Research On Forecast Of Electricity Demand Of Electric Power Customers

With the reform of the power system,power customer demand forecasting is a very important basic work in power marketing activities.The power demand of customers can help the power grid understand customers’ individualized and differentiated service needs and conduct two-way interactive services.The power company uses interactive channels to remind customers to pay in time.Customers respond to the company’s payment reminders to make payments to avoid arrears and blackouts.The company and electricity customers can form a good and effective electricity tariff reminder and payment interaction mechanism,thereby improving the company’s service level and customers Satisfaction.For large users and users who are at risk of arrears,the power company can accurately predict the future electricity bills of users by accurately predicting their future electricity demand.The company uses interactive channels to push the company’s forecast results on users’ electricity bills,and customers respond The company’s forecast results,reasonable arrangements for production and business plans,so as to avoid arrears.Therefore,accurate power customer demand forecast results will promote good interaction between users and the power grid.This article first processes the electricity demand data of power customers,and detects,eliminates and corrects bad data based on the abnormalities,defaults,and negative values of the data;secondly,it analyzes the factors that influence the electricity demand of users,Feature division and aggregation of similar days,that is,considering the impact of weather conditions,economic conditions,and holidays on user power demand,users are divided into three categories,urban users,commercial users,and industrial users,and k-means similar days are aggregated.Classes,using the COP index to analyze the number of different clusters,and get the best number of clusters and clustering results.First,construct the user demand prediction model based on the user’s different categories and date characteristics;then,according to the user’s electricity demand characteristics,comprehensively considering the user’s data characteristics,forecasting duration and model characteristics,a parallel Elman extreme learning machine model is established;and finally,In order to find the model corresponding to the forecast date,the DTW algorithm is used to match similar dates and predict different load characteristics.The prediction results show that the established prediction model is practical and and feasibility for the prediction of user power demand under massive data.