Wiener Process and Extreme Learning Machine Swarm Based Transfer Learning for Remaining Useful Life Prediction of Lithium Battery

As an effective clean energy source, lithium battery (LIB) has been widely applied in practical engineering. Thus, it is very meaningful to estimate the remaining useful life (RUL) of LIB, so that the reliability of engineering can be guaranteed. However, the limited degradation data of historical LIBs may lead to inaccurate result of RUL prediction. In this article, we propose an extreme learning machine swarm and Wiener process based transfer learning model (WP-ELMS- TL) to settle this trouble. Firstly, we utilize the two-phase WP model to expand the dataset of historical LIB. Based on the expanded dataset, a training strategy of repeated training and optimal selection is proposed to obtain the ELMS. Then, a novel strategy of TL is put forward to promote the accuracy of ELMS in predicting RUL. We illustrate two experiments of RUL prediction, where a simulated experiment is used to display the algorithm flow of our WP-ELMS-TL model, and a practical example is put forward to verify the feasibility of our WP-ELMS-TL model in practical engineering. By comparing with other available models, the superior of our WP-ELMS-TL model in generality and accuracy has been proved.