Electrical Load Profile Analysis And Load Forecasting Technology Based On Data Mining

With the rapid development of advanced measurement systems for power systems and the construction of smart grids,the electric utilities currently collected a large amount of raw power load data,which contains a wealth of valuable information.Therefore,it is of great significance to fully exploit the effective information in the massive load data.The electrical load profile analysis and load forecasting of power users are the most basic and most difficult to accurately model the two traditional problems in power system planning and operation.Under the background of the current smart grid and big data of electric power,new challenges and requirements are put forward to these two traditional problems.How to use the new generation of data mining technology and artificial intelligence technology in the user load profile analysis and load forecasting to improve the refined load management of the electric utilities and the development of better differentiated power supply strategies are of great significance.Firstly,this paper introduces the data sources and data quality of load data,weather data and electricity price data used in this study,and focuses on data preprocessing of various types of raw data,including load data cleaning,weather data up-sampling and data normalization,etc.The above work provides a data basis for the user load profile analysis and load forecasting later.Then,a user load profile analysis method based on time series data mining is proposed,which can effectively extract the typical load profile of a single user and analyze the load profiles of different types of users.First,based on the Piecewise Aggregate Approximation(PAA)algorithm and the Symbolic Aggregate appro Ximation(SAX),the user’s daily load data is reduced and re-expressed,and a string of character strings is used to represent a daily load curve,and then the user’s most frequent load pattern is screened out to obtain the user’s typical load profile;Then extract the time series features of the user load profiles,and combine the traditional load shape indices as the input feature of the clustering algorithm,and finally analyze the load profile of different types of users based on the K-means algorithm.Next,this paper proposes novel algorithm which is based on the stacking model fusion algorithm to forecast the short-term load of classified users.It takes into account the differences between different types of users,in order to improve the accuracy of user load forecasting.First,based on the maximum information coefficient and Pearson correlation coefficient,the relationship between different types of users and various influencing factors,including peakvalley time-of-use electricity price factors and meteorological factors,are studied to provide a basis for the features selection of load forecasting models;Then,a stacking model fusion algorithm based on the short-term load forecasting of different types of users is proposed,and different input features and optimal model parameters are selected in consideration of user differences,which can effectively improve the accuracy of user load prediction.Finally,a set of load characteristic analysis software system is developed based on the above theoretical research foundation.This system is a set of load characteristic analysis software system with strong applicability and flexible expansion based on Java language development,with micro-services and micro-applications as the core architecture.It can provide some important reference information for power grid planning and construction to realize the intelligent and refined management of the electric utilities.