Estimation of engine torque based on improved BP neural network

Aiming at the mass-energy power assembly control system in HEVs, a method is designed to estimate the engine torque, which is based on improved BP neural network. Based on the experiment results in engine dynamometer, and strong nonlinear characteristic of the engine is taken into account, traditional BP neural network error function is improved, and it is trained by optimal stopping, as a result over-fitting will be avoided. The engine torque output model is established with MATLAB, and it has high estimated accuracy and nice generalization ability. After all, validity of the algorithm mentioned above is verified by experiments.