迅速获取精确的锂电池的剩余使用寿命和健康状态,对于维持锂电池的可靠性至关重要.针对锂电池剩余使用寿命(remaining useful life,RUL)预测精度较低,传统的高斯过程回归(Gaussian process regression,GPR)模型的超参数寻优结果不理想、预测效果差等问题,使用蚁狮优化算法(ant-lion optimization algorithm,ALO)对高斯过程回归的超参数进行寻优,实现锂电池剩余使用寿命的精确预测.首先,根据电池充电过程中电池电压的循环曲线,提取了6个参数作为电池的健康因子,然后采用Pearson相关系数验证健康因子与电池容量的相关关系,最终选出平均放电电压、恒流充电阶段电池存储的充电量、整个充电阶段电池存储的充电量以及时间积分中的放电温度这4个参数作为健康因子.最后,利用支持向量回归(support vector regression,SVR)、GPR和ALO-GPR对锂电池RUL进行预测,对各项指标进行比较分析.并将本工作所提出的模型与其他文献所提出的模型进行了比较.通过NASA锂电池数据集验证了模型的有效性,实验结果表明,所提出ALO-GPR的RUL预测模型误差小,均方根误差控制在1%以内,平均绝对误差控制在0.65%以内,泛化性强,具有良好的应用前景.
Prediction of the remaining useful life of lithium batteries based on Antlion optimization Gaussian process regression
Rapidly obtaining accurate information about the remaining useful life(RUL)and health status of a lithium battery is critical to maintaining its reliability.To solve the problems of low prediction accuracy regarding the RUL of lithium batteries,unsatisfactory hyperparameter optimization results,and poor prediction effect of the traditional Gaussian process regression(GPR)model,in this study,the Antlion optimization algorithm was used to optimize the hyperparameters of Gaussian process regression(hereinafter referred to as"ALO-GPR")to accurately predict the RUL of lithium batteries.First,according to the cycle curve of battery voltage during battery charging,six parameters were extracted as the health factors of the battery;subsequently,the correlation between these factors and the battery capacity was verified by using the Pearson correlation coefficient.Finally,the following four parameters were selected as the health factors:the average discharge voltage,the amount of charge amount stored by the battery in the constant current charging stage,the amount of charge stored by the battery in the whole charging stage,and the discharge temperature in the time integral.Finally,support vector regression,GPR,and ALO-GPR were used to predict the RUL of lithium batteries,and various indicators were compared and analyzed.The model proposed in this study is compared with models proposed in other literatures.The effectiveness of the proposed model is verified by using the NASA lithium battery dataset.The experimental results show that the RUL prediction model of ALO-GPR has a small error;the root mean square error is controlled within 1%;and,the average absolute error is controlled within 0.65%.Thus,ALO-GRP shows strong generalization and a good application prospect regarding the prediction of RUL of lithium batteries.
lithium-ion batteryGaussian process regressionant lion optimized algorithmremaining useful life
国家科技期刊平台
NSTL
万方数据
