Research And Application On Load Forecasting Method Based On Grid

Load forecasting is the premise and basis of distribution network planning.Accurate load forecasting results have very important guiding significance for the construction of distribution networks.In the traditional load forecasting work,it is usually to analyze the historical load data of planning area to grasp the development law and growth trend of the regional load,then use the highly adaptive single load forecasting method to obtain the future regional load value.The specificity of each sub-region in the region has not been fully considered.This article builds a grid-based load forecasting system to solve this problem,selecting the highly adaptive model for prediction based on the results of grid division,with the data characteristics of each grid in the forecast area and the current load development level are fully considered.Based on the historical load data characteristics and load development levels of different grids,and combined with the differences in load forecasting years,this paper constructs a grid-based load total forecasting system for short-term and medium-term load forecasting of mature grids and developing grids,a grid-based space load forecasting system for long-term load forecasting of all grids,then combined the two to build a complete grid-based load forecasting system.The main improvements of this paper are:First,in the total load forecasting system,in addition to verifying the prediction accuracy of a single model based on the historical data of different types of grids,and using this as a basis to screen highly adaptable models,it also introduces the combined forecasting method further optimizes the results.Second,in the space load forecasting system,based on the space load forecast results of each functional area,the K-means clustering algorithm is used to cluster the functional areas to divide the unit.In the process of using the traditional K-means clustering algorithm to cluster the functional areas to divide the unit,there are two shortcomings,one is that the selection of initial clustering center is random,and the other one is that the distance was used as the only condition to cluster the functional areas.The improved K-means clustering algorithm was used to divide functional area clustering units.Analyze and compare the results obtained by the two improved methods of the initial clustering center.Finally,a clustering mode is used with the maximum load point as the initial clustering center,and the load factor is added to the clustering conditions,and the load distribution is relatively balanced.The results of unit division are used to guide the later planning work.