Series-parallel hybrid power system energy management method based on DQN variant

Abstract

The invention discloses a series-parallel hybrid power system energy management method based on a DQN variant and belongs to the technical field of series-parallel hybrid electric vehicles, thereby improving the training convergence speed and the vehicle fuel economy. The method comprises the steps that a series-parallel hybrid electric vehicle model is established, and environmental parameters, including road gradient and vehicle-mounted quality, influencing an energy management strategy are obtained; an optimal energy management strategy is calculated by utilizing a dynamic programming (DP)algorithm, experience is stored into an optimal experience pool (OEB), a model is trained by combining a hybrid experience playback (HER) technology and adopting a Deuling DQN strategy to obtain a trained deep reinforcement learning agent, and energy management is carried out on the series-parallel hybrid electric vehicle under different working conditions. The HER technology and the DQN variant Dueling architecture constructed by the method can effectively improve the training convergence rate, the automobile fuel economy and the algorithm robustness.

Description

technical field [0001] The invention belongs to the technical field of hybrid hybrid vehicles, and in particular relates to an energy management method for a hybrid hybrid system based on a DQN variant. Background technique [0002] In today's increasingly serious energy crisis, vehicle emission standards are gradually becoming stricter, and the use of pure fuel vehicles has been challenged, while hybrid vehicles take into account the long driving range of fuel vehicles and the emission-free electric vehicles to solve the problem of fossil fuel combustion Therefore, the energy management of hybrid power has always been the key to research. [0003] At present, the energy management of hybrid electric vehicles is mostly a rule-based strategy. By setting a certain energy management threshold, the most common rule for plug-in hybrid electric vehicles is to first consume battery energy, and then maintain battery power. energy control. The representative benchmark of the optimi...

AI Technical Summary

Problems solved by technology

[0004] However, there are some shortcomings in the methods in the prior art. The energy management effect based on rules is often not significant enough. For a single working condition, a lot of experience and knowledge are required. The optimization-based DP requires the global working conditions to be known, and it cannot be real-time due to the long calculation time. Online application, the existing model prediction can be optimized and carried out in real time, but the step size of the prediction control cannot be selected too large, and there is still a large gap compared with the optimization result of DP

And many optimization methods are not considered comprehensively, ignoring the road slope information and the change of the car's own vehicle quality

Images