Forecasting of Domestic and Commercial Power Demands Using Machine Learning for the LESCO Region in Pakistan

This paper presents a machine learning-based model for load forecasting by the Lahore Electric Supply Company (LESCO). The specific electrical load data in the LESCO region were trained through ML to achieve optimal predictive solutions through multiple training mechanisms. LESCO is the largest power distribution company in Punjab, Pakistan and handles both domestic and commercial loads. The main problem with LESCO is its manual system for handling load recording, which results in poor load management and consequent forced load shedding. The model proposed in this study works on real-time electrical data collection and the application of the Long Short-Term Memory (LSTM) algorithm for forecasting. The dataset comprises numerical load values over 20 years from major supply feeders in the Lahore city of the LESCO region. ANACONDA was used as a simulation platform to provide sophisticated and effective load forecasting through the training of recent data for the prediction of long-term future load demand. The examination of the accuracy and performance of the LSTM-based load forecasting model using acceptable criteria demonstrated its efficacy in gathering and forecasting complex load patterns. The traditional load monitoring mechanisms adopted by LESCO are less efficient, besides, no system for accurate load estimation is currently used by the distribution companies in Pakistan. The proposed LSTM based model can provide efficient load prediction addressing the issues of energy management and avoid forced load shedding.