Micro-Spatial Electricity Load Forecasting Using Clustering Technique

Abstract

Low growth of electricity load forecast eliminates cost opportunity of electricity sale due to unserviceable load demands. Meanwhile, if it is exorbitant, it will cause over-investment and incriminate investment cost. Existing method of sector load is simplified and easy to implement. However, the accuracy tends to bias over one area of which data is limited and dynamic service area. Besides, the results of its forecast is macro-based, which means it is unable to show load centers in micro grids and failed to locate the distribution station. Therefore, we need micro-spatial load forecasting. By using micro-spatial load forecast, the extrapolated areas are grouped into grids. Clustering analysis is used for grouping the grids. It generates similarity matrix of similar data group. Clustering involves factors causing load growth at each grid; geography, demography, socio- economic, and electricity load per sector. Results of every cluster consist of different regional characteristics, which later the load growth is projected as to obtain more accurate forecast.