Research On New Energy Sales Strategy Based On Deep Reinforcement Learning

With the continuous improvement of the technical level of renewable energy utilization and the gradual expansion of the industrial scale,new energies such as wind,water and solar have gradually become important resources to promoting energy diversification and achieve sustainable development.In the electric power industry,rapid development of new energies generation technology not only relieves the rapid consumption of non-renewable energy,but also brings new challenges to the commercial use of new energies.A large number of new energy sources are connected to the grid,and the reform of the electricity market tends to be competitive.As a result,new energy is faced with many problems when participating in market transactions.It is not only necessary to pursue the maximization of benefits,but also necessary to consider the risks caused by the uncertainty of new energy such as wind power generation.The sales strategy of new energy is to consider and find the way to obtain the maximum profit of new energy from the perspective of market trading.This paper takes into account the uncertainty of new energy generation and the interaction between supply and demand of the system,focuses on the two models,one is the model of wind power generation and energy storage system cooperation for electricity sale abroad;One is the model of wind power generation,photovoltaic power generation and energy storage system cooperation for electricity sale abroad,The main work is as follows:(1)The electric energy sales model of wind storage hybrid system is established.In the model,the wind-power winning profit in the day-ahead market,the discount sales profit of electric energy,the up-down purchase of reserve and the calculation of penalty cost are considered.When calculating the revenue of discount sales,the influence of the user’s demand response based on the discount of electricity price on the revenue is considered.The objective function is to maximize the revenue of the wind-storage hybrid system,and the constraint condition is the charge and discharge power limitation and storage quantity limitation of the energy storage system.according to the proposed model with the characteristics of high complexity and uncertainty,Q-learning algorithm of reinforcement learning was used to solve the problem,the results show that the effectiveness of the model and the optimized sales strategy is given.(2)The electric energy sales model of wind-solar storage hybrid system is established.The income of the hybrid system in the day-ahead market includes the income from electricity sale,the up-down standby purchase fee and the penalty fee.In addition,the operating cost of the photovoltaic system is also considered in the revenue model.Objective function: The wind-solar storage hybrid system has the largest electricity sales revenue,and the constraints include the charge and discharge limits of the energy storage system,the stored amount of electricity,and the photovoltaic power generation output limits.DQN algorithm of deep reinforcement learning is used to solve this problem.The results show the practicability of the proposed model and the power sales strategy of the optimized wind-wind-storage hybrid system is given.