Development of a distribution network electricity supply regulation system using machine learning at PT. PLN (Persero) UP2D North Sumatra

Abstract

The regulation of efficient energy supply within the distribution network is a significant difficulty for PT. PLN (Persero) UP2D North Sumatra in delivering dependable and stable services. This project seeks to create an automated system for regulating power supply in the distribution network through machine learning techniques. This system aims to forecast and enhance the allocation of electrical energy utilising historical data and current network circumstances. The dataset is partitioned into training data (training set) and testing data (testing/validation set) in specific ratios, such as 70%:30% or 80%:20%. The division is executed to preserve the temporal sequence (in time series scenarios) or to ensure a balanced representation (in classification scenarios). The employed machine learning approach is K-Means Clustering, utilised to analyse electricity consumption patterns and identify probable problems in the distribution network. The new centroid computation is based on the fact that each cluster contains a single data point, specifically C1 = (193, 205, 213), C2 = (153, 167, 170), and C3 = (179, 196, 200). This study's findings are anticipated to enhance the efficiency of energy distribution management, minimise downtime, and elevate the quality of service for consumers. By using a machine learning-driven automation system, PT. PLN UP2D North Sumatra can enhance its adaptability to load fluctuations and optimise the utilisation of current power supplies. The division is executed to preserve the temporal sequence (in the case of time series) or to ensure a balanced representation (in the case of classification).