Revision as of 09:56, 27 November 2025

Abstract

Load modelling is a crucial element of power system study that significantly affects the field’s planning, operation, and control methods. With the increasing penetration of renewable energy sources, electric vehicles, demand-side management, and distributed generation (DG), the traditional static and dynamic load model approaches are being replaced. This paper reviews extensively the existing load modelling techniques, namely, component-based load modelling, measurement-based load modelling, and hybrid methods. In addition, advancements tuned by artificial intelligence (AI) and machine learning (ML) are critically reviewed, emphasizing improving the accuracy, flexibility, and real-time adaptability of load models. For instance, Long Short-Term Memory (LSTM) networks have demonstrated significant improvements in forecasting accuracy, while Reinforcement Learning (RL) techniques enable adaptive and real-time control of load dynamics. Special focus is laid on load modelling in conditions of imbalance, dynamic parameter identification, and integration with smart grids and active distribution networks (ADNs). The review also discusses the importance of uncertainty embedded in probabilistic and data-driven models, customer behaviour, and the stochastic nature of distributed energy resources (DERs). The areas of future study emphasized AI-assisted adaptive architectures, hybrid frameworks, and digital twin applications for resilient and intelligent load modelling.OPEN ACCESS Received: 01/08/2025 Accepted: 26/09/2025 Published: 27/11/2025

Document