Short-Term Power Load Forecasting Based On IWOA_II-ELM

Short-term load forecasting provides important data support for the power system planning,dispatching,and power market transaction,which is related to the stable operation and economy of the power system.However,with the integration of the high-penetration renewable energy generations in China,the situation of the power supply and demand presents significant randomness,and the uncertainty of the power system supply and demand is further enhanced.The short-term load forecasting faces new challenges.The accuracy of load forecasting needs to be further improved,and the traditional forecasting methods cannot meet the requirements.In addition,with the deepening of the reform of power supply side and power selling side,as well as the promotion of the spot market,the "source-load" interaction is increasingly significant,and the complex coupling behavior between load demand and real-time transaction price is formed.The real-time electricity prices have gradually become one of the indispensable factors for the short-term load forecasting.If this influencing factor is ignored at this time,the accuracy of the load forecasting results will be greatly reduced.This dissertation focuses on the research of the high-precision short-term load forecasting with the real-time electricity price,mainly including: the analysis of influencing factors of short-term load forecasting and the selection of model inputs,the preprocessing of model data,and the data-based short-term load forecasting models and the comparison and verification of models.(1)Influencing factors analysis and the model input selection of the short-term load forecasting.Considering the influence of the real-time electricity price on the short-term load forecasting,the real-time electricity price is regarded as one of the input variables of the forecasting model.(2)Preprocessing of the model data.The K-Means clustering algorithm is used to cluster the original data according to the real-time electricity price,and the similar day category data of the prediction day is used as the training sample,which lays the foundation for improving the prediction accuracy of the model;By using the principal component analysis(PCA)method to reduce the input variables of the model,the problems such as the increase of input variables dimension,the more complex structure of the model,and the increase of training time are solved after considering the factors of the real-time electricity price.(3)Short-term load forecasting model based on the improved incremental extreme learning machine(II-ELM).In order to reduce the complexity of the short-term power load forecasting model and improve the real-time performance of the model,an offset is added to the hidden layer output of incremental extreme learning machine(I-ELM)network model to optimize the output weight.(4)Short-term load forecasting model based on the improved whale optimization algorithm for improved incremental extreme learning machine(IWOA_II-ELM).In order to further overcome the shortcomings of random acquisition of input weights and hidden layer thresholds of the II-ELM,the improved whale optimization algorithm(IWOA)is used to optimize the initial weights and thresholds of the II-ELM,and the IWOA_II-ELM model is developed which is integrated the IWOA algorithm and II-ELM.Finally,the load data of a region from September 1,2009 to August 31,2010 are used as load forecasting samples for case analysis,and compared with the I-ELM,II-ELM,and whale optimization algorithm for improved incremental extreme learning machine(WOA_II-ELM).The results verify the effectiveness of the improved neural network model.