Clustered Energy Management Strategy of Plug-In Hybrid Electric Logistics Vehicle Based on Gaussian Mixture Model and Stochastic Dynamic Programming

Abstract

Energy management strategy is one of the most important technologies for hybrid electric vehicles. An energy management strategy based on the Gaussian mixture model and stochastic dynamic programming is proposed. The driving conditions are grouped into 5 clusters according to the velocity and power demand by Gaussian mixture model. 5 corresponding transition probability matrices can be obtained by the nearest neighbor method. Then, strategies generated by stochastic dynamic programming are linearly combined using posterior probabilities of Gaussian mixture model as weights. The simulation results show that, compared with the conventional strategies, the proposed strategy can conserve the equivalent fuel consumption and is more adaptive when driving cycle changes. The minimum equivalent fuel consumption of the proposed method can reach 101.07%, using the strategy based on dynamic programming as a baseline. As a result of the Jensen-Shannon divergence analysis, the proposed strategy improves performance by increasing the representativeness of the transition probability matrix for test driving cycles. Experimental test results indicate that the proposed strategy can be implemented on automotive standard microcontrollers. Compared with benchmark strategies, engine operating efficiency of the proposed strategy has been improved by up to 3.56% and engine operating time is reduced by up to 2.96%.