About the project

Bangladesh has expanded its generation capacity and grid coverage substantially over the past two decades. The sector still faces rising generation costs, growing dependence on imported fossil fuels and load shedding at peak demand. Much of this is made worse by a lack of real-time information: plant performance is reviewed after the fact, and decisions on capacity, tariffs and power purchase agreements rest largely on static models and manual assessment.

Elsewhere, artificial intelligence and data analytics are already used to forecast demand, detect faults and losses, and run power systems more efficiently. In Bangladesh these methods have so far stayed in academic studies and small pilots. Power-Net AI sets out to close that gap with a platform built for, and tested in, the country's own power plants.

The project, led by the Department of EEE at Daffodil International University and funded by the Bangladesh Energy and Power Research Council, will build a web-based platform with three connected parts. The first is remote monitoring: devices installed at the plant measure voltages, currents, power, power factor, frequency and harmonics, and AI models flag abnormal behaviour as it happens.

The second is generation cost modeling. Instead of fixed spreadsheets, cost per unit is calculated from fuel, maintenance, capital and operating data and updated as conditions change, with machine-learning models used to predict costs and test scenarios. The third is decision support: forecasts of generation and demand, and recommendations that help operators plan dispatch and maintenance and reduce load shedding.

What makes the approach distinctive is that monitoring, costing and decision support share one data foundation. The same measurements that raise an alarm also feed the cost model and the forecasts, so operators, planners and regulators work from one consistent picture instead of separate reports. The platform is designed with BPDB, which provides the pilot sites and validates the results, and with the Bangladesh Computer Council, which supports data hosting, API standards and training, so that it fits the way the sector actually works.

The platform is piloted at three BPDB power plants, one each running on coal, oil and gas. The pilot covers hardware and software integration at the plants, validation of the forecasting, anomaly-detection and cost models against live operating data, and an assessment of accuracy, responsiveness and scalability. The aim is a proven framework that can later be extended to other plants and to transmission and distribution networks.

Building local skills is an objective in its own right. Through trainer programmes, hands-on stakeholder trainings and national workshops, engineers from generation, transmission and distribution companies, and students from universities and engineering colleges, will learn to use and maintain AI tools for the power sector.

Throughout the project, graduate students and research assistants work on the data and the models. They learn from real plant operation, not textbook examples, and the knowledge stays in Bangladesh after the project ends.

Better information should lead to better operating decisions: faults caught earlier, plants run closer to their most economical point, and maintenance planned rather than reactive. Automated checks reduce the room for human error. Because the hardware and software are developed locally, the sector depends less on imported technology.

Shared, consistent cost data also makes investment and procurement decisions easier to scrutinise, which supports a level playing field for public and private producers. Together, these outcomes support a more secure, affordable and sustainable energy supply for Bangladesh.