Research On Electro-load Forecasting Based On The Improved Neural Network By Water Wave Algorithm

Electro-load forecasting and distribution is the basis of transmission line design and substation layout.With the rapid development of electricity industry,electro-load planning has become one of the important indexes of power system stability.However,user-level load fluctuations are severe,random,and difficult to predict,so load forecasting based on user-level is very difficult.The user-level electrical load value is not only susceptible to environmental factors such as weather,temperature,and rainfall,but also susceptible to external uncertainties such as holidays and economic production.Therefore,the key point of accurate prediction of electro-load is to select appropriate load influence factors and establish a suitable model.In this dissertation,a electro-load forecasting model based on water wave optimization algorithm(WWO)and improved radial basis function neural network(RBFNN)is designed to solve the problem of inaccurate electro-load forecasting based on user-level.By optimizing the hidden layer center of RBF neural network and expanding constant parameters,It is further verified that the improved RBF neural network based on the water wave optimization algorithm can achieve better results in the electro-load forecasting based on the user-level.The main contents are as follows:1.Improving the defects of fuzzy c-means algorithm(FCM).By learning the algorithm model of fuzzy clustering,the realization principle,algorithm features and other knowledge.There are two weaknesses in FCM fuzzy clustering: First,it takes a long time to process large data sets.Second,it is very sensitive to initialization and easy to fall into local minimum.The global search ability of the water wave optimization algorithm is used to make FCM jump out of the local extreme value and speed up the accuracy and speed of clustering.2.Improve the determination of the center of the hidden layer nodes of the RBF neural network and the problem of randomly determining the spreading constant.An improved RBF neural network electro-load forecasting model based on WWO-FCM is designed and implemented.The real data is processed and the factors affecting the electro-load forecasting are analyzed.It solves the problem that the hidden layer parameters of RBF neural network need to be determined according to the actual problems in electro-load forecasting based on the user-level.When the model designed in this dissertation is verified on the UCI standard data set,the forecasting effect of this dissertation is better and the precision is higher;When the model is used for user-level's electro-load forecasting,the prediction average error is lower and the prediction stability is stronger.At the same time,the proposed model network training time is shorter and the convergence speed is faster.The forecasting ability can meet the needs of the power supply enterprise to plan and schedule the electro-load forecasting.The prediction effect is reliable and practical,and it is superior to the traditional forecasting model and can be widely applied to the actual electro-load forecasting system.