Performance Analysis of Different ANN-based Weight Updating Algorithms in Forecasting Short-Term Load Demands in Cluster Microgrids

Abstract

In today's deregulated energy markets, artificial intelligence (AI) and machine learning (ML) methods are frequently employed for estimating short-term load demand. Out of all the AI/ML techniques available, Artificial Neural Networks (ANN) provide different features such as controller design, analysis of regression and forecasting, etc. The efficacy of ANN depends on how well the network is trained. So, the proper selection of training algorithm is required for effective load demand forecasting in the system considered. So, with this intent in this paper, the performance of different ANN-based weight updating algorithms such as Levenberg-Marquardt (LM), BFGS Quasi-Newton (BFGS-QN), and Resilient Back Propagation algorithms (RBP) are investigated to forecast the load demands on 24-hours basis in a microgrid cluster. All the simulations are done in “MATLAB/Simulink 2021a software”. To validate, the performance indices of all the algorithms are measured and compared, thereby the superior algorithm is recommended.