The Smart4RES vision corresponds to a holistic approach, where developed tools answer to key challenges faced by the intelligence layer in modern energy systems. These tools interact within a generic model and value chain from data to decision and generate structuring responses for the power system.

While our public deliverables are being reviewed by the EC, we offer below a summary of Smart4RES’ main achievements.

The first challenge is the imperfect knowledge of the physical system. Weather forecasting is an essential input for RES forecasting but is not always tailored to energy applications. Smart4RES proposes forecasts based on high-resolution and multiple data sources. These solutions are adapted to the energy sector and enable to obtain a higher modelling accuracy such as 15% of improvement for the Pseudo-Deterministic forecast from Ensembles.
In addition, the use of sky-imager data allows to improve PV forecasting accuracy at minute-ahead horizons of 20% while the seamless approach for RES forecasting over multiple time scales, improves forecasting accuracy of 16% over 15-min to 6-h ahead horizons.

The second challenge addressed by Smart4RES is the vulnerability of existing solutions to the real-world quality of information, e.g. the lack of information available to forecasting solutions. Smart4RES solutions optimise decisions considering lacking information, therefore improving the resolution of RES-dedicated solutions. Within WP4, it has been demonstrated that:

  • The use of additional and distributed data (e.g. from neighbouring RES plants) may yield RES forecast quality improvements of up to 10-15%
  • Privacy-preserving distributed learning and data markets are key to realize the value of distributed data
  • Data markets bring new revenue streams for renewable energy producers and other actors of the energy systems

Considering the multiple necessary modelling steps and relevant data sources for prediction and optimization in RES applications, existing solutions highly depend on specific data sources and have low interpretability. Seamless forecasting, prescriptive analytics and reduced information for human operators enable to achieve simplicity of RES-dedicated solutions.

Uncertainties related to RES processes and applications (markets and grid management) are multiple and evolve dynamically over time. Better predictions reduce uncertainties and optimisation tools accommodate the remaining level of uncertainties.

Finally, prediction and decision-making face several barriers leading to suboptimal decisions: predictions are tuned to maximize accuracy and not decision value, and privacy constraints prevent models from harvesting the full potential of relevant data sources. Privacy-preserving approaches, value-oriented forecasting and prescriptive analytics form a bundle of solutions that enable to maximise value in RES trading and power system management.

Deliverables will be made available soon. Consult our Resources Center!