Residents’ Electricity Behavior Analysis And Load Forecasting Based On Big Data

With the development of internet,internet of things,wireless sensors and other new generation of technology,accelerate the speed of smart grid construction.In the process of smart grid construction,advanced metering equipment,intelligent terminal equipment has also been installed and used by large,the electricity approach of residents tends to diversification.In-depth perception of the actual power consumption patterns of residents to improve the accuracy of load forecasting and to protect the normal operation of power systems,energy management and planning is essential.Firstly,the source of the residential electricity big data was analyzed,and points out that the residential electricity big data is faced with challenges in the storage,processing and so on,because of the large data volume,the complex type,the fast speed and the strong interaction.Then,One model with electricity behavior analysis of residents and load forecasting based on big data is proposed.The model uses smart meter data,meteorological data,holidays and other data as input data,and uses big data processing framework Spark which is based on memory calculation to data mining and analysis.Finally,the prototype system of behavior analysis and load forecasting based on big data is designed and developed.The system includes Spark cluster management,load data management,algorithm analysis,forecast result display and so on.The K-Means clustering algorithm based on Spark is used to realize the residential electricity pattern cluster experiment,the experimental results show that the model has a high accuracy rate for residential users.Compared with the traditional K-Means algorithm,the experimental results show that the K-Means algorithm based on Spark shows good performance and reduces the time of clustering execution and improve the accuracy of clustering with the continuous expansion of the data set size.And analyzes the behavior of electricity consumption for different categories of residents.Based on the above experiment,the load forecasting model is established for each type of resident user,and the load forecasting of each resident user is realized by using Spark-based multi-layer sensor neural network algorithm(MLP-NN)and SVM algorithm based on Spark.The experimental results show that MLP-NN has a high prediction accuracy.Two different datasets were used to verify the feasibility of the model,and the scale of the two datasets was extended.Twenty feature vectors were extracted from the smart meter data,meteorological data and holiday data of each class as the input layer data of the algorithm.The experimental results show that the proposed method can improve the accuracy of load forecasting to a certain extent and show better prediction effect in big data environment.