Affordable RES production forecasts with PROGO

By September 3, 2019 Artificial Intelligence, Energy, R & D

Our R&D team is actively working on innovative IT projects, cooperating with the most prestigious technical universities and research facilities. The solutions for different industries are implemented in many companies across the world. One of our projects for the energy sector is PROGO – a system which will produce affordable Renewable Energy Sources (RES) energy production predictions. Recently we have finished the second phase of this project.

What’s PROGO?

PROGO is a solution for accurate area-based prediction of RES production. Opposed to point-based predictions this method allows significant cost reduction, allowing to include many small and medium-sized prosumers with the prediction. Currently, due to lack of profitability – it was impossible and their energy potential remained unused.

What difference will it make?

Implementing PROGO will instantly lower the price of the RES energy and increase its supply at the same time. Prediction done by this system will include new sources being created and the current network load. The ability to forecast consumption and production at different time horizons and different frequency depending on the needs will help with network stability and safety.

Another milestone

Recently we’ve completed the second stage of this project. The phase was finished with validation and testing of PROGO’s prediction data. We were cooperating with Enea, a Polish DSO, from where we received data necessary for checking the quality of prediction in real conditions of customers’ usage.

Our system was tested on a north-western part of Poland – a perfect testing grounds. The region experiences great tourist surge in the summer, which means huge energy demands. Also, this area consists of various RES including wind turbines and a couple of small water plants with about 100 prosumers. It allowed us to test our prediction in a demanding real-life environment.


The tests began in April of 2016 and lasted for 33 months. After that, the gathered data was transferred and analyzed. We used it to create and upgrade new deep learning models, as well as to confront the predicted values with the actual demands.

The acceptance threshold that was stated in our project preparations was 20%.


Square Root Mean Error (SRME):

8,5% - 10,1
For wind turbines
For small power plants
For the entire testing area

After set area corrections, we managed to lower SRME to 6.1% Those prediction accuracy results were better than the acceptance threshold – meaning we have succeeded in the second phase of the PROGO Project.

If you want to know more about our recent developments in our R&D projects contact us! We’d be happy to answer your questions.