Abstract: For the problem of low accuracy of lithium-ion battery capacity prediction, a seeking optimization algorithm-support vector machine (SOA-SVM) based capacity prediction method is proposed. By analyzing the random discharge process of lithium-ion battery, two indicators, the mean and standard error of random discharge capacity reflecting the capacity change of lithium-ion battery are constructed which are used as the feature parameters for capacity prediction. The principle component analysis is used to analyze the correlation between the feature parameters and extract the principle components. Based on the first principle component and the capacity data of part of tested batteries, SOA is used to optimize hyper-parameters of SVM and train the model. The optimized model combined with the first principle component date of other batteries is adopted to predict the capacity of lithium-ion batteries. The prediction results show that the proposed capacity prediction method has high prediction accuracy

Key words: lithium-ion battery, random discharge condition, capacity prediction, SOA, support vector machine