REGRESSION MODEL-BASED SHORT-TERM LOAD FORECASTING FOR LOAD DISPATCH CENTER

Abstract

In the wake of grid modernization initiatives, such as the integration of renewable energy sources and demand response, as well as the increasing popularity of electric vehicles, a greater degree of uncertainty has been introduced due to the fact that electricity demand has become more active and less predictable, and forecasting load has become increasingly difficult. Since the historical data is irregular, non-linear, non-smooth, and noisy, it is difficult to achieve a satisfactory result. The present study overcomes the challenges with the help of an improved energy demand forecasting model for load dispatch centers as part of an Artificial Intelligence (AI) driven project supported under research grants from the Department of Scientific and Industrial Research. A real-time hourly load consumption dataset was collected from Regional Load dispatch centers from July 1, 2020, to August 22, 2022. In this paper, 24 regression model-based day-ahead load forecasting algorithms are developed and evaluated using the load consumption and meteorological data collected from NASA Power (https://power.larc.nasa.gov/). MATLAB Regression Toolbox offers 24 regression models divided into five families: Linear Regression, Tree Regression, Support Vector Machines (SVM), Gaussian Process Regression (GPR), and Ensemble of Trees. Since GPR models are nonparametric kernel-based probabilistic models, they show the best load forecasting performance. The study recommends two GPR models for load forecasting: Rational Quadratic GPR and Exponential GPR.