Cluster Analysis Of Controllable Load And Optimal Scheduling Strategy Considering Space-time Distribution Characteristics

The high proportion of renewable energy and a large number of demand side resources connected to the power grid aggravate the bilateral uncertainty of source load.In this context,controllable load resources can maintain the balance between supply and demand of the power grid and operate safely and stably by participating in demand response,so as to improve the movement flexibility of the system.However,when participating in power grid regulation,the time variability of controllable load and the uncertainty of its spatial location distribution bring difficulties to further tap the potential of load side resources.Therefore,this thesis comprehensively considers the spatial distribution and time characteristics of controllable load,excavates the common characteristics of controllable load through reasonable power load characteristic clustering,analyzes the dynamic characteristics of controllable load,and grasps the power consumption law of users.On this basis,this thesis puts forward the optimal regulation strategy with power grid stability and economy as the comprehensive index.The main research contents are as follows:(1)Clustering algorithm based on controllable load spatial distributionThe spatial distribution of controllable load affects the accuracy and economy of load forecasting and optimal dispatching,which is an important factor that can not be ignored in load management and control.Therefore,aiming at the actual spatial distribution of controllable load,this thesis proposes a spatial density clustering algorithm based on the spatial characteristics of load.In the process of clustering,combined with the idea of grid division,first divide the sample data into regions,and then determine the parameters of spatial density clustering algorithm through kernel density estimation method,so as to reduce the time complexity of the algorithm and the subjectivity in the process of parameter determination,overcome the inherent defects of DBSCAN algorithm and better display the spatial distribution attribute of controllable load.(2)Clustering algorithm based on controllable load time characteristicsWith the rapid advancement of ubiquitous sensing technology in power grid,power data gradually presents the characteristics of massive,high-dimensional and diversified.Direct processing and analysis will face the restriction of computing efficiency.Therefore,aiming at the high-dimensional daily load curve,this thesis proposes a controllable load clustering algorithm based on load time characteristics.Firstly,the principal component analysis method is used to reduce the dimension of the controllable load daily load curve;Then the adaptive function is introduced to automatically determine the optimal number of clusters,avoid the subjective blindness of fuzzy c-means clustering algorithm,improve the time efficiency of the algorithm,and can still effectively extract the typical daily load curve under the background of increasingly massive power data.(3)Controllable load distributed optimization scheduling strategy based on multi-intelligenceThrough reasonable cluster analysis,we can mine the power consumption information reflected in multi-source heterogeneous power data,extract the common characteristics of controllable load,and provide theoretical support and guidance for load forecasting and optimal dispatching strategy.Therefore,according to the spatio-temporal clustering results of controllable load,a spatial load forecasting based on load density index is proposed.Combined with the load clustering algorithm,a more accurate load density index is established to improve the accuracy of spatial load forecasting;A controllable load optimal scheduling strategy based on multi-agent consistency is proposed.By taking the compensation cost function of controllable load polymer as the consistency variable,the controllable load demand response model is established,and the optimal regulation with power grid stability and economy as the comprehensive index is realized.