcement learning model to obtain a target control strategy, wherein the target control strategy comprises a process parameter adjustment amount and/or an equipment operation parameter adjustment amount. Preferably, the intelligent prediction control method for the full-process energy consumption of the lithium battery diaphragm production further comprises the following steps: In the execution process of the target control strategy, calculating an adjustment step length of a process parameter adjustment amount and/or an equipment operation parameter adjustment amount based on an absolute value of the energy consumption deviation value and a history adjustment record, wherein the history adjustment record comprises adjustment amounts in a plurality of past control periods and corresponding absolute values of the energy consumption deviation; executing the adjustment amount of the target control strategy in the adjustment step si