Research On Algorithms For Power Load Anomaly Detection And Prediction Method Under Cloud Computing Environment

With the continuous improvement of power system informatization,more and more data are accumulated in the power grid system,forming the power big data.The analysis and mining of power big data plays an important role in the construction of smart grid and the efficient,safe and stable operation of power grid.Power load anomaly detection and prediction is an important research content of power big data mining.This thesis studies the method of power load anomaly detection and prediction and realizes it in the cloud computing environment.The achievements of this thesis are as follows:1.A parallel anomaly detection algorithm based on clustering by fast search and find of density peaks is proposed.In order to solve the problem of high time complexity caused by not considering the local characteristics of data and traversing the whole data set,the fast density peak algorithm is used to detect outliers.A rule is designed to judge outliers for power load.The algorithm divides the data set to be clustered into several data partition with relatively balanced data volume through spatial grid Spark parallel programming model is used to design parallel algorithm.The improved density peak clustering algorithm is used to detect whether the power load is abnormal in the corresponding data partition of each computing node.Then the abnormal detected by each node is combined and the real power load data set is used for experiments.The results show that the above algorithm can detect the power load data efficiently and accurately exception.2.A short-term load forecasting algorithm based on K-means radial basis function(RBF)neural network is proposed.Aiming at the problem of low prediction accuracy of radial basis function neural network data,this paper optimizes the training samples of radial basis function neural network and the network structure parameters such as radial basis function center,width,connection weight and number of hidden neurons.Using the prediction algorithm of radial basis function neural network based on K-means,different types of data,such as historical power data,date characteristics,etc.,are passed through the characteristics Extraction,data cleaning,building model for short-term power load forecasting,and parallel the K-means of RBF neural network algorithm,accelerate the training speed of RBF neural network,improve the efficiency of RBF neural network power load data forecasting.Using the existing power load data set for experiments,the experimental results show that the radial basis function neural network algorithm is superior to other algorithms in execution efficiency,and in the face of large-scale power load data set,the results show that it is effective and feasible.