go to top scroll for more

Projects: Projects for Investigator
Reference Number	NIA_NGSO0032
Title	Control REACT
Status	Completed
Energy Categories	Other Power and Storage Technologies(Electricity transmission and distribution) 100%;
Research Types	Applied Research and Development 100%
Science and Technology Fields	PHYSICAL SCIENCES AND MATHEMATICS (Computer Science and Informatics) 50%; ENGINEERING AND TECHNOLOGY (Electrical and Electronic Engineering) 50%;
UKERC Cross Cutting Characterisation	Not Cross-cutting 100%
Principal Investigator	Project Contact No email address given National Grid plc
Award Type	Network Innovation Allowance
Funding Source	Ofgem
Start Date	01 July 2020
End Date	01 July 2021
Duration	ENA months
Total Grant Value	£400,000
Industrial Sectors	Power
Region	London
Programme	Network Innovation Allowance
Investigators	Principal Investigator	Project Contact , National Grid plc (100.000%)
	Industrial Collaborator	Project Contact , National Grid plc (0.000%)
Web Site	https://smarter.energynetworks.org/projects/NIA_NGSO0032
Objectives	Workstream 1 – REACT (led by Smith Institute)The first three work packages of Workstream 1 are structured around demonstration of increasing advanced versions of a Proof of Concept (PoC) for a prototype visualization tool called REACT (Real-time Error and Cost Tracking). The concept under investigation is that binary SORT files can be processed to present useful and visually appealing forecast error and cost information for CR engineers. This concept will be proven in a Python implementation of REACT and delivered alongside plans for its future integration into the Control Rooms Business Network.For each of work packages 1-3 the deliverables will include: An agreed cost model for forecast errors. Agreed visualizations to be included in the REACT PoC. Demonstration of the REACT PoC. Report summarizing learning and insights gained in work package. There is an option dependent on innovation funding in RIIO2 to proceed with Work Package 4 in April 2021, it will focus on support to NGESOs implementation of the REACT PoC as a tool in the Control Room. Workstream 2 – Probabilistic Forecasts (led by TNEI)Probabilistic forecasts are predictions that include uncertainty quantification; in contrast to more familiar “point” forecasts which provide a single-valued prediction with no indication of how probable errors of any size may be. Workstream 2 will explore how existing point forecasts can be extended to produce innovative probabilistic forecasts for demand and wind, and will demonstrate how these probabilistic forecasts could lead to more efficient decisions in the Control Room.It consists of four work packages and the deliverables will include: Work Package 1 – Initiation: Project feasibility study including literature review. Work Package 2 - Demand: Scripts and models for producing probabilistic demand forecasts Work Package 3 - Generation: Scripts and models for producing probabilistic generation forecasts Work Package 4 – A report detailing the methods for producing these forecasts, evaluating them, and showing the benefits of using them in decision-making. Workstream 1Error Identification and Quantification: The Control Room receives a number of data streams to inform the actions to balance the grid. Many of these data streams are forecasts. Identifying forecast errors and quantifying them, with a view to understanding their impact on the cost of balancing the grid is the main aim of this workstream. A selected number of forecast data sets will be analysed. For each category of forecast the errors will be considered (i.e. difference between forecast and actual) at time horizons of 8 and 4 hours ahead and 90, 60, 30 and 10 minutes ahead. Cost Modelling: The costs associated with forecasting errors will initially be modelled using simple linear relationships agreed with the NGESO Subject Matter Experts (SMEs) for use in the development of the first basic PoC. In later versions of the PoC, more sophisticated cost models will be developed. For example, to more accurately represent the fact that certain types of error lead to more costly interventions to reduce the risks to system stability. This could include consideration of the costs associated with unwinding balancing instructions and emergency power plant use. Visualisations and User Interaction: The method of presenting information about forecast errors and their cost impacts to the CR is not yet agreed. Throughout the project, Proof of Concept visualisations will be developed by the Smith Institute and regularly demonstrated to the NGESO stakeholders and the other project partners to obtain feedback on its fitness for purpose and suggestions for improvement. The initial PoC will provide visualisation of forecast errors and estimated contributions to balancing costs. The extent to which the user can interact with the visualisations will also be explored and incorporated into the Proof of Concept developed and implemented in the final PoC version.Proof of Concept Implementation: A Proof of Concept (PoC) for REACT will be implemented by the Smith Institute in the Python programming language in this workstream. The PoC will assume that the eventual deployable visualisations will sit in the CRs Business Network and that the required data will be streamed from the SORT and SPICE systems. We envisage using open source Python packages to develop the visualisations in the PoC. Integration Planning: The PoC created in Work Package 3 is not intended for deployment directly into the CR. Following the development of the PoC, a draft plan for the implementation and integration activities will be prepared. Work Stream 2Development of new and sophisticated forecasting models has delivered increased accuracy within NGESO. However, this has been through the improvement of single value forecasts representing “what we expect will happen” and managing the system on that basis. Material additional value can be achieved by explicitly acknowledging the uncertainty within these forecasts, and making decisions that account for this uncertainty. In this workstream, the focus will be on developing probabilistic forecasts for demand and generation to demonstrate that this is possible, and to explore how these could be implemented in future and used within a more advanced decision support framework. Initiation: This WP will review relevant literature, develop methodology further and assess implementation challenges, engaging with NGESO Subject Matter Experts (SMEs). This may include exploring aspects such as forecast lead-times, the possibility of using ensemble Numerical Weather Prediction, how to model generation without operational metering and the most appropriate approach to probabilistic forecast construction e.g. demand net of embedded generation, separate DG and VG and demand.Demand: WP2 will explore algorithms for constructing probabilistic forecasts of demand, using historic data comprising of demand point forecasts, actual demands, weather forecasts and actual weather variable values. This will include formulating forecast models, and writing the code/scripts to produce and validate forecasts. Furthermore, it will be necessary to account for the spatial correlation between different locations if the use-case (decision-support) in question involves multiple locations, power transfers between locations, or aggregations of probabilistic forecasts. Generation: In WP3, state-of-the-art methods for probabilistic wind and solar forecasting will be developed, building on work already done by University of Strathclyde. Methods for capturing the correlations between generation and net-demand will also be explored. This will be necessary for use-cases involving multiple forecasts, as in WP2. Demonstration and path to deployment: The principle that using probabilistic forecasts can result in more efficient decisions will be demonstrated using real historical data and the cost modelling approach developed in workstream 1. This will support the development and scoping of an advanced decision-making framework for forecasting, which, if successful, will be explored further in follow-on work. Implementation challenges will be updated based on the outputs of WP 2 and 3 and strategies developed to address these. Objective 1: provide insight into the cost impacts of the forecast errors. Allowing NGESO to prioritise schemes for improving forecasting accuracy and managing uncertainty in future, such as those which will be suggested as outputs from Workstream 2 of this project. Objective 2: prototype enhancements to the current Control Room capability in managing uncertainty by developing visualisations of forecast errors and their associated cost impacts (REACT PoC). Objective 3: show how existing point forecasts can be extended to produce probabilistic forecasts for demand and windObjective 4: demonstrate that using probabilistic forecasts can lead to more efficient decisions.
Abstract	The uncertainty that Control Room (CR) engineers must handle in their decision-making is growing rapidly due to increases in renewable and embedded generation. At the same time, the CR has seen a huge rise in the number of units involved in their balancing decisions (from 40 to over 1,000). It is inevitable then, that the costs of balancing the grid has also been rising and will continue to do so until an approach is adopted which allows CR engineers to effectively manage uncertainty. It is believed that if information about forecast uncertainty was presented in real-time to CR engineers, that this would provide opportunities for them to make more economic and secure balancing decisions.
Publications	(none)
Final Report	(none)
Added to Database	02/11/22