Electric Vehicle Battery Evaluation Based on Long-Term Memory Recurrent Neural Network
Aiming at the low evaluation accuracy of existing new energy battery state evaluation algorithms,an evaluation meth-od based on long-term memory recurrent neural network was designed.Firstly,the indicators related to the health degree and the working voltage of the power battery were extracted and used as the original data set to evaluate the state of the battery.Secondly,a recurrent neural network framework is constructed to replace neurons with short and long term memory modules,but the longitudinal communication ability between neurons is retained.According to the gating mechanism of the long and short time memory module,the status of the module unit is determined and the final evaluation result is output.The experimental results show that the model it-eration training capability of the proposed power battery evaluation algorithm is stronger than that of recurrent neural network and feedforward neural network,and the accuracy of battery state evaluation is close to the theoretical curve,and the voltage kurtosis value deviations of the training set and the test set is 0.002 and 0.004,respectively.
long and short term memoryrecurrent neural networkpower batterystate assessmentiterative capability
万方数据
维普
