Research On Optimal Model Of Power Customer Behavior Based On Clustering Algorithm

Facing the rapidly rising domestic energy service market,how to use existing customers,technology,data and other advantages,through data mining and analysis,to build a high-quality and high-viscosity customer-centric energy efficiency service system with value-added and energy-saving as the core The current business operation model and the creation of a competitive business with strong market influence are urgently needed by the current comprehensive energy company for research and solution.Based on the existing data assets,this paper adopts the cross-industry data mining standard process to analyze and apply the massive real-time energy consumption data on the demand side.Based on the analysis of potential and high-quality energy customers,it builds customer electricity behavior analysis and The prediction algorithm can optimize the electricity consumption behavior of peak-shaving and valley-filling for customers with energy storage conditions,and form an overall and comprehensive comprehensive energy management solution.First,based on existing internal data(marketing,metering,production,etc.),combined with electricity market transaction rules,industry background,GDP growth trends,and related crawler data,three levels of electricity consumption,power factor,and electricity consumption are defined Indicators,and refined into nine secondary indicators to describe customer characteristics.Aiming at the shortcomings of K-means clustering algorithm which needs to determine the clustering a priori,the Auto-Kmeans algorithm for automatic clustering is improved based on the Davidson-Bouding Index(DBI).The clustering results show that the optimal number of categories is 4,and all four categories are representative,realizing the classification,grouping and customer portraits of existing customers.Secondly,on the basis of customer classification,combined with historical and forecast data such as load,output value,weather,events,etc.,the electricity consumption behavior of the first-class customers,that is,high-quality customers,in different periods are analyzed and load forecasted.Combined with the characteristics of high-quality customers’ electricity consumption behavior in different periods,linear function,exponential function,and quadratic function fitting are used to achieve annual load forecast through cumulative generation.Based on the customer’s electricity plan,the envelope load forecast method is used to achieve monthly load forecast.The workdayclassification and fitting algorithm implements daily load forecasting.The results of the calculation examples show that the prediction algorithm model can provide reliable data support for the optimization research on the electricity consumption behavior of high-quality customers.Finally,based on the above customer classification and load forecasting model,this paper uses the existing main distribution network,dispatching,and marketing related data to build a peak-filling electricity consumption optimization model for high-quality customers with energy storage conditions.The model takes the charge and discharge power of the energy storage system,the capacity of the energy storage system,the customer load curve and the consistency of charge and discharge as constraints,constructs a function with the lowest electricity cost as the goal and finds the optimal solution.The analysis of the example shows that without changing the previous production plan of the enterprise and ensuring the normal supply of the load,the model of this paper can be combined with the customer’s electricity preference to give a better energy-saving practice for the customer.