Digital Infrastructure
AI for Visibility and Forecasting of Renewable Generation
The AI for Visibility and Forecasting of Renewable Generation project aims to enhance UK Power Networks' ability to forecast unmetered solar energy generation. By developing a machine learning (ML) algorithm, the project will infer the capacity of unmetered solar generation behind substations. This capacity estimate will feed into a solar forecast algorithm, producing forecasts at primary substations.
Improved visibility and forecasting will help UK Power Networks better anticipate network power flows, reduce procurement costs, minimize renewable curtailment, and inform long-term planning. This project aims to lower costs for end users and support a sustainable energy system.
Key deliverables
- Project is in it's infancy, so no outputs have yet been produced. However the objective of the project is for the DSO to have improved forecasts across intra-day and day ahead horizons compared to the current model in use, to improve day ahead flexibility decisions and real time network control. This will be delivered through: • Improved accuracy of generation forecasts for metered generation • Improved understanding of capacity of unmonitored/behind the meter generation
Target outcomes
- As the project is just beginning, there are no outcomes or learnings to report
Measures of success
- The project aims to develop a forecasting model that provides a meaningful improvement in forecast accuracy for metered sites and provides an approach to unmonitored renewable capacity estimation. The data will be made available through an API which can be used with our existing forecasting system. The success criteria are: -Forecasting accuracy improvement Meaningful improvement in accuracy compared to forecasts using Global Forecast System (GFS) weather forecasts, demonstrating the effectiveness of the newly developed forecast algorithms. Probabilistic forecast accuracy will be assessed by measuring quantile exceedances and pinball loss. 10th and 90th percentiles will be assessed. -Unmonitored renewable capacity estimation UK Power Networks to improve its energy accounting of behind the meter wind and solar generation capacity at the DNO level. Report (project milestone 3.
- will include an estimate and the uncertainty on the estimate. -Evidence of potential efficiencies Initial assessment has suggested that this implementation has the capacity to reduce flexibility purchasing costs by ~1.5% after the system is operational leading to cost reductions as detailed in Section 3.2. The level of improvements seen in forecasting accuracy and capacity estimation, over the course of the project, should provide an indication that this forecasted cost reduction is attainable over the longer term
Stakeholder Engagement
Stakeholder engagement plan is in development
Timeline
Status
On TrackMilestones
- Discovery Phase Kick-Off (Apr 25)
- M1 - Completion of WP1 (Jul 25)
- M3 - Completion of WP3 (Mar 26)
- M2 Completion of WP2 (Nov 25)