An artificial intelligence-driven model developed by Halima Haque, graduate student at Independent University, Bangladesh, can help authorities predict power consumption trends in Dhaka, optimise electricity distribution – particularly during peak summer demand – and reduce load shedding, said a press release.
Halima, a graduate student at the Department of Electrical and Electronic Engineering (EEE) at IUB, came up with the machine learning (ML) model after an extensive study on energy demand forecasting for Dhaka City.
The research, supervised by Professor Md Abdur Razzak from IUB’s Department of EEE, leverages over 6.5 million real-world electricity consumption data points collected from Dhaka Electric Supply Company Limited (DESCO) between 2020 and 2023. It examines electricity usage patterns across residential, commercial, and industrial sectors, incorporating factors such as weather conditions, tariff categories, and consumption trends.
‘With Dhaka facing rising energy challenges due to rapid urbanization and industrial expansion, accurate demand forecasting is crucial for ensuring stable power distribution, minimizing shortages, and integrating renewable energy into the national grid,’ said Halima.
The study tested five machine learning models – Multiple Linear Regression (MLR), K-Nearest Neighbors (KNN) Regression, Random Forest (RF), Light Gradient Boosting Model (Light-GBM), and Extreme Gradient Boosting (XGBoost). The KNN model delivered the highest prediction accuracy, achieving an R² value of 72% with the lowest Root Mean Square Error (RMSE) of 222 kWh. Light-GBM and RF also showed strong performance, making them viable alternatives for energy forecasting.
‘The ability to predict energy demand with high accuracy can significantly improve power distribution strategies and support renewable energy integration,’ said Halima Haque. ‘By utilizing AI-driven forecasting, we can help policymakers and utility companies make informed decisions that ensure a stable and sustainable power supply.’
The findings, presented at IEEE conferences in the Maldives and Thailand and published in the Q1-ranked IEEE Transactions on Industry Applications, highlight the potential of AI-driven forecasting to optimize power distribution, enhance grid management, and support smart grid technology.
‘This research is an important step toward modernizing Bangladesh’s energy infrastructure,’ said Professor Razzak. ‘Machine learning models can help us better understand consumption trends and develop more efficient policies for energy management, reducing shortages and making way for smarter, more sustainable urban planning.’
The research provides actionable insights for policymakers and energy utility companies to improve energy planning, reduce shortages, and enhance sustainability. By applying machine learning to energy forecasting, Haque’s work offers a scalable solution for addressing Bangladesh’s growing power demands. The study underscores the role of data-driven decision-making in ensuring a sustainable and efficient energy infrastructure for the country’s urban and industrial sectors.