Short-term Power Load Forecast Based On Similar Days And Elman Neural Network

Short-term load forecasting is an important means to ensure the safety of power system and the stable operation of power economy.Accurate load forecasting can effectively improve the use of electrical energy and reduce unnecessary losses.It is very important for the smooth operation and development of the whole power grid and even the whole society.In the current energy shortage environment,Energy has become the core issue of research in various countries.Accurate power load forecasting can save more energy and can make energy use more effectively.This thesis starts with the analysis of short-term load characteristics,the power load and the related influencing factors at the following moment,and establishes a dynamic short-term power load forecasting model to obtain more accurate short-term power forecast.The thesis first introduces the background and significance of short-term power load research,the current research status of power load at home and abroad,and the application of neural network in short-term power load forecasting.The basic steps of short-term load forecasting are briefly described,including the determination of target and prediction scale,the acquisition of historical load data and influence factors,the preprocessing of short-term power load and influence factor data and the establishment of short-term prediction model.The network structure,principle and algorithm of BP neural network and Elman neural network are introduced.Then,on the basis of the analysis of the characteristics of short-term load,the thesis expounds the similar day algorithm and its prediction process,and combines the characteristics of the similar day algorithm and the Elman neural network.The thesis studies the prediction model based on the combination of similar day and Elman neural network.The influence of weather factors,date gap and week type on load is analyzed,and the correlation symbol factors are quantified.The calculation method of similarity degree is given,and the prediction process of the new model is illustrated.Finally,Based on the historical data of partial power load of a local power bureau,the three models are simulated respectively.First of all,the experimental environment and power data are described in general.The preprocessing of the data includes the complement of missing data,the processing of abnormal data and the normalization of the original data.In view of different day types and meteorological types,the three models were tested respectively,and the average absolute percentage error was used to evaluate the prediction effect of the three models.Experiments show that the model based on similar days and Elman neural network can show higher prediction accuracy in short-term power load forecasting.According to the experiment of different day types and weather types,the model has the feasibility of practical application,and the program based on this model can predict the short-term power load according to the set input.