ARIMA-Based Road Gradient and Vehicle Velocity Prediction for Hybrid Electric Vehicle Energy Management

Abstract

Road gradient and vehicle velocity are critical information in deciding the power demand of hybrid electric vehicles (HEVs), and greatly impact the powertrain energy management performances. Generally, the road gradient is assumed to be known in previous studies. This paper presents a data-driven autoregressive integrated moving average (ARIMA) based method, aiming to predict the short-term future velocity and road gradient in real time for the predictive energy management of HEVs. The established ARIMA-based learning model provides vehicle velocity and road gradient prediction references during each control horizon for the controller. Model predictive control is employed to construct the predictive energy management strategy, with dynamic programming to resolve the optimal powertrain control problem during each control horizon. Real driving cycle and road gradient data are collected via experiments, and used to establish the ARIMA predictors. Simulation results show that the ARIMA model is able to predict the future velocity and road gradient with reasonable accuracy. With the ARIMA predictor used in the predictive energy management strategy, the HEV fuel consumption is effectively reduced by about 5%–7% compared with when no predictor used.