A hybrid ensemble deep reinforcement learning model for locomotive axle temperature using the deterministic and probabilistic strategy

Abstract This paper proposes a hybrid deep reinforcement learning framework for locomotive axle temperature by combining the wavelet packet decomposition (WPD), long short-term memory (LSTM), gated recurrent unit (GRU) and generalized autoregressive conditional heteroskedasticity (GARCH) algorithms. The WPD is utilized to decompose raw nonlinear series into subseries. Then predictors LSTM GRU are established predict future temperatures in each Q-learning could generate optimal ensemble weights integrate finish deterministic forecasting GARCH used conduct based on residual. These parts of structure contributed modelling accuracy provided effective support real-time monitoring fault diagnosis transportation.