How IIIT-B’s algorithms are teaching the grid to think green

Researchers at the International Institute of Information Technology Bangalore (IIIT-B) are using machine learning and mathematics to tackle one of renewable energy’s toughest problems - how to generate enough clean power without driving up costs or risking grid instability.

By developing optimisation models that balance carbon reduction with affordability, their work aims to make India’s transition to solar and wind energy both reliable and practical. Their models not only forecast solar or wind power generation, but they also balance multiple objectives at once such as accuracy, cost, and reliability, helping grid operators make fairer, more transparent decisions in real time. 

Datasets from Germany

Aswin Kannan, assistant professor, IIIT-B, who led the research, along with his students, have worked on datasets from Germany (Netztransparenz, SMARD), the United States of America (NREL), and India, linking weather variables such as irradiance, temperature, and pressure to real power-output data. 

From multiple research papers, the team found that accuracy alone is not enough.

“In energy markets, over-predicting reduces reliability, while under-predicting increases operational costs. We also found that bias in data can quietly distort results. By combining optimisation with learning, we can detect these biases and build forecasts that balance cost, reliability, and fairness for real-time grid operations,” Prof. Kannan explained.

While much of his early work was in Europe, Prof. Kannan says India presents a far more dynamic challenge. “India’s renewable data quality is actually very good, sometimes better than Europe, but its variability is much higher,” he said, pointing out that unlike Germany’s uniform weather, India’s solar and wind conditions vary drastically across States and seasons.  

He also noted that in India, publicly managed transmission systems are better suited to handle such vast and diverse networks compared to Europe’s privatised model.

A transition of scale

Higher solar radiation doesn’t automatically mean higher output here. Humidity, dust, and terrain play a much bigger role. In fact, India already generates a larger share of power from renewables than many realise, he added.

According to Prof. Kannan, India’s energy transition is not difficult because of policy or unpredictable supply, but because of scale. “In Europe, the transition meant retrofitting existing pipelines for hydrogen. In India, the challenge is creating new microgrids, battery systems, and transmission lines for variable renewable power,” he said.

Prof. Kannan’s ongoing research now focuses on solar, wind, and hydro systems, and how they can work together within a joint hydrogen–electricity network. While industry tools typically aim only for accuracy, in this framework, the models weigh trade-offs between cost, bias, and risk of error. They also switch algorithms based on data quality or changing weather, an approach that makes them more resilient to sudden shifts or uncertainty. 

The research has clear implications for grid operators, policymakers, and renewable developers. Better forecasts, as per the team, can prevent costly imbalances in power markets, reduce wastage, and allow for more flexible energy pricing.