Research On Demand Response And Distributed Energy Optimal Scheduling In Cloud Environment

Under the background of ubiquitous power Internet of things,how to achieve the demand response ability of residential users and the optimal scheduling of distributed generation and distributed energy storage in the cloud environment has become a research hotspot.However,due to the strong fluctuation and uncertainty of distributed energy output,how to improve the new energy consumption rate is also a big problem.At the same time,the demand response of residential users is difficult to quantify.Different users have different response trends in different scenarios,which makes it difficult to make full use of the demand response ability of residential users.With the development of ubiquitous power Internet of things technology,it is possible to use refined demand response network to deeply mine the user side demand response capability,which provides a new solution to the optimal scheduling of distributed energy.Based on the above background,firstly this paper establishes a cloud platform architecture system for the optimal scheduling of distributed power supply,proposes the Cloud-Edge-Terminal communication architecture and operation mode based on MQTT protocol,and divides it into two stages according to the different operation modes: the Perception and the Control.At the same time,the cloud platform of distributed energy optimization and scheduling decision-making based on cloudera architecture is built,on which distributed computing framework and the Tensorflow machine learning architecture are deployed,providing communication and computing support for the subsequent distributed energy optimization decision-making.Secondly,according to the analysis of the user side demand response scenario,the user side load characteristics are studied,and the demand response behavior is characterized by the demand response benefit model.In the demand response benefit model,the demand response benefit coefficient represents the user’s response willingness to the incentive signal at different times and under different conditions.In order to get the accurate demand response benefit coefficient of users,two LSTM networks are constructed to predict the power consumption of users and calculate the demand response benefit coefficient of users,and a way to help LSTM network update parameters is proposed.The results show that the two LSTMs can complete the prediction correctly.Through the analysis of the demand response model,the user’s willingness to respond to the incentive signal and the ability to adjust the demand response of a single user can be directly reflected.Finally,the characteristics of distributed energy in the distribution network are studied,and the energy characteristics of wind power,photovoltaic and energy storage are analyzed.And the distributed energy optimal scheduling model is established,which takes the economic optimization as the optimization goal,while considering the operation cost,demand response cost and environmental cost.In addition,in order to solve the problem that there is no explicit analytical expression in the refined demand response model,the Deep Deterministic Policy Gradient(DDPG)algorithm in deep reinforcement learning is used to solve the optimal scheduling model proposed in this paper,forming an efficient decision-making method that can be used for real-time control.Through the example analysis,the refined demand response model can deeply mine the user side demand response ability.Compared with the actual incentive,the distributed energy optimal scheduling based on the refined demand response model can obtain better comprehensive economic benefits.