Neural network-based driving condition prediction and energy management strategy for power-split HEVs

Compared with the other Hybrid Electric Vehicles (HEVs), the advantage of power-split HEVs is that the various kinds and combination ways of the power-train and the benefits of the mechanical transmission and the electric transmission are combined. The energy management strategy is the key for the excellent economy and dynamic performance. Global optimization-oriented energy control algorithms can achieve the best fuel economy of HEVs, but it needs the priori knowledge of the driving condition. Therefore, in this paper, a neural network-based Driving Condition Prediction (DCP) method and an energy management strategy based on DCP have been proposed. BP (Back Propagation) neural network is used to predict the future driving condition in a short time horizon. The BP algorithm which is one of the most widely used algorithms has very high nonlinear fitting ability, and it can be used to predict the developing trend of driving data in practical application and simulation. For the DCP, the BP neural network was trained by the past driving condition data, and the output of the BP neural network is the future driving information in the limit time horizon. The optimization is executed in every sample time according to the prediction results. Simulation tests are carried out to verify the optimal energy management strategy. The results are compared with a rule-based logic threshold control strategy in the same driving cycle and prove that this energy strategy based on DCP is viable.