RUL prognostics method based on real time updating of LSTM parameters

Traditional LSTM model cannot effectively use the non-life-cycle data to establish an excellent RUL Prognostics model since it cannot utilize the online data reasonably. For small sampling data LSTM learning, this paper proposes an improved LSTM method with real-time parameters updating by using new online observation data to minimize the cost function. Taking NASA lithium-ion battery data as an example, the applicability of the improved LSTM model with real-time parameter updating in the field of remaining useful life Prognostics is verified.