Photovoltaic generation prediction models
PV generation prediction models for small power installations such as self-consumption systems and district-level assets in Local Energy Communities or Positive Energy Districts
Description of the service
Demand forecasting provides substantial value for planning electricity purchase positions in the daily and intraday markets.
At CIRCE-Technological Center we develop photovoltaic generation forecasting models for small power installations such as self-consumption systems and district-level assets in Local Energy Communities (LECs) or Positive Energy Districts (PEDs).
Our approach uses a "grey box" model, which combines weather forecasts, production history and physical characteristics of the facility to provide an accurate and reliable forecast.
Share your challenge with us!
Value proposition
Predictive optimization and generation efficiency
This algorithm provides essential information input for optimal predictive management, enabling more efficient utilization of renewable generation sources, in particular photovoltaics. This leads to a significant improvement in profitability and an increase in self-sufficiency and self-consumption ratios. Likewise, it is also useful for forecasting possible surpluses of large self-consumption that could access the daily and intraday electricity market.
Customizable forecasts and real-time learning
We offer two-level forecasts: quarter-hourly values for the next hour and hourly values for the next 24 hours. The time window and time steps can be adjusted as needed, relying on the quality and accuracy of the weather forecasts. In addition, the algorithm implements an online learning layer that improves its accuracy and adapts to changes in the delivery capacity of the facility.
Adaptability to new installations
The model is able to adapt and predict the generation capacity of new installations, even in the absence of historical data, based on the physical model of the installation.
Our Clients' Feedback
Frequently Asked Questions (FAQs)
Predictions are provided at two different levels: four quarter-hourly values for the next hour and 24 hourly values for the next 24 hours. This provides a detailed, short-term view of the expected generation.
Yes, both the time window of the forecasts and the time steps can be adjusted. This depends on the quality and accuracy of the available weather forecast, which allows customization according to the specific needs of each installation.
Our algorithm implements an online learning layer that allows it to adjust and improve its accuracy over time. This means that as the system operates, the algorithm adapts and identifies changes in the delivery capacity of the facility, ensuring increasingly accurate predictions.
Yes, the algorithm is designed to predict the generation capacity of new facilities, even without historical data. It uses the physical model of the facility to make its predictions, making it ideal for facilities that are newly constructed or in early stages of operation.
