
Monitoring photovoltaic plants with Machine Learning
Perpetum
Challenge
Perpetum, a leading entity in the energy sector with a focus on photovoltaic
panel (PV) installation and monitoring in Belgium, faced a significant challenge in monitoring
and managing their vast array of PV power plants.
Their primary goal was to reduce the downtime of PV plants and perform
early identification of anomalies. The objective aimed to enhance the
capabilities of the maintenance team, optimize energy output, and guarantee
the satisfaction of their clients.
Their secondary aim was to forecast the production of photovoltaic
energy up to a week in advance. This was intended to mitigate
imbalances in the electric grid and improve their financial performance.
Outcome
Our team implemented a solution which involved
several key steps and achievements.
We first created a scalable big data infrastructure to
handle the large amount of data from PV power plants. The
infrastructure forms the backbone for the monitoring web platform,
facilitating seamless data collection, processing, and analysis.
We then incorporated intelligent anomaly detection mechanisms
into the web platform, using advanced machine learning (ML) algorithms.
The integration allows Perpetum’s maintenance team to receive real-time
alerts for detected anomalies in electricity production. Promptly addressing
these issues minimizes downtime, optimizes performance, and improves operational
efficiency; ultimately leading to reduced financial losses
Moreover, the platform we developed offers Perpetum's customers the
ability to view their production up to a week in advance. Our team developed
this forecast functionality using machine learning-based algorithms, taking into
account weather forecasts and historical production data to ensure accurate predictions.
By leveraging these predictions, Perpetum's customers can optimize the reselling
of their exceeding production, resulting in increased financial returns.
We also built a custom monitoring and reporting platform that
retrieves data from multiple third-party systems. With its intuitive interface,
this platform is accessible to both technical and non-technical users within
Perpetum, allowing the team to concentrate on ensuring the smooth operation
and optimal performance of their PV plants.
Currently, the application we developed and deployed is in active
use in production, effectively collecting and aggregating data while forecasting
the activity of hundreds of megawatts of solar power in Belgium and abroad.
Conclusion
Perpetum has successfully transformed their PV power plant
monitoring capabilities. Their maintenance team now benefits from immediate
alerts about anomalies, enabling prompt issue resolution and minimizing
downtime. The enhanced efficiency has significantly contributed to Perpetum's growth.
Accurate energy production forecasting has also positively impacted
their financial performance. With real-time reporting and predictive capabilities,
Perpetum's clients can make timely actions, ensuring their satisfaction and trust.
Our collaboration with Perpetum showcases the transformative power
of advanced data-driven solutions.