A Demand Response Scheme In Smart Grid With Clustering Of Residential Customers

As a significant improvement of the traditional power grid,smart grid connects all components of the system through two-way power and information flow,showing obvious superiority to the traditional power grid in reliability,availability,efficiency,economic benefits and various other aspects.As an important part of the smart grid management technology,demand response can reduce peak demand,smooth power supply and demand curve,reduce the overall cost of the system and improve the stability of the system.The existing demand response schemes in the literature mainly focus on optimizing user’s load profiles but not enough attentions are paid to important factors such as energy consumption patterns of residential appliances,users’ satisfactory level,fairness and energy consumption habits.In order to tackle with the above issues,this thesis first proposes a demand response scheme based on clustering method,which aims to reduce the energy consumption cost of users and the peak to average ratio of the system.To decrease the huge computation burden in the above scheme,an optimization method based on equal cluster size clustering algorithm and deep learning is then proposed.The main contributions of this thesis are as follows:(1)In view of the diversity of users’ demand,this thesis proposes a centralized residential demand response scheme for reducing the cost of users’ energy consumption and the peak to average ratio of the system.The scheme comprehensively considers the energy consumption pattern of residential appliances,power consumption,users’ satisfaction level and fairness.The scheme extracts new features from users’ historical data and then utilizes such information to classify users into several clusters.Based on the fact that users in different clusters have different behavior tendencies,the demand response scheme becomes more flexible and effective.Simulation results show that the proposed demand response scheme performs well in the management of residential appliances,saving users’ expenditures and reducing the peak to average ratio of the system.(2)In order to solve the huge computation burden problem in the above demand response scheme,an optimization method based on equal cluster size clustering algorithm and deep learning is proposed.This method uses equal cluster size clustering algorithm to classify users into several smaller equal size clusters,and then executes the above-mentioned demand response scheme in the small clusters to obtain the energy consumption data of each consumer after demand response scheduling.The data of small cluster users is used to recalculate in large cluster to get the final energy consumption data.This data is adopted to train a long short-term memory network for each user to predict the energy consumption after demand response scheduling and then modify the prediction results.This prediction data is adopted by users to calculate the fairness.It decomposes the centralized solutions for multiple users into multiple single user solutions,which significantly reduces the computational complexity.This method could effectively reduce the power cost of users and the PAR of the system.Although it sacrifices a small amount of accuracy(less than 1%)compared with the optimal scheme,it significantly reduces the computational complexity and greatly enhances the application range.