NIT Rourkela Team Patents AI System to Cut Water Use in Solar Panel Cleaning

Researchers at the National Institute of Technology (NIT) Rourkela have secured an Indian patent for an artificial intelligence system designed to make solar panel cleaning more targeted and resource-efficient. The invention, titled "Federated Learning-based Autonomous System and Method for Monitoring and Cleaning Solar Plant," was developed by Prof. Arun Kumar, Prof. Bibhudatta Sahoo, Dr Lopamudra Hota and Dr Biraja Prasad Nayak from the Department of Computer Science and Engineering.

Dust, bird droppings and industrial pollutants can cut energy generation by up to 40% at solar farms in dry and dusty regions, according to the research team. Conventional maintenance often relies on fixed cleaning schedules that wash every panel regardless of actual soiling, consuming large volumes of water and labour.

The new system uses federated learning, a machine learning approach where local devices analyse data on-site and share only encrypted model updates rather than raw operational data. This architecture aims to preserve data privacy, reduce bandwidth needs and improve cybersecurity while allowing the platform to detect faults, assess real-time soiling conditions and recommend cleaning only for affected panels.

The researchers say the integrated platform combines federated learning, edge computing, AI-driven predictive maintenance and autonomous cleaning controls. In simulations, the technology has reached Technology Readiness Level 3, meaning the concept has been validated as a proof of concept under controlled experimental conditions.

Prof. Sahoo stated that existing market solutions often involve high capital costs and limited intelligence, whereas the NIT Rourkela system integrates multiple AI functions for autonomous operation. The team estimates the technology could eventually deliver its features at roughly 10% of the cost of current systems once scaled for field deployment.

Potential applications include utility-scale solar plants, floating solar farms, rooftop installations, industrial parks, smart city infrastructure, defence sites and remote off-grid systems. The researchers note the approach could support India's National Solar Mission and net-zero targets by reducing the resource intensity of panel maintenance.

Next steps involve building a physical hardware prototype and integrating Internet of Things sensors to collect live data from solar installations. The team plans to move from simulation to pilot projects in real-world conditions and is seeking collaboration with government agencies and industry partners for field testing and technology transfer.

Future development may incorporate drone-based inspection, coordinated cleaning robots and AI models for energy yield prediction. The researchers emphasise that successful real-world trials will be necessary before the system can be deployed at scale.