Study On Engine Torque Estimate For Parallel Hybrid Electric Vehicle

The quantities of motorcars are more and more, along with the problem of serious energy sources and motorcar let. For limiting pollution of cars and saving fuel, Hybrid Electric Vehicle (HVE) is developing inside and outside of the world now. Demarcated power of HEV engine only can satisfy the natural run, when it is accelerating, climbing slope and high speed, the increased power is provided by electrical engine which is drove by storage battery, so the specification of the HEV is much smaller, then the oil consume and pollution of smoke much smaller. Otherwise, because the secondary cell supplies electricity just when it is needed, the capacitance is much smaller, and then the problem of small secondary capacitance is resolved.In this paper, BP (Back Propagation) network——a sort of neural network (NN) is discussed, which is used to estimate engine torque for Parallel Hybrid Electric Vehicle (PHEV). The engine performance data measured on test-bench are trained in Neural Network tool of Matlab. A lot of BP networks for neuron cells number in different layers are established, and the networks are used to train the data which is measured when PHEV engine was in steady and dynamic control. During the process of the establishment of network models, a new training method is proposed. After analyzing the different models'training results, the model with higher accuracy and generalization ability is selected out, by use of which, the actual torque is satisfactorily estimated. Then, the engine steady models and dynamic models which are putted into the whole vehicle models are validated rationally.