Sliding Window Regression based Short-Term Load Forecasting of a Multi-Area Power System

TL;DR

This paper introduces a sliding window regression method for short-term load forecasting in multi-area power systems, addressing load demand variability due to distributed generation and outperforming existing techniques.

Contribution

A novel dynamic load forecasting model using sliding window regression that adapts to changing load patterns in multi-area power systems.

Findings

The proposed method outperforms four existing techniques in accuracy.

Validation on New York ISO data demonstrates improved prediction performance.

The model effectively captures the dynamic nature of load demand.

Abstract

Short term load forecasting has an essential medium for the reliable, economical and efficient operation of the power system. Most of the existing forecasting approaches utilize fixed statistical models with large historical data for training the models. However, due to the recent integration of large distributed generation, the nature of load demand has become dynamic. Thus because of the dynamic nature of the power load demand, the performance of these models may deteriorate over time. To accommodate the dynamic nature of the load demands, we propose a sliding window regression based dynamic model to predict the load demands of the multiarea power system. The proposed algorithm is tested on five zones of New York ISO. Results from our proposed algorithm are compared with four existing techniques to validate the performance superiority of the proposed algorithm.