Research On Real-time Charging Scheduling Strategy Based On Deep Learnin

With the continuous improvement of industrialization and transportation infrastructure in society,cars have become one of the indispensable transportation tools for people.However,traditional fuel-powered cars have gradually lost their market due to high emissions of pollutants and low fuel efficiency.The emergence of new energy electric vehicles provides people with an environmentally friendly,efficient,and economical way of transportation.Despite the widespread popularity of new energy electric vehicles,"charging difficulty" has become the most troubling issue for every electric vehicle user with the increase in the number of electric vehicles.In this context,it becomes crucial to address the timeliness,uncertainty,and cost reduction of charging.There are three main challenges in designing a reasonable electric vehicle charging strategy in intelligent transportation systems.Firstly,it is challenging to design an effective real-time electricity price prediction model to assist electric vehicles in completing charging tasks due to the volatility of electricity prices.Secondly,it is difficult to design an effective charging strategy to guide electric vehicles to select suitable charging stations for charging tasks based on the constantly changing road traffic conditions.Lastly,it is challenging to improve the efficiency of charging strategy generation by reducing the operating time in a Gym based traffic simulation environment.To address the above challenges,the main content of this paper is as follows:Firstly,this paper designs a GRU model based on the attention mechanism to achieve real-time prediction of charging station electricity prices,thus completing intelligent-assisted decision-making.The experiment validates that the proposed method has better predictive performance than other methods.Secondly,this paper describes the charging scheduling problem of electric vehicles as a Markov decision process and builds an electric vehicle charging environment based on Gym.We propose a model free deep reinforcement learning algorithm to dynamically learn the optimal charging strategy,which can cope with constantly changing road traffic conditions and weather conditions,and ultimately achieve lower total charging costs for users.From the analysis of experimental results,it can be seen that the method proposed in this paper has improved the reward convergence value by about 30% compared to other advanced deep reinforcement learning algorithms such as PPO algorithm,and has increased it by about 5% compared to A2 C algorithm.Thirdly,to reduce the impact of the large-dimensional system state transition calculation in the established electric vehicle charging environment on the generation of real-time charging strategies.In this paper,the LSTM deep learning model is used to predict the state transition of the system,thus resealing the Gym based charging scheduling environment,and finally use a model-based Soft Actor-Critic algorithm to achieve efficient electric vehicle charging decision-making.From the analysis of the experimental results,in terms of the size of the reward convergence value,the reinforcement learning method based on the model is about 100 higher than the reinforcement learning method based on the model,and the time efficiency of strategy generation is also higher than that of the reinforcement learning method based on the model.method improves by about 0.05 seconds.