Research On The Muti-objective Energy Management And Parameter Design In Data-driven Scenarios

With the increasingly severe fossil resource shortage and environmental pollution problems,the development of hybrid electric vehicles is an effective way to realize the transformation and upgrading of the energy-saving and new energy vehicle industry,and to improve the energy utilization efficiency of powertrain components.The power-split hybrid electric vehicle is the research focus of current hybrid electric technology route,which can achieve complete decoupling of the engine and the load side,and the fuel economy and dynamic performance potential of hybrid powertrain can be fully utilized.In this paper,the research on the multi-objective energy management and parameter design of dual-mode power-split hybrid electric vehicle in data-driven scenarios is carried out,and the details of this research are as follows.(1)The study on the construction of traffic scenarios reflecting the real driving patterns and driving habits of specific regions is conducted based on the data-driven approach,and a more refined upper-layer design is carried out for the driving cost evaluation of hybrid electric vehicles.In the upper layer,a typical test driving cycle of specific regions is constructed in a much smaller time span based on the real traffic flow information,which includes the division and typical characteristic extraction of driving cycle segments,dimensionality reduction and cluster analysis.In the lower level of component control,the control-oriented mathematical model of powertrain components including the transmission system are established.(2)The multi-objective energy management strategy considering battery aging based on the deep reinforcement learning is developed for the input power-split configuration.By establishing a control-oriented semi-empirical capacity decay model with battery cycle aging,the sensitivity analysis of battery severity factor reflecting battery health status is carried out,and a unified quantification method that comprehensively considering battery aging,fuel consumption and state of charge maintenance is proposed.The multi-objective energy efficiency optimization framework based on deep deterministic policy gradient algorithm with expert experience assistance is built,and the optimization performance,computational efficiency and multi-objective trade-off decision-making under different weight coefficients are analyzed with dynamic programming algorithm and deep Q-learning algorithm.(3)The steady-state power split characteristics of dual-mode configuration under different configuration modes are comprehensively analyzed,and the mode switching strategy aiming at maximizing powertrain efficiency is proposed.The dynamic programming optimization framework incorporating the mode switching strategy based on direct transfer point is built to explore the energy-saving potential of different powersplit configurations,which avoids the occurrence of power recirculation and improves the computational efficiency while obtaining the optimal control strategy.Combined with backward simulation,the effectiveness of the proposed mode switching strategy is verified.and the performance superiority of dual-mode power-split configuration is comprehensively analyzed and validated by taking into account the economic evaluation indexes of fuel economy and state of charge maintenance,and the dynamic performance evaluation indexes of acceleration ability.(4)The dual-layer multi-objective optimization framework based on the integration of MOEA/D(Multi-objective Evolutionary Algorithm Based on Decomposition)and dynamic programming is proposed for the first time for the optimized powertrain parameter design of dual-mode power-split configuration.In the upper layer,The MOEA/D algorithm explores and selects the powertrain parameters based on the Pareto optimality principle.In the lower layer,the dynamic programming algorithm is applied to evaluate the fuel economy and dynamics performance in terms of population adaptability value of different dual-mode configuration parameters.It explores the performance potential of the dual-mode power-split hybrid electric vehicle with comprehensive consideration of coupling mechanism between the energy management strategy and component parameter optimization.The computational efficiency and Pareto front are verified by the comparison with the NSGA-II multi-objective optimization method,and the obtained Pareto front can provide a wider design space for the multiobjective trade-off strategy and powertrain parameter optimization of dual-mode powersplit configuration.