Remaining useful life prediction of PEMFC systems under dynamic operating conditions

The Prognostic and Health Management (PHM) has been developed for more than two decades. It is capable to anticipate the impending failures make decisions in advance extend lifespan of target systems, such as Proton Exchange Membrane Fuel Cell (PEMFC) systems. a critical stage PHM. Among various prognostic methods, data-driven ones could predict system based on device’s knowledge historical data. In Remaining Useful Life (RUL) prediction, Indicators (HIs) should be able reflect health states PEMFC stack. Moreover, an effective HI help define explicit degradation state improve prediction accuracy. HIs voltage power are usually used under static conditions due their monotonic decreasing characteristics. Besides, measurements current implemented easily practice. Nevertheless, unable directly dynamic operating because they sensitive mission profiles. To overcome weakness HIs, convenient practical named Relative Power-loss Rate (RPLR) proposed herein. According polarization curve at beginning life, initial different profiles can identified. Then actual obtained by monitoring continuously. Finally, RPLR calculated power. Afterward, RUL predicted some Artificial Intelligence (AI) algorithms. approaches, Echo State Network (ESN) provided efficient promising solution Compared with classical Recurrent Neural (RNN), it accelerate convergence rate reduce computational complexity. traditionally single-input ESN structure feeble handle varying As scheduling variable, interesting parameter since represents working properties extent. Considering system’s characteristics, stack regarded another input ESN, output matrix’s dimension increased same time. Therefore, double-input enhance performance. Based RPLR, three micro-cogeneration (?-CHP) durability tests systems verify improved structure.