Predictive equivalent consumption minimization strategy based on driving pattern personalized reconstruction

Range-extended electric vehicles combine the benefits of electric propulsion with the convenience and range flexibility of conventional internal combustion engine vehicles. In the energy management system, an equivalent consumption minimization strategy (ECMS) ensures balanced utilization of various energy sources to meet driving demand in real time, thereby enhancing energy efficiency. This paper proposes a driving pattern personalized reconstruction (DPPR) method to maximizes the accuracy of velocity prediction and to serve a predictive ECMS (P-ECMS) with high adaptability to uncertain scenarios. In this context, open-source driving cycles are segmented, and features of each driving pattern are extracted for cluster analysis. Utilizing the results from clustering, new driving cycles are reconstructed for the purpose of training adaptive neuro-fuzzy inference system (ANFIS) models. Personalized velocity prediction is achieved through bias fusion in the multi-ANFIS model, adapting to various driving conditions. The performance of the proposed P-ECMS is validated across various driving conditions and compared with conventional ECMS (C-ECMS), adaptive ECMS (A-ECMS), dynamic programming (DP), and fuzzy logic (FL) strategy. The results indicate that using the proposed P-ECMS is strongly robust in terms of battery pack SoC stabilization under various driving conditions, whose SoC difference achieves 97.02% of using the DP strategy. Compared to using other strategies, using P-ECMS achieves better SoC's stability and lower fuel consumption than using the FL strategy (11.1%), using A-ECMS (1.7%), and using C-ECMS (26.6%).