Applying Innovative Models for Forecasting Small-Area Peak Electrical Loads

Abstract

The number of Distributed generators is currently increasing, and the electrical industry is trending toward regional supply-and-demand and resource integration. Thus, a model that can forecast small-area peak electrical loads is an indispensable part of power infrastructures. This study constructs innovative model for forecasting small-area peak electrical loads. The main aspects considered were the accuracy of the forecasting model and the convenience of follow-up maintenance and management of the model and data. This study used yearly peak load value and total power data from substations to construct regression tree models. These acted as models for the small-region peak electrical load of substation districts in the Taipower distribution systems. The errors of these forecasting models were substantially smaller than those of the least squares model originally used by Taipower to forecast peak load. The addition of exogenous factors was unnecessary. Additionally, our results were superior regardless of whether once or incremental models were adopted for the data. This confirms the usability of our models.