Abstract
This study addresses the challenge of optimizing power distribution during load reduction periods within the electrical grid, focusing on the integration of machine learning and advanced smart grid technologies. With increasing demands on energy infrastructure, traditional load reduction protocols often lead to inefficient power allocation and disruptions in service. This research proposes a model that employs machine learning to analyse household consumption patterns and dynamically adjust power allocation, providing a more efficient and reliable load reduction strategy.
The study involved collecting data from Eskom’s load reduction protocols and using simulated household power usage data obtained in MATLAB® to reflect real-world consumption behaviours. A MATLAB® and Simulink model was developed, incorporating time series forecasting techniques like ARIMA and decision trees, these techniques predict power consumption and distribute available power among households based on need rather than fixed limits, improving the efficiency and fairness of load reduction.
The model was evaluated through various load reduction scenarios, analysing its performance in optimizing power distribution, mitigating overload risks, and reducing blackouts. Results show that the proposed machine learning model can significantly enhance power distribution during load reduction, contributing to a more resilient and responsive grid. This research offers a promising approach for utilities to manage demand flexibly and equitably, ensuring both operational efficiency and customer satisfaction in future smart grid deployments.