General Description
The energy market is undergoing a profound transformation driven by multiple structural and regulatory factors. Key elements include:
-
Changes in historical price patterns: Price dynamics no longer follow traditional historical trends. Recent market modifications, combined with the broader contextual environment, have resulted in significant shifts, making correlation with past trends increasingly difficult.
Transition to quarter-hourly markets: The shift to 15-minute intervals increases the demand for timely data updates for system operation and requires higher precision in the models that support decision-making.
These factors call for new predictive models and approaches capable of adapting to the current market volatility and complexity.
With over 10 years of experience in renewable generation and demand forecasting, Gnarum in 2025 took a step further to enhance its value proposition by offering a specialized service for price forecasting and fundamental market variables in European energy markets.
This has been made possible by combining extensive expertise and know-how with cutting-edge technology in time series prediction.
Our client profile
Day-ahead price forecasting is a critical variable for Market Agents, enabling them to optimize buying and selling strategies for each 15-minute interval while integrating generation and demand forecasts.
The service is primarily targeted at Traders managing large renewable portfolios, particularly those involving battery optimization. In both cases, users require algorithms that explicitly incorporate risk management, providing not just point estimates but the full probabilistic distribution.
The focus is on short-term forecasting, prioritizing maximum accuracy for the 96 quarter-hourly periods of the following day.
Our challenge
The market lacks truly differentiated predictive solutions. Traditional providers operate with established models that, while functional in past contexts, increasingly show limitations in the face of:
This gap between existing offerings and the current needs of market agents—particularly regarding probabilistic risk management—creates a clear opportunity for innovative, methodologically rigorous solutions.
Gnarum’s Solution
Gnarum’s approach is built on four pillars:
-
AI/ML Integration: Combines foundational models with classical statistical and machine learning techniques, creating a hybrid and robust ecosystem.
-
Updated Scientific Basis: Incorporates the latest advances from specialized literature in time series forecasting, ensuring alignment with state-of-the-art methods.
-
Advanced Architecture: Designs structures that enable collaboration between multiple models, optimizing calibration and training to maximize precision and stability.
-
Dynamic Adaptability: Self-calibrating systems designed to respond effectively to changing market conditions.
Gnarum’s Solution
Gnarum’s approach is built on four pillars:
-
AI/ML Integration: Combines foundational models with classical statistical and machine learning techniques, creating a hybrid and robust ecosystem.
-
Updated Scientific Basis: Incorporates the latest advances from specialized literature in time series forecasting, ensuring alignment with state-of-the-art methods.
-
Advanced Architecture: Designs structures that enable collaboration between multiple models, optimizing calibration and training to maximize precision and stability.
-
Dynamic Adaptability: Self-calibrating systems designed to respond effectively to changing market conditions.
Two distinctive features of the solution are:
-
Delivery of the full probabilistic distribution through 19 percentiles (P5 to P95) for each of the 96 quarter-hourly periods.
-
Calibration of the distribution using Conformal Prediction techniques, without assuming any predefined distributional shape.
This approach is critical for trading operations, where risk management is key to decision-making, positioning Gnarum’s solution as a novel offering in today’s market.