In a major technical breakthrough for India’s renewable energy infrastructure, researchers at the National Institute of Technology, Rourkela, have developed an artificial intelligence-driven autonomous system designed to monitor and clean solar panels selectively. The innovation, which has been granted an Indian patent, utilizes advanced machine learning techniques to identify soiling and technical faults in real time, triggering cleaning operations only when necessary rather than relying on rigid, calendar-based washing schedules.
The development comes as India rapidly expands its solar power capacity, deploying vast utility-scale photovoltaic installations across arid and dusty terrains. By replacing labour-intensive and water-heavy maintenance routines with targeted, automated intervention, the system promises to significantly reduce operational expenditure while conserving billions of litres of water across large-scale clean energy projects.
Overcoming the Soiling Tax in Renewable Generation
Solar power generation across the Indian subcontinent faces severe operational bottlenecks caused by atmospheric dust, industrial particulates, and avian droppings. In dry regions such as Rajasthan and Gujarat, surface accumulation on photovoltaic modules can diminish power output efficiency by up to 40 percent, imposing a continuous financial drain on solar park developers. Traditional mitigation strategies depend on manual labour or automated washing robots that clean entire solar arrays uniformly on fixed schedules, incurring steep labour costs and consuming immense quantities of freshwater in drought-prone landscapes.
To solve this persistent operational inefficiency, a research team from NIT Rourkela’s Department of Computer Science and Engineering—comprising Assistant Professor Arun Kumar, Professor Bibhudatta Sahoo, and research graduates Dr Lopamudra Hota and Dr Biraja Prasad Nayak—engineered a condition-based maintenance framework. Their patented methodology, titled “Federated Learning based Autonomous System and Method for Monitoring and Cleaning Solar Plant,” shifts the maintenance paradigm from periodic schedule-driven washing to selective, need-based action.
Privacy-Preserving Architecture and Edge Computing
At the core of the innovation is Federated Learning, a decentralized artificial intelligence architecture that enables real-time diagnostic analysis without compromising data security. Conventional monitoring platforms routinely collect and transmit raw operational telemetry from solar fields to centralized cloud servers, exposing critical utility infrastructure to potential cyber threats and incurring significant network bandwidth costs. In contrast, the NIT Rourkela platform processes operational telemetry locally on edge computing nodes, sharing only encrypted mathematical updates across the network to preserve cyber resilience and operational privacy.
The system continuously evaluates output metrics across individual photovoltaic cells, pinpointing specific modules affected by heavy dust accumulation, physical defects, or localized shading. Once a dirty or faulty panel is identified, the platform generates precise cleaning recommendations and activates localized automated mechanisms to clean only the affected surface area. This granular approach eliminates unnecessary washing cycles across clean panels, offering a far more sustainable operational blueprint for utility-scale solar farms, floating photovoltaic arrays, and urban rooftop systems.
Disruptive Cost Dynamics and Scalability
Beyond its water-saving and security benefits, the system presents compelling economic advantages for commercial solar developers. The research team projects that once fully commercialised, the integrated sandbox platform will deliver advanced predictive maintenance and autonomous cleaning capabilities at approximately 10 percent of the capital cost associated with existing market alternatives, which frequently suffer from limited intelligence and prohibitive upfront hardware expenses.
Currently validated through controlled experimental simulations at Technology Readiness Level 3, the project is advancing toward physical hardware prototyping. Future development phases will incorporate Internet of Things sensor arrays, drone-assisted aerial inspections, and multi-agent collaborative cleaning devices to prepare the technology for field trials across diverse geographic environments.
Strategic Alignment with Climate Ambitions
As the Union Government accelerates initiatives under the National Solar Mission to achieve its target of 500 gigawatts of non-fossil fuel energy capacity by 2030, maintaining peak generation efficiency across existing installations has become an economic imperative. By minimizing water dependency and optimizing energy yields across the nation’s solar installations, innovations originating from Indian academic institutions are poised to play a crucial role in supporting the State’s long-term net-zero emission commitments.
