基于最优权阈值ELM算法的锂离子电池RUL预测

Prediction of Lithium-ion Battery RUL Based on Optimal Weight and Threshold Using ELM Algorithm

针对锂离子电池剩余使用寿命RUL(remaining useful life)预测结果不准确及极限学习机ELM(extreme learning machine)权阈值随机选取等问题,提出利用ELM模型间接预测锂离子电池RUL的方法 ,并利用遗传蚂蚁算法GAAA(genetic algorithm ant algorithm)选取ELM的最优权值与阈值,建立基于等压降放电时间间接寿命特征参数的最优GAAA-ELM锂离子电池RUL预测模型。基于NASA锂离子电池数据集预测和评估锂离子电池的RUL,并与BP模型预测方法、ELM模型预测方法和GA-ELM模型预测方法相比较,结果表明该方法能够更准确有效地实现锂离子电池RUL预测。

Since the predictions of lithium-ion battery remaining useful life（RUL） are inaccurate and the selection of weights and thresholds for an extreme learning machine（ELM） is random, an indirect prediction method for lithium-ion battery RUL is proposed based on ELM model. Moreover, the optimal weight and threshold of ELM are selected using genetic algorithm-ant algorithm（GAAA）, and the prediction model for lithium-ion battery RUL using GAAA-ELM is established based on the time interval to equal discharging voltage which is a kind of indirect life feature character. Finally, the lithium-ion battery RUL is predicted and assessed based on NASA data sets of lithium-ion battery, which is further compared with that obtained using BP, ELM, and GA-ELM prediction model methods, showing that the proposed method can accurately and effectively predict the lithium-ion battery RUL.