Pattern Recognition Oriented Model Predictive Control Of Series Hybrid Vehicle Energy Management

As the core problem of hybrid vehicles,energy management has been a hot topic in recent years.The paper took series hybrid vehicles as the research object,and focused on the research of vehicle speed forecasting based on pattern recognition and energy management strategy based on model predictive control.First of all,the hybrid system was modeled with a reasonable simplification and optimization.After that,the paper focused on the classification and identification of vehicle driving patterns and the prediction of future vehicle speed.The K-means and EM(Expectation Maximization Algorithm)clustering algorithms were used to classify the driving conditions.After comparing the intra-class distance and the inter-class distance,the EM clustering algorithm was finally selected.The Euclidean distance was used to determine the current vehicle driving pattern.In the prediction of vehicle speed,the predictive effects were compared between three methods: exponential prediction,Markov chain model based prediction combined with pattern recognition,and Markov chain model based prediction without combined pattern recognition.The necessity of pattern recognition was verified and the final selection of predictive method was made.Energy management strategy based on model predictive control was the most important content of the paper.To meet the needs of vehicle dynamics and extend battery life,an energy management optimization problem was constructed,and quadratic programming was used to optimize the solution.Using the predicted vehicle speed to obtain the demand power,through the simulation and verification of the control strategy based on MPC,the power generated by the vehicle power elements could meet the demand power,the battery SOC adjustment process was obvious and the fluctuation range did not exceed 0.06,the number of working speed points was limited.And the engine always worked in efficient areas,and decoupling from the ground was achieved.In addition,an energy management strategy based on explicit model predictive control was studied to further reduce online computing time.Convert the system model into a linear time-invariant model,used the MPT toolbox to construct a constrained optimization problem,obtained the explicit function between the optimal control variables and the state variables through offline calculation,and generated an explicit model predictive controller.This controller was verified by simulation and it was found that both goals of the vehicle’s power and battery SOC’s maintenance were met,and the on-line calculation time was only 1.94 seconds in the entire cycle condition.