The renewable energy sector is in full development, and the use of Artificial Intelligence is improving and optimizing all its processes. The use of Artificial Intelligence is transforming the commissioning and maintenance of solar power plants worldwide.
In2AI, one of the leading Artificial Intelligence companies, and Exanter, a company specializing in improving the maintenance and efficiency of solar parks, have jointly developed an Artificial Intelligence-based solution that is transforming the commissioning and maintenance of solar power plants in Spain and Latin America.
The innovative software automates the analysis of images captured by drones and provides precise and detailed information about the status of each solar panel through a platform accessible to solar park managers. This solution, which combines high impact with cost-effectiveness, allows for improved profitability of photovoltaic installations.
The use of drones for image capture is a faster, more economical, and more accurate alternative to traditional options, such as taking photos from manned aircraft or manual on-site analysis of solar panels. According to Exanter, “their fleet of drones fly over the solar panels and are equipped with state-of-the-art infrared and daylight cameras.”
Meanwhile, In2AI has focused on automatic analysis of these images, avoiding errors in manual or semi-manual processes, and achieving high data processing speed to detect incidents as soon as possible.
Lino González, project leader and AI expert, points out that, “while the starting point of the project is obtaining good images, the most innovative aspect is the advanced analysis of these images using computer vision, machine learning, and an open-source environment, without proprietary technologies, for automatic detection of shadows and other anomalies.”
Computer vision is a field of artificial intelligence that allows for extracting meaningful information from digital images, videos, and other visual inputs. Thanks to this, measures can be taken or recommendations can be made based on that information. For this project, computer vision technologies were used to identify underperforming solar panels due to connection issues, such as shadows, stains, or manufacturing defects, thus maximizing plant efficiency.
The foundation of the solution is machine learning, which has made it possible for the solution to be ready to quickly and accurately detect any type of incident after a few months of learning and inventorying incidents. The software continues to learn and increase the accuracy of the analysis as it is used.
Additionally, the software analyzes new images and presents the results through a platform accessible to the client, allowing for filtering of error types according to their needs.
The project encompasses the development, training, and deployment of computer vision models specialized in shadow and defect detection. Using the Icevision framework and pre-trained models from Open MMLab’s MMDetection, the solution is deployed on the AWS platform.
The first clients who have implemented the solution highlight its decisive impact on reducing costs associated with maintenance and periodic inspections necessary for the optimal operation of photovoltaic solar parks. They also report a 25% improvement in the efficiency and performance of their solar installations.
Referrer: MiMub in Spanish
