Study On The Demand-side Flexibility Resources Prediction And Application Based On Cluster Analysis

In March 2021,the central government proposed to build a new power system with new energy as the main body.New energy belongs to renewable energy,mainly including wind power and photovoltaic,which are the main body of the future power system.At the same time,demand-side flexible resources such as electric vehicles and temperature-controlled loads will also grow on a large scale,ushering in rapid development of energy digitalization.The power grid needs to carry out all work around data,realize its own quality and efficiency improvement and integrated development through data-driven,and enable the traditional power grid with informatization and digitalization.Therefore,this paper mainly discusses the application of data mining and artificial intelligence technology in the research of demand-side flexible resources.Firstly,this paper makes use of the advantages of principal component analysis and hierarchical clustering algorithm to improve the k-means clustering algorithm to improve its clustering performance.Principal component analysis can reduce the dimension of power data,and condensed hierarchy clustering can initially divide all objects and determine the initial clustering center,so as to make the clustering effect of k-means algorithm more stable.Curve clustering analysis is the basis of data mining in power system.This algorithm is used to cluster load curves and analyze the characteristics of various curves,which is helpful to further explore the application scenarios of data mining technology under the background of energy Internet.Then,this paper combines the improved curve clustering method with BP neural network to predict the load of the demand-side flexibility resources and improve the prediction accuracy.Flexibility resources usually has its own characteristics and influencing factors,when predicting the output of the temperature load such as air conditioning,to include the date of the meteorological environment clustering,to predict one day output is the flexibility of the load forecasting,delimit and meteorological characteristics of similar kind of cluster,it with the class cluster history daily load data as the input of the BP neural network model of the training sample,get the flexibility resource prediction curve.This method can reduce the interference of irrelevant samples,optimize the sample structure,and improve the accuracy of load prediction.Finally,with the construction of energy Internet,more and more distributed resources have the ability of regulation,and demand-side flexibility resources are easier to be aggregated.Virtual power plant,due to its characteristics of easy coordination and control and strong aggregation ability,can realize the efficient utilization of all kinds of resources and guarantee the economy and stability of power grid operation.Based on the proposed improved algorithm,this paper extracts evaluation indexes according to the curve features obtained by clustering,measures the aggregation effect of the virtual power plant on the demand-side flexibility resources through comprehensive evaluation,and determines the optimal aggregation scheme.