Use of AI

AI – Artificial Intelligence in KYOS’ models

Recently our market analyst team at KYOS took a deep dive into the world of Artificial Intelligence (AI), focusing on FCR forecasting using the latest machine learning libraries.

Through feature engineering, we’re gaining insights into the key drivers of this dynamic market. FCR is notoriously challenging due to its ever-changing nature and the absence of recurring historical trends. A prime example is the significant price surge since March 2024, fueled by renewable oversupply (look at the first figure).

Example of using AI in FCR forecast

But we’re up to the challenge! By refining our training sets, we’ve achieved accurate short-term forecasts. Our approach combines forecasting Day-ahead, Intraday, and Imbalance prices, as well as renewable volume with our KySim model taking the historical developments of Gas, EUA and Dutch power markets into account. In a later stage we integrate these simulations as features in our ML model.

Check out the second figure which showcases our FCR forecast for 2 days ahead, trained up until one day before delivery (prices are predicted prior to market settlements).

And this is just the beginning. We’re committed to improve our feature engineering and exploring new avenues to deliver accurate long-term forecasts. Stay tuned for more exciting developments from KYOS as we explore the power of AI to transform market analysis!