A Hierarchical Data Driven Method for Spatial Electric Load Forecasting

Spatial Load Forecasting (SLF) is the basis for distribution network planning. Most of the previous SLF methods, including empirical load density index method and the top-down simulation method, fail to give sufficient accurate forecasts on the quantity and distribution of the future load growth. Motivated by the increasing data sources and widely fulfilled land use planning in cities, a data driven approach for SLF is proposed in this paper. Firstly, a hierarchical bottom-up load accumulating approach is utilized based on clustered typical load profiles. Then land plots are described with multiple indices including intensity, development, climate and location, and the numerical spatial load densities of different land use types are obtained through a nonparametric kernel density estimation. Finally, a machine learning method is introduced to forecast the future plot load. The proposed data driven approach can greatly enhance the accuracy and operability of SLF for distributed network planning.