Sector-Wise Electricity Consumption Analysis at District level in Bangladesh Using Machine Learning: Patterns, Trends, and Predictions

Abstract

In this paper, I have examined electricity consumption patter and segmentation in Manikganj which has achieved full electrification but is still struggling with the issue of arrears. ‘Manikganj Palli Bidyut Samity (MPBS) dataset’ and various machine learning techniques, such as regression, classification, clustering, anomaly detection and causality, are applied in order to discover the patterns of electricity usage, consumer duties and support the efficiency and the awareness. XGBoost was the best in predicting (≈0.84) the variance of electricity but not the clusters hung high capacity low usage and low capacity high dues consumer. The Anomaly detection resurged an anomalous customer usage pattern which can indicate a suspective fraud or an inefficiency, whereas the causal effect proposed that the digital meter was associated with lower dues?.. as against analogue meter. Based on these findings, we have developed a web-based interface that envisages two views: The administration panel for back end support staff at the district to evaluate and regulate the data, and the user panel, where the customers can sign in to check their line capacity, the meter type, the usage report and payment clearance while a guest get access to the district-wide electricity information for educational learning and teaching as a learning resource. In sum, the studies and system exemplify the potential for data-driven intelligence and virtual tools to shed light on utility management and rural energy use and practices, make visible and transparent rural energy usage, and raise public awareness of both imaginary and actual rural energy use.