BengalGrid: A Decade-Scale Hourly Dataset of Bangladesh Power Supply by Source Type, Including Cross-Border Imports from India and Nepal

Electricity is the invisible current that powers modern life, flowing from power plants through a vast network to lights, factories, and homes. For engineers and policymakers to keep this system running smoothly, they need a clear, continuous record of exactly how much power is being made, where it comes from, and when it is needed. This record is called a time series, a simple list of numbers that tells a story over time. Without such a list, it is impossible to predict future needs or understand how the mix of energy sources—like gas, coal, or wind—is changing. However, in many parts of the world, this vital information is scattered, hidden in formats meant for human eyes rather than computer analysis, or simply missing for long stretches of time.

In Bangladesh, a nation with a rapidly growing economy and a complex energy landscape, such a clear record was missing until recently. The country's power grid relies on a mix of domestic fuel sources and electricity imported from neighboring countries, but the official data was locked away in a public website that presented information in a way that computers could not easily read. The dates were written in a different script, the numbers were inconsistent, and the data was broken into thousands of tiny pages. To study the country's energy future, researchers needed to unlock this data, clean it up, and turn it into a single, reliable timeline.

A team of researchers from East West University in Dhaka has now done exactly that. They created a new, open dataset called BengalGrid, which offers a decade-long, hour-by-hour view of Bangladesh's power supply. By writing a computer program to systematically visit the official government portal, they collected over 104,000 records spanning from April 2015 to October 2026. This massive effort transformed a chaotic collection of web pages into a clean, organized file that anyone can use to study how the country generates and imports electricity.

The process began with a digital sweep of the Power Grid Company of Bangladesh's public website. The researchers wrote a script to visit every page of the site, one by one, mimicking a human visitor to avoid overwhelming the server. They extracted the tables containing generation data, which listed the total power output and the breakdown by fuel type, such as natural gas, liquid fuel, coal, hydro, solar, and wind. They also captured data on electricity brought in from India and Nepal through specific cross-border connections. The raw data they gathered was messy. The dates and times were written in Bengali numerals, which computers do not recognize as numbers, and the time formats were inconsistent. Some entries were duplicates, and others had missing information.

To make the data usable, the team built a second program to clean and organize it. They converted the Bengali numerals into standard digits, fixed the time stamps so they all followed the same format, and removed the duplicate entries. In total, they discarded 438 duplicate rows and a handful of records with impossible dates, leaving a final set of 104,134 unique hourly records. Crucially, they decided not to fill in the missing numbers or fix the strange outliers they found. Instead, they left the gaps and the odd values exactly as they appeared in the source, trusting that future researchers would want to see the raw truth of the data and make their own decisions about how to handle it. This approach ensures that the dataset remains a transparent mirror of the official records, rather than a polished version that hides potential errors.

Once the data was ready, the researchers spent months checking its quality to ensure it told a true story. They verified that the total amount of power reported for any given hour matched the sum of the individual sources, such as gas and coal, plus the electricity imported from neighbors. In nearly all cases where all the numbers were available, the total matched the sum of its parts perfectly, with a difference of less than one megawatt. This internal consistency confirmed that the data collection process had preserved the relationships between the different columns of information. They also checked if the data behaved like a real power grid should. They found that electricity generation followed a predictable daily rhythm, dipping in the early morning, rising during the day, and hitting a distinct plateau in the evening when demand peaks. They also saw a weekly pattern, with power usage dropping noticeably on Fridays, which is the weekly holiday in Bangladesh, and a seasonal pattern where supply is higher in the hot summer months and lower in the winter.

The dataset reveals a significant shift in how Bangladesh powers its growth. Over the ten years covered by the data, the country's total electricity supply has grown steadily. The researchers observed that while natural gas has traditionally been the dominant source, its share of the total mix has begun to decline. At the same time, coal has surged, growing from a tiny fraction of the supply to a major contributor that now rivals gas in volume. The data also shows the rising importance of cross-border trade. Electricity imported from India, particularly through a high-voltage connection called Bheramara, has become a stable and essential part of the grid. Newer connections, including lines from Adani and from Nepal, appear in the later years of the dataset, reflecting the country's expanding energy partnerships.

The researchers also documented the limitations of their work, ensuring users know exactly what they are looking at. They found that some newer sources, like wind power and solar energy, have large gaps in the early years of the dataset because they were not being reported at that time. The data also contains a few extreme values that seem physically impossible, such as a single record showing a total generation of over 64 million megawatts, which is far beyond what the entire country could ever produce. These are likely errors in the original reporting, and the researchers have left them in the file so that users can identify and handle them. Similarly, the data shows that the country's power supply is not just about how much is made, but also about when it is made, with clear peaks in the evening that require careful management.

By releasing this dataset along with the computer code used to create it, the team has provided a foundation for future research. Scientists can now use this data to forecast future energy needs, study the environmental impact of switching from gas to coal, or analyze how cross-border trade affects national security. The work turns a fragmented, difficult-to-read collection of web pages into a clear, long-term record of a nation's energy journey. It shows that even when data is hidden behind messy interfaces, it can be uncovered and organized to reveal the steady, complex pulse of a modern economy. The result is a tool that allows anyone, from students to policy experts, to see the past decade of Bangladesh's power history in high definition, ready to inform the decisions of the next decade.