Online Prediction Of Battery Manufacturing Capacity Based On Formation State Analysis And Reinforcement Learning

In the field of predicting the manufacturing capacity of power batteries,traditional neural network models are difficult to accurately describe and predict the manufacturing capacity of power batteries due to the influence of various external environmental factors,signal transmission delays,complex production conditions,and uncertainties.Artificial intelligence methods such as reinforcement learning are commonly used in manufacturing prediction,and predictive analysis of power battery manufacturing capacity is of great significance for enterprises to arrange manufacturing cycles reasonably and improve the safety and profitability of power battery manufacturing.Therefore,research on the prediction of power battery manufacturing has always been a hot topic in the field of power battery manufacturing,and has certain theoretical significance.This paper focuses on the online prediction of the manufacturing capacity of batteries based on formation state analysis and reinforcement learning,and the main research work is as follows:(1).The accuracy of battery formation data is the key factor affecting the online prediction of power battery manufacturing capacity.Aiming at the unavoidable issue of delay in the charge-discharge signal transmission process during the formation process in the manufacturing of power batteries,the iterative deduction method is introduced in this study to establish the convex space structure of the system.Then,from the perspective of convex space,the noise and disturbance at the current moment are used to construct the strip space,and obtain the convex space structure that satisfies the conditions for state prediction and update.And the linear programming inequalities are established based on proposed constraints,the best conservation convex space body of the feasible set for wrapping the states is obtained after the solution,and the state estimation algorithm for noise uncertainty time-delay systems based on convex space contraction filtering is proposed,and provide data source for online prediction of power battery manufacturing capacity.(2).Aiming at the problem that traditional neural networks have difficulty predicting the optimal fit for the manufacturing capacity of power batteries,reinforcement learning is used to construct the learning environment of hidden layer nodes in recurrent neural networks and long short-term memory network models,and an optimal hidden layer node optimization algorithm is studied to further reduce the prediction bias of the hidden layer.Then,a weight learning environment of the combination model is constructed,and the optimal weight is obtained after iterative training,further reducing the prediction bias of the combination weight.The analysis shows that the power battery manufacturing capacity combination prediction method based on the neural network model has high reliability and prediction accuracy.(3).Aiming at the real-time problem of power battery online prediction,online training optimization is introduced.Using the double exponential smoothing prediction method,the manufacturing capacity predicted in the previous step is recursively updated to predict the next step,avoiding prediction errors caused by data fluctuation trends and thus improving online prediction accuracy.In order to solve the problem of unknown online weights during the online prediction update process,the reinforcement learning idea is used to establish the optimal online weight learning environment and propose an online prediction algorithm for power battery manufacturing capacity based on reinforcement learning.In conclusion,this paper conducted research on the online prediction of manufacturing capabilities for batteries based on formation state analysis and reinforcement learning,and the effectiveness and feasibility of the proposed method were verified through simulation examples.Finally,the paper summarized and looked forward to the research content of this topic.