Authors: Du Guangbo1, Cai Mao2, Zhang Xin2, Fan Xingming2, Cheng Jianghua1Affiliations:1. China United Engineering Corporation, Hangzhou, Zhejiang 310052;2. Guilin University of Electronic Technology, Department of Electrical Engineering and Automation, Guilin, Guangxi 541004..Abstract:This study focuses on lithium batteries used in new energy vehicles, establishing a state of charge (SOC) estimation model based on Echo State Network (ESN) to predict the SOC of lithium batteries. The cross-validation method is employed to optimize the parameters of the Echo State Network, addressing the challenges of parameter selection in network models. The ESN model, trained using a recursive least squares method with a forgetting factor, updates the output weight matrix in real-time to enhance the adaptability and accuracy of the network. Simulation analysis of the model validates the feasibility of the prediction algorithm, further comparing the SOC estimation performance of the established ESN prediction model with that of BP neural network and Radial Basis Function (RBF) network algorithms under UDDS, US06, and NYCC conditions. The results indicate that the ESN model outperforms both BP and RBF algorithms in terms of performance and effectiveness for lithium battery SOC estimation, showing promising application prospects and providing a reference for long-term SOC prediction and assessment of lithium batteries.Introduction: Lithium batteries used in new energy electric vehicles are widely applied due to their advantages of being pollution-free, having strong endurance, and being rechargeable multiple times. The prediction of the state of charge (SOC) of lithium batteries is a key technology in the field of new energy vehicles. SOC reflects the remaining usable energy of the lithium battery[1-2], and accurately predicting SOC provides a reliable basis for studying the endurance of new energy electric vehicles, reasonable charging and discharging of lithium batteries, and battery health management[3]. The internal chemical reactions of lithium batteries are complex, and the changes in SOC are influenced by various factors such as temperature, the number of charge-discharge cycles, charge-discharge rates, and aging, making SOC prediction challenging[4]. Common methods for predicting SOC include ampere-hour integration, open-circuit voltage method, internal resistance method, and battery modeling method. However, these methods have limitations such as significant error accumulation[5-6], restricted application conditions[7-8], inability to directly measure actual SOC[9], and difficulties in parameter identification[10-13], which restrict their application scenarios. The cross-validation method is used to optimize the reservoir size N, spectral radius SR, input scaling IS, and input displacement IF of the Echo State Network (ESN), and the recursive least squares method with a forgetting factor is employed to adjust the network output weight matrix in real-time. To verify the feasibility and superiority of the ESN algorithm, simulations are conducted comparing the ESN algorithm with BP and RBF algorithms under UDDS conditions using different training and testing sets, followed by comparative analysis of the three algorithms under UDDS, US06, and NYCC conditions.Source: “Electronic Technology Application” Magazine, January IssueClick belowto read the original text and download the paper PDF
☞ Business Cooperation: ☏ Please call 010-82306118 / ✐ or email [email protected]
Click
here to “read the original text”, directly to the Electronic Technology Application official website