Stochastic Model Predictive Control Strategy With Short-Term Forecast Optimal SOC for a Plug-In Hybrid Electric Vehicle

Both the stochastic traffic information and state of charge (SOC) greatly impact the plug-in parallel hybrid electric vehicle performance. Uncertain cycles and driving styles affect the effectiveness of velocity prediction and further cause the instability of SOC estimate. These uncertain stochastic factors interfere with the solution of torque demand at different degrees in each control cycle. To address this issue, a stochastic model predictive control (SMPC) considering short-term forecast optimal SOC is proposed. First, multiple linear regression of engine and battery is developed for energy management strategy (EMS). Then, the velocity prediction model is developed based on Markov chain considering the driver styles, and reference SOC is optimized by dynamic programming (DP) with the forthcoming information. Finally, the SMPC-based EMS with the short-term optimal SOC is constituted. The verification results show that Markov based on driver styles has better predictive performance than radial basis function neural networks and backpropagation neural networks. The fuel economy of the proposed strategy improves by about 11.8% compared with normal model predictive control and is close to that of the globally optimal DP. The test results indicate that the SMPC with the short-term optimal SOC can promote EMS to improve fuel economy.