Load Forecasting for Dhaka City Using RNN, LSTM, and GRU Architectures with Meteorological and Temporal Data

This paper develops a machine learning model to predict the daily maximum load demand in megawatts (MW) for Dhaka city. By leveraging historical time-series data alongside meteorological and temporal factors, the model employs advanced Recurrent Neural Network (RNN) architectures, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) layers. These architectures are crucial for capturing both the short-term fluctuations and long-term trends in electricity demand. Our dataset integrates critical weather parameters such as temperature, humidity, wind speed, and precipitation, as well as significant temporal factors like government holidays, which are known to affect electricity consumption patterns significantly. Through meticulous data pre-processing, which includes scaling, feature engineering, and the application of a sliding window technique, our model can recognize intricate patterns in the data. We incorporate techniques such as batch normalization and dropout to prevent overfitting and enhance the model’s training efficiency. The results demonstrate the model’s capability to deliver highly precise load forecasts, which are essential for utility companies in planning resources, reducing operational costs, and improving overall energy management in Dhaka. This paper highlights the creation of a specialized dataset and a robust deep learning model, offering significant contributions to the enhancement of modern electrical power management systems.