Artificial Intelligence offers a huge potential for the solar sector, especially in Operation and Maintenance.
Some companies are already working with it, and new applications are expected to be launched soon, with growing business opportunities.
We talked about the deployment of AI in solar energy with Mattia Dallapiccola, Research Assistant at EURAC Research Institute.
There is no field where AI cannot be used. In solar energy, we can envisage applications ranging from cell design, production, logistics and transportation to operational system maintenance. Most potential applications can be found in operational maintenance because it relies on a lot of complex data. It costs a lot. You need a dashboard. You need a lot of algorithms to detect problems, to label tickets.
Of course, AI cannot do the complete job, but it can really simplify it by clustering and summarizing tickets. This kind of application will soon be available, and it will provide new models for the solar sector.
There are already applications which use AI from the first phases of the cell design process where dozens of parameters must be optimized at the same time. In this case AI can help by providing suggestions and forecast based on previous experiments or even its own knowledge with the final goal of reducing material usage and cost of production while maximizing efficiency.
Moreover, in cell manufacturing it is possible to utilize computer vision applications that monitor the production line with a camera and analyze images to detect defects. These applications then create reports or tickets, which boost both efficiency and quality.
What are the opportunities for AI in solar energy? You just mentioned improvements in quality and efficiency.
AI can be a revolutionary tool since it can give access to everyone in the value chain to knowledge and expertise in the sector, which is necessary to reach our solar deployment goals. Vertical - or industry-specific - AI can create tools for experts that make them more efficient while maintaining high quality. Moreover to reach the solar deployment goals the sector needs a lot of experts that now are not available. AI can help close this gap by providing access to expertise and knowledge for the persons that will enter the solar sector soon.
What are the risks and limitations of artificial intelligence in solar energy?
The major limitation is data availability. Obtaining data from third parties is difficult, and the number of plants that provide reliable data is limited. Another limitation is the lack of standards. Every system and every operator have their own system for storing data. To train an AI model, you need uniform, high quality and reach data.
So, a lot of training data are not available or were collected in the wrong way, which make them unreliable. This has a negative effect on the quality of the resulting general models and in some cases is also risky.
Another risk is, of course, the way we use data. Data owners need reliable ways to share data and to train models. Not every company can afford to train their own model, only big companies can. But big companies may not be willing to share their models with competitors. So, we really have to base the development of AI in the solar sector on trust and the willingness to share data with the common goal of improving quality and efficiency in the whole sector.
At the moment, AI is just at the beginning in the solar sector, so we don’t have a comprehensive answer. But let’s use as an example O&M. Staring at dashboards for hours carries the risk of a maintenance operator missing issues or sometimes not looking at the data at all for days. Thanks to the improved efficiency and simplicity that AI brings, one person can take care of a large number of plants helping the operator to detect, cluster and find a solution quickly when an issue occurs. That is why operational maintenance is the first application where AI can be very useful and probably profitable.
But AI can be applied in almost every step of the value chain. Let's say the solution is there, now we have to find the right problem to apply it. There will be a lot of companies investing in AI for the solar sector. In three years' time, we will know in which parts of the solar sector AI can be profitable.
How can solar installers benefit from AI? And what about operators of solar installations on residential buildings?
We have created an AI chatbot for operators of solar installations on residential buildings who have no experience in solar management, but also for installers. The person interacts with it via text or voice message without having to understand a complex dashboard or installing additional software. They just ask their solar assistant, “How is my system doing?”. We get the data, we analyze them, and then we report back. In the same way a solar installer can interact with the system and then provide support to their customer.
Of course, many systems have a monitoring app or platform, but some installers and system owners do not really look at the data. This means that solar installers do not provide the best possible maintenance support to their clients and sometimes it takes weeks or months to detect an issue. In this way AI can also improve the efficiency of installers’ businesses.
For example, you may have a portfolio of 100 systems. AI may then tell you that you have ten systems that have been running for ten years, one of which is not performing well compared to the others. It can suggest possible issues and solutions. Or maybe AI could tell you that the some components are not working well and that it might make sense to buy some of them to offer the customer replacements quickly after the issue arise or even before.
These are just examples of applications of AI in the solar sector for improving the quality of support services while also boosting efficiency.