Projects
Projects: Projects for Investigator |
||
| Reference Number | NIA2_NGESO002 | |
| Title | Solar PV Nowcasting | |
| Status | Completed | |
| Energy Categories | Other Cross-Cutting Technologies or Research(Energy Models) 20%; Renewable Energy Sources(Solar Energy) 50%; Other Power and Storage Technologies(Electric power conversion) 15%; Other Power and Storage Technologies(Electricity transmission and distribution) 15%; |
|
| Research Types | Applied Research and Development 100% | |
| Science and Technology Fields | PHYSICAL SCIENCES AND MATHEMATICS (Computer Science and Informatics) 30%; ENGINEERING AND TECHNOLOGY (Electrical and Electronic Engineering) 70%; |
|
| UKERC Cross Cutting Characterisation | Systems Analysis related to energy R&D (Energy modelling) 100% | |
| Principal Investigator |
Project Contact No email address given National Grid ESO |
|
| Award Type | Network Innovation Allowance | |
| Funding Source | Ofgem | |
| Start Date | 01 September 2021 | |
| End Date | 31 July 2023 | |
| Duration | ENA months | |
| Total Grant Value | £500,000 | |
| Industrial Sectors | Power | |
| Region | London | |
| Programme | Network Innovation Allowance | |
| Investigators | Principal Investigator | Project Contact , National Grid ESO (100.000%) |
| Industrial Collaborator | Project Contact , National Grid plc (0.000%) |
|
| Web Site | https://smarter.energynetworks.org/projects/NIA2_NGESO002 |
|
| Objectives | The project proposes to create more accurate forecasts for solar electricity generation by applying machine learning (ML) to satellite imagery and numerical weather predictions. By undertaking the following work packages:Trial and prototype a functional solar forecasting systemTest delivered forecasts to the UK National Grid Control Room. Working closely with users to ensure we meet operational needs.Development and measure the impact on the grids carbon intensity and costs.The project will look to develop a Deep Learning machine learning model which takes a sequence of recent satellite images and numerical weather predictions. The model will output probabilistic solar electricity nowcasts for each PV system in the country, these will be calibrated in near-real-time using live solar electricity data. Deep Learning models can handle huge amounts of data, so we will train the model across the entire geographical extent of the satellite imagery (not just the areas which happen to have solar electricity systems). As such, the model will be trained to predict the next few frames of satellite imagery as well as solar electricity generation. The project is split into three different work packages:WP1 - DesignResearch & develop the use of machine learning & satellite images to nowcast PV power generation at GSP-level. Research will be conducted in close collaboration with academia & industry.WP1 will include the following:Use historical data to train machine learning models.Evaluate forecast skill using historical data.Compare against ESOs current approach; and ESOs approach + gridded NWPs; and CM SAF.ML model output: Probabilistic predictions for total solar PV power generation for each GSP at 5-minute intervals.The system will also estimate PV outturn now (situational awareness).Calibrate forecasts in near-real-time using live PV power data.Static designs for user-interface (UI)WP2 – Prototype DevelopmentResearch & develop a prototype of real-time PV nowcasting system. Research how to present nowcasts to Control Room users via an interactive web user-interface. Build a prototype of API & web UI & nowcasting engine capable of running in real-time.WP2 will include:Develop operational requirements with Control Room engineersResearch ways to run the nowcasting ML models every 5 minutes for all of the 1 million PV systems in the UK.Develop a suite of PV nowcast performance validation metricsBuild a functional prototype of nowcasting system & web user interface.Expose PV nowcasts via the API for PEF implemented in work package 1.The UI will give an overview of the countrys UK PV fleet and also allow the user to drill into details.WP3 – Prototype DemonstrationEvolve nowcasting system through multiple rounds of user feedback. Quantify impact on grid balancing.WP3 will include:Feedback, evolve, develop with Control Room engineers to learn what they do and dont like about the prototype PV nowcasting system. Quantify the effect of PV nowcasting on balancing costs & CO2 emissions (building on the Control REACT NIA project).Explore using probabilistic PV nowcasts to dynamically set reserves (working with the Dynamic Reserve NIA project).Explore ways to measure users interactions with the PV nowcasting system.Research ways to blend PV Live with PEFs half-hourly PV forecasts with OCFs 5-minutely PV nowcasts. The project proposed to complete the following objectives:Research report comparing the performance of our Deep Learning nowcasting system against other forecasting techniques (including ESOs current approach).Cost estimates for running business-as-usual (BaU).Prototype web UI & PV nowcasting service running in real-time. Including a suite of validation metrics.Report on the feasibility of running ML-powered PV nowcasts in real-time.Developed and socialized agile/CI/CD methodology.Workshop on implementing satellite-powered nowcasts BaU.Workshop on agile/CI/CD in critical energy applications.Joint proposal with ESO for candidate adjacent project with planning on cadence and target level of agile/CI/CD. | |
| Abstract | Using Deep Machine learning techniques, this project is exploring whether if we had more accurate predictions for solar electricity generation then we could reduce the amount of spinning reserve required. This would reduce carbon emissions and reduce costs to end-users, as well as increase the amount of solar generation the grid can handle | |
| Publications | (none) |
|
| Final Report | (none) |
|
| Added to Database | 19/10/22 | |
