多核相关向量机优化模型的锂电池剩余寿命预测方法

For the remaining useful life (RUL) prediction method based on relevance vector machine (RVM), kernel function is the important item of RVM model for the final prediction result. The current RVM prediction models are dominated by single kernel, and the selection of RVM kernel is a little bit subjective. So, the prediction performance of the constructed RVM model is limited. To address this problem, a multi kernel RVM model is proposed for the RUL estimation, using the fruit fly optimization algorithm (FOA) to find the best corresponding coefficients of multi kernel in the linear combination of multi kernel functions, and to improve the prediction performance of RVM model applied in the RUL estimation of lithium-ion battery. The battery test data sets of the national aeronautics and space administration (NASA) and the center of advanced life cycle engineering (CACLE) in the university of Maryland are used respectively. Experiments have been carried out to test the performance of the proposed method. The results show that the mean absolute error (MAE) and root mean square error (RMSE) of multi kernel RVM method are both less than the single kernel RVM algorithm.