An Improved Energy Management Strategy for HEV Based on Driving Condition Prediction Within a Finite Time Horizon

Weida Wang, Yanqin Wang, Changle Xiang, Chao Wei, Yulong Zhao

Abstract


Energy management based on driving condition information or driving condition prediction within a finite time horizon is a way of further improving the fuel economy of hybrid electric vehicles (HEVs). The future power demand of vehicles is predicted by using support vector regression (SVR) and fuel economy is further optimised on-line to acquire the optimal operating points of both engine and motor. According to the speed characteristics, the driving cycle is classified as a certain speed mode by using the K nearest neighbour (KNN) algorithm. By taking the speed characteristics as training data, the prediction model for vehicle speed based on SVR under a certain speed mode is obtained through training by applying partial swarm optimisation (PSO). The PSO-SVR method can greatly improve speed prediction accuracy within a finite time horizon both accurately and efficiently. The power demand within the finite time horizon is calculated and the assigned values of various control variables at every sample time are acquired by using dynamic programming (DP) within a finite time horizon. The research results indicated that the control strategy based on predicted information shows a significantly positive effect in optimising fuel consumption and battery utilisation.

Keywords


hybrid electric vehicles; support vector regression; energy management strategy; partial swarm optimisation; K nearest neighbour algorithm


DOI
10.12783/dteees/iceee2018/27904

Refbacks

  • There are currently no refbacks.