基于改进灰狼优化与支持向量回归的锂电池健康状态预测

Prediction for state of health of lithium-ion battery by improved grey wolf optimization and support vector regression

为了提高锂电池健康状态(SOH)的预测精度,将改进的灰狼优化(IGWO)算法与支持向量回归(SVR)相结合,提出了一种基于改进灰狼优化和支持向量回归(IGWO-SVR)的联合算法。该算法的核心思想是运用改进的GWO算法解决SVR模型中的参数联合寻优问题。IGWO-SVR随机产生1个灰狼种群,灰狼个体的位置向量由SVR模型的3个参数C,σ,ε组成。根据每只灰狼的位置信息进行学习,并计算适应度。按照适应度值对狼群进行分级,对灰狼个体位置进行更新,然后进行差分进化操作,选择优秀个体进入下一代种群,重新计算灰狼个体在新位置的适应度。迭代过程结束后,提取狼群中适应度最优的灰狼位置信息作为最终的SVR模型参数进行训练。在美国国家航空航天局(NASA)锂电池数据集上的实验表明了所提SOH预测方法的有效性。

To improve the prediction accuracy of state of health(SOH)for lithium-ion batteries,a joint algorithm based on the improved grey wolf optimization and the support vector regression(IGWO-SVR)is proposed by combining improved grey wolf optimization(IGWO)algorithm with the support vector regression(SVR).The core idea of this algorithm is to use the IGWO algorithm to solve the problem of parameter joint optimization in the SVR model.The IGWO-SVR generates a wolf population randomly.The position vector of the wolf is composed of C,σ,εthree parameters of the SVR model.The location information of each grey wolf is used to learn,and the fitness is calculated.The wolves are graded according to the fitness value,and the locations of individual grey wolves are updated.Then the differential evolution operation is carried out to select excellent individuals to enter the next generation population,and the fitness of gray wolf individuals in the new position is recalculated.At the end of the iteration process,the grey wolf location information with optimal fitness in the wolves is extracted and trained as the final SVR model parameters.Experiment on the lithium-ion batteries datasets of NASA demonstrates the effectiveness of the proposed SOH prediction method.