基于GA-ELM模型的锂电池SOH预测

针对锂电池健康状态(State of Healthy,SOH)预测精度低的特点,利用遗传算法改进的极限学习机(Extreme Learning Machine,ELM)算法可提高锂电池SOH的预测精度。ELM输入层到隐含层的权值及隐含层单元的阈值随机产生,ELM算法只需设置隐含层单元的数目及隐含层激活函数类型。相比传统BP算法,ELM算法具有学习速率快、泛化性能好等优点。但由于ELM网络输入层到隐含层的权值和隐含层阈值产生的随机性,ELM算法的稳定性较差。ELM算法中引入遗传算法(GA)优化输入层到隐含层的权值和隐含层单元的阈值,该方法可增强ELM算法的稳定性。实验对比了GA-ELM算法与ELM算法、BP算法、RBF算法及SVR算法对锂电池SOH的预测,结果显示GA-ELM算法相比其他算法在预测精度和算法稳定性上均有提升。

Lithium Battery State of Healthy Prediction Based on GA-ELM Model

In view of the low prediction accuracy of lithium battery state of health (SOH),the extreme learning machine (ELM) algorithm improved by genetic algorithm can improve the prediction accuracy of lithium battery SOH.The ELM input layer to the hidden layer weight and the hidden layer unit threshold are randomly generated.The ELM algorithm only needs to set the number of hidden layer units and the hidden layer activation function type.Compared with the traditional BP algorithm,the ELM algorithm has the advantages of fast learning rate and good generalization performance.However,due to the randomness of the input layer to the hidden layer of the ELM network and the hidden layer threshold,the stability of the ELM algorithm is poor.The genetic algorithm (GA) is introduced into the ELM algorithm to optimize the weight of the input layer to the hidden layer and the threshold of the hidden layer unit.This method can enhance the stability of the ELM algorithm.In the experiments,the lithium battery SOH prediction of GA-ELM algorithm is compared with that of ELM algorithm,BP algorithm,RBF algorithm and SVR algorithm.The results show that the GA-ELM algorithm can improve the prediction accuracy and algorithm stability compared with other algorithms.