Energy Management Strategy For Extended-Range Electric City Buses Based On Driving Condition Adaptation

Extended range electric city buses(E-REBs)have two energy sources,including a range extender and a battery pack,and their operation characteristic is that the driving distance is fixed.In order to improve vehicle fuel economy,energy management strategies need to manage the battery charge while allocating the power between the two energy sources,and make it decline slowly to last to the end of the trip to increase the ratio of the battery pack operating in the high efficiency range.Currently,there is still a gap on how to effectively establish adaptive control methods based on driving conditions in this slow charge release mode.As a result,the performance of energy management strategies is sensitive to driving condition types.Focusing on the above-mentioned problem,this thesis proposes an on-line energy management strategy based on driving condition adaptation,and validates the proposed strategy through simulations and bench tests.Real-time optimization algorithm is the key of constructing an on-line energy management strategy.The Pontryagin's Minimum Principle(PMP)is applied to solve the vehicle energy optimization problem under different types of driving cycles.By analyzing the characteristics of the optimal solutions,it is found that the ratio of the co-state to the open circuit voltage of the battery pack is approximately constant during the whole driving cycle,and the value of such constant is related to the driving condition type.Based on this finding,a real-time PMP algorithm is proposed,whose control parameter is associated with driving condition type.The performance of this proposed algorithm is close to that of the PMP.Driving conditions clustering is the foundation of constructing driving condition adaptive control parameters.Taking micro-trip between two starts as the unit to divide driving condition samples,a driving condition clustering method is proposed which focuses on the fuel consumption differences of samples.Through correlation analysis,redundant driving feature parameters are removed and driving feature parameters related to fuel consumption are selected.Based on the components extracted from Principal Component Analysis(PCA)and in combination with the Euclidean norm,a normed linear space is constructed.Using the K-means algorithm and a density algorithm to cluster driving condition samples and evaluating the corresponding clustering effects,the results show that the density algorithm increases the fuel consumption difference between two types of driving condition samples.Based on the clustering results of the density algorithm,three types of driving condition samples with different fuel consumption levels are obtained.Based on the clustering results and the characteristics that the micro-trip is difficult to be correctly recognized in the early stage,an on-line hybrid method combining prediction and recognition for driving condition identification is proposed.In the early stage of a micro-trip,using a Markov chain model to predict its type.In the later period,using a Naive Bayesian model to dynamically recognize its type.Through researching the accuracy estimation methods for the two models and the best switching time between the results of the two models,the proposed hybrid identification method obtains a higher driving condition on-line identification accuracy than existing methods.Typical driving cycles are constructed to represent the three types of driving condition samples.Giving different remaining mileage and initial values of State of Charge(SOC),the corresponding optimal control parameters for the above typical driving cycles are calculated.Based on this,an energy management strategy based on driving condition adaptation is established.For on-line control,the energy allocation is realized by identifying the type of driving condition,and then calculating the optimal control parameter using look-up tables based on the type of driving condition,the remaining mileage and SOC.The simulation and bench test results show that the proposed strategy improves fuel economy,and its performance is insensitive to driving condition types.