State of health assessment for echelon utilization batteries based on deep neural network learning with error correction

The accurate prediction of the state of health for retired batteries is the premise to ensure the safe and efficient operation of echelon utilization batteries. Aiming at the problems of limited battery cycle data and coupling of health status parameters, an assessment method of the state of health for echelon utilization batteries based on deep neural network learning with error correction is proposed in this paper. According to the reference discharge curve of echelon utilization batteries, the main characteristic parameters characterizing the performance aging of echelon utilization batteries are mined, and the state of health evaluation model of echelon utilization batteries based on deep neural network learning is established after the dimensionality of these characteristic parameters are reduced by the grey correlation analysis method. Markov chain error correction is used to further improve the prediction accuracy of the established deep neural network model. The effectiveness of the proposed method is verified by the simulation analysis of lithium-ion battery cycle test data from NASA Ames Prediction Center of Excellence. The simulation results show that the average absolute errors of the state of health prediction for echelon utilization batteries are less than 0.8% after the deep neural network learning prediction model is modified by Markov chain error, which provides a theoretical basis for the safe and stable operation of echelon utilization batteries.

Automated summary

This paper proposes an assessment method for the state of health of echelon utilization batteries based on deep neural network learning with error correction. The main characteristic parameters characterizing the performance aging of echelon utilization batteries are mined, and a state of health evaluation model is established using deep neural network learning. The prediction accuracy of the model is further improved using Markov chain error correction. The simulation analysis verifies the effectiveness of the proposed method, with average absolute errors of the state of health prediction for echelon utilization batteries being less than 0.8%.