Data-driven Algorithmic Performance Monitoring of Indian Solar Energy Systems

Solar Photovoltaics (PV) are on the rise as the better and more reliable way to boost the renewable energy industry. Among its various types, rooftop solar is arguably the one with maximum untapped potential and rapid scalability. In this scenario of increasing demand for solar energy systems, the need to predict their feasibility and monitor performance is more than ever.

Although these systems are constantly developed to achieve greater reliability & lower maintenance, they are highly susceptible to their surrounding adverse conditions. Combined with their material properties, we encounter several lossy phenomena resulting in reduced performance.

In this article, I bring your attention to understand some of these losses. All these are perfect use-cases to show how data-driven algorithms can play a crucial role in the future of solar research & business.

Solar rooftop is an energy system very much open to its surroundings. This characteristic makes its generation intermittent and highly contingent on having suitable weather. Incidentally, there are also intrinsic limitations governed by their material physics. Since any decrease in solar PV generation can result in considerable financial & operational issues, it is important to predict the possible generation.

Eergy estimation is only possible when we can effectively model the system taking into account its various parameters as well as losses. The most important parameter influencing generation is the incident irradiation. Apart from that, Photovoltaic losses play a crucial role in benchmarking the maximum generation possible.

These losses are governed by a plethora of natural and operating conditions such as cloud cover, module operating temperature, ambient temperature, particulate deposition rate on the modules and sensors at the site, grid unavailability (in case of hybrid solar systems) and many more.

A major part of the total loss in solar rooftop generation is due to its module operating temperature. Incident irradiation rises the module temperature thereby decreasing its operating efficiency. This initiates a negative feedback wherein reduced efficiency further increases its temperature resulting in a significant thermal loss. Soiling of modules due to particulate deposition is another important phenomenon contributing to considerable loss.

Loss due to soiling, in contrast to thermal loss, is almost exclusively defined by the plant surroundings and its module cleaning frequency. Loss due to degradation and ageing is a more elusive one with its roots in the solar module’s electrical characteristics. It is intrinsic in Photovoltaics nature to give depreciated performance as time passes even in the most pristine operating conditions. This unavoidable loss is another long-term risk both in terms of potential revenue loss and reduced plant life.



Given the significant trouble these losses can pose towards optimal plant performance, it is imperative to have a robust understanding of them from a design as well as asset monitoring standpoint. Therefore, modelling these PV losses is no longer just a theoretical interest in energy science. It should rather be a norm in all solar industrial operations giving insights on how to have better operational efficiencies.

This is precisely the motivation behind the pursuit of employing data-driven studies on solar projects to recognize patterns indicative of these losses. It is noteworthy at this juncture to point out the importance of having an extensive solar data acquisition system. Modern technologies like the Internet of Things (IoT) are indispensable in leveraging this immense power of data. Such data is invaluable to answer and corroborate almost all kinds of decision-making exercises. Yet, this kind of data-driven approach remains relatively new to many Indian solar rooftop operations.

One main point of this article is also to motivate individuals & organizations in the solar energy segment towards the greater untapped potential of data-driven analytical decision making. This serves in both respects, as a consistent and tangible system to validate pre-existing domain notions & as a tool of innovation and pushing the envelope of solar research numerically.

There is quite a bit of research and effort in devising such data-centric empirical (as well as machine intelligent) algorithms. These can be replicated and deployed as per the requirement to estimate aforementioned PV losses on a real-time basis. These when implemented will give invaluable feedback aimed at optimizing performance of solar system

- Vijay Bhaskar, Senior Associate, Analytics,  Amplus Solar