NIT Rourkela develops AI system to make solar plants smarter, efficient

Patented technology uses AI, edge computing and federated learning to optimise solar panel cleaning, reduce water use and improve power generation efficiency

As India races towards achieving 500 GW of non-fossil fuel power capacity by 2030 and expanding solar energy as the backbone of its clean energy transition, researchers at the National Institute of Technology (NIT), Rourkela, have developed an artificial intelligence (AI)-powered autonomous system that could significantly improve the efficiency and economics of solar power plants.

 

The patented technology addresses the challenge of maintaining clean solar plants to ensure peak power generation without regular human intervention. It intelligently identifies only the portions of solar panels that require cleaning, monitors their health in real time and performs selective cleaning, reducing water consumption, lowering maintenance costs and maximising electricity generation.

 

As solar energy capacity in India expands rapidly, large-scale solar farms, particularly in arid and dusty regions, face energy losses of up to 40 per cent caused by dust, bird droppings, industrial pollutants and other debris accumulating on solar panels.

 

Conventional maintenance practices generally rely on periodic cleaning schedules irrespective of the actual condition of the panels. Besides being labour-intensive, these methods consume enormous quantities of water and increase operational costs while often resulting in unnecessary cleaning of already efficient panels.

 

The researchers have achieved a significant breakthrough in addressing this major yet often overlooked challenge in the solar energy sector. "Our system replaces this conventional approach with an intelligent decision-making framework. Instead of cleaning every panel at fixed intervals, the AI continuously analyses the operational condition of each panel, detects faults or contamination, and determines whether cleaning is actually required," said Arun Kumar, assistant professor, Department of Computer Science and Engineering.

 

More importantly, he said, the AI system performs selective, need-based cleaning by targeting only the affected areas, thereby conserving water and reducing wear and tear on equipment. This condition-based maintenance approach enables operators to optimise both plant performance and maintenance expenditure.

 

According to the researchers, the system integrates multiple advanced technologies, including federated learning, edge computing, artificial intelligence, autonomous cleaning and predictive maintenance, into a single platform capable of real-time monitoring and intelligent decision-making.

 

What makes the technology noteworthy is its use of federated learning, an advanced branch of artificial intelligence regarded as the future of privacy-preserving AI. Unlike conventional AI systems that transmit raw operational data from every solar installation to a central server for processing, federated learning allows AI models to be trained locally at the edge while sharing only encrypted model updates rather than sensitive operational data.

 

"The combination of autonomous fault detection, selective cleaning, lower water consumption and reduced maintenance costs makes the innovation a first-of-its-kind integrated solution for solar infrastructure. The technology reduces cybersecurity risks, protects operational privacy, lowers communication bandwidth requirements and makes the system more scalable for deployment across geographically distributed solar farms," said Bibhudutta Sahu, professor, NIT Rourkela.

 

While existing automated solar cleaning systems available in the market involve high capital investment and relatively limited intelligence, the AI-driven system can deliver superior functionality at a fraction of the cost. "Once commercialised and deployed at scale, the technology is expected to offer enhanced performance at nearly 10 per cent of the cost of current market solutions, making intelligent solar maintenance economically viable even for medium-sized installations," Sahu said.

 

The technology has wide-ranging applications extending beyond conventional utility-scale solar parks. It can be deployed in floating solar farms, rooftop photovoltaic systems, industrial solar parks, smart city energy infrastructure, defence establishments and remote off-grid renewable energy installations where routine maintenance is either expensive or logistically challenging.

 

The researchers are now working towards developing a full-scale hardware prototype integrated with Internet of Things (IoT) sensors and validating the technology through pilot deployments. The team also plans to incorporate drone-assisted inspection, multi-agent collaborative cleaning and AI-driven energy yield forecasting in the next stage of this innovation. 

The AI-powered system detects dust, activates automatically and begins cleaning when required | Photo: NIT Rourkela

 

This is how the system works | Photo: NIT Rourkela