go to top scroll for more

Projects

Projects: Projects for Investigator
Reference Number NIA2_NGESO037
Title Forecasting the Risk of Congestion
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 ENGINEERING AND TECHNOLOGY (Electrical and Electronic Engineering) 100%
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 June 2023
End Date 31 December 2023
Duration ENA months
Total Grant Value £300,000
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/NIA2_NGESO037
Objectives "This project focuses on the probabilistic forecasting of congestion after the clearing of the day-ahead market, assuming that the day-ahead scheduled flows are known at the forecasting time. By quantifying the possible deviations between the scheduled flows and the physical flows, the project will assess the impact of congestion across the network and predict the probabilistic risk of congestion on specific branches of the power grid. Overall the project aims to provide a methodology and associated tool to assess the risk of congestion, also considering the uncertainties in the nodal physical flows. The project will assume that the day ahead scheduled flows are known at the forecasting time, with the hourly nodal injections and offtakes resulting from clearing of the day ahead market of the next day are known. The main deliverable will be a report documenting the methodology, the application to the UK power grid, and the analysis of the results. The developed tool will be tested on a sample of weeks and will provide the congestion risk profile for each pair of critical contingencies (i.e. assets considered for unplanned outages) and critical branches (i.e. transmission lines, cables, transformers monitored against overloading) identified during the project. Each risk profile will give both the risk of exceeding the thermal limit of the line, and the different excess scenarios with their respective probabilities. This risk profile will be compared against the binary contingency analysis of the Day Ahead Congestion Forecast, and the predicted risk will be tested against the congestion outcome from historical physical flows. Where confidential data sharing is required, the relevant data sharing requirements and procedures shall be followed. The following work packages (WP) will be completed: Work Package 1 – Development of forecasters for the interconnectors A supervised learning model to predict the probabilities of the physical flow deviations with respect to the day-ahead scheduled flows. This will focus on the interconnectors between the UK and EU countries with at least 2 years of operating time to give a good size of training data. This WP will include the following tasks: Data mining – to build the input features of the model, an extensive set of time series data will be mined. When the data are collected, they will be stored in a clean and structured database that can be referenced by the forecasting solution. All-time series data will be cast in the standard format using UTC timestampsProcessing of input features – once required manipulations have been performed, new data can be engineered from the existing data so that it can be used by the forecasting solution. A selection of time series used as input features will be built, and for each time feature the time lags used are specified.Development of the forecasting algorithm – select the relevant approach for the development of the algorithm to ensure that the chosen technology fulfils the project use case requirements. Anticipation that tree-based regression techniques are well-positioned to deliver the highest quality probabilistic forecast, performance will be compared against other multivariate models that also pass the identified use case requirement criteria.Testing of the model – the testing of the AI model must be done following strict rules to ensure the results are reliable and reproducible. A proportion of the dataset will be set aside and never used in the testing of the models until the very end, this is normally chosen as the chronologically last 20% of the dataset. To ensure model accuracy is not determined on the same data as the model is trained on, the remaining dataset will be further divided up into multiple training and validation datasets.Storage of the predictions, performance analysis and reporting – for the time-period in scope, the model will generate the probability distribution for the spread between the physical flow and scheduled flow for each interconnector between UK and an EU country, for each hour of the day, to be stored in the database. Model performance can be evaluated based on accuracy of the percentile predictions. With a focus on bias and variance, an analysis of the model performance will be conducted and summarised in a report. Work Package 2 – Generation of scenarios Generate explicit possible day-ahead scenarios for the physical interconnector flows, filter the produced scenarios, check the correlations, and interface the scenarios with the power grid model. This WP will include the following tasks: Generation of scenarios for physical flows associated with interconnectors – A sample of scenarios for physical flows will be generated to ensure the spread between generated flows and scheduled flows is distributed according to the AI models built in WP1. In this task it is more practical to work with unweighted scenarios, so scenarios will be generated with the same weight.Post-processing and testing of the scenarios – filter the scenarios to keep only the relevant information and reduce the burden of performing many load flow analyses in later WPs, and perform the necessary sanity checks on the filtered sample of scenarios. To perform reduction of scenarios without affecting the distributions, events in the higher probability region will be filtered out and weight of events kept in the sample will be rescaled using techniques in Monte Carlo generation. Tests will be documented to validate both the generation and filtering of scenarios.Preparation of interface with the power grid model – write scenarios in a format that can be analysed by the load flow solver module and interfaced with the power grid model Work Package 3 – Load flow analysis Reconstruct the probability distribution for the loading of the line for each contingency and each critical branch within the project scope. This WP will include the following tasks: Check load flow with solver – run the flow solver on the received grid model, checking the convergence for contingencies and validation of the power flow based on key reported physical quantities (voltage, power values, thermal limits of transmission lines)Interface with network topological changes – validated version of the grid model will be complemented with scenario data that represents changes in network topology over time. This includes but is not limited to changes in topology of the grid, elements in outage, and re-adjustment of operational setpoints of controllable assets in the system. To ease integration with the grid model, a consistent format must be defined to represent evolution of the network over different time stamps.Interface with generation and demand profiles – validated scenario module capable of adjusting the demand and generation profiles associated with the day ahead scheduled exchanges for each hour in scope. This includes but is not limited to load and generation scenarios, simplificationsto consolidate different generation units, and changes in the flow via interconnectors with EU countries.Load flow sanity checks – validate power flow results are consolidating the model grid with the scenario data from task 2 & 3 above. A set of criteria for acceptance must be defined at the start of this task to characterise the required level of accuracy underlying the power grid model with scenario data adapted in this project.Build the custom load flow solver for the generated scenarios – load flow will be assessed in the scope of the steady-state power flows. Depending on the number of scenarios for the physical flows, running an exact AC power flow for each scenario and contingency may be slow, techniques to overcome this speed issue will be investigated if required.Reconstruction of the risk of congestion – risk profiles associated with the loading of each critical branch and contingency in scope. Process the outcome of the power flow, collect the loadings of the critical lines under different scenarios and contingencies, and store results in a database using a suitable format. Work Package 4 – Analysis of results and benefits Analyse the predicted probability distribution for congestions, run the comparison with the outcome from the point forecast and assess the ability of the probabilistic congestion forecast to better anticipate historical N-1 congested cases. This WP will include the following tasks: Comparison with current Day Ahead Congestion Forecasts – quantitative assessment of the congested lines in the probabilistic approach, compared with the assessment in the current congestion point forecast Comparison with congestion resulting from physical flows – quantitative comparison between anticipated risk of congestion based on the probabilistic forecast, and the security analysis based on the actual physical flows in the interconnectorsCheck actions of the operator – for selected cases, feedback from the control room operators on what may have been the different actions triggered by the probabilistic forecast had it been available. Actions taken by operators are based on a high level of skills and experience, so it may be challenging to quantify this impact.Work Package 5 – Address uncertainties from wind generation Update the overall procedure leading to the prediction of the congestion risk profile for each contingency and critical branch in scope to include uncertainties from key wind generation units.This WP will include the following tasks: Update the probabilistic forecasting model – forecast algorithm upgraded with the prediction of the spread for key wind generation units Update the scenario data – regenerate scenario data covering the uncertainties in the physical flows for key wind generation unitsLoad flow analysis and impact on the congestion risk – updated congestion risk profiles after folding the uncertainties from wind generation, and quantification of the impact from the wind. This includes re-running the load flow solver on the new set of scenarios, reconstructing the congestion risk profiles for all critical branches and contingencies in scope, and assessing the extent the resulting congestion risk profiles differ from those obtained from just uncertainties from interconnectors. In line with the ENAs ENIP document, the risk rating is scored Low: TRL Steps = 1 (2 TRL steps)Cost = 1 (<£500k)Suppliers = 1 (1 supplier)Data Assumptions = 2Total = 5 (Low) " This project will develop new probabilistic forecasts to anticipate the possible spread of values between the day ahead scheduled energy flows and the actual energy flows. The project will first consider development of forecasters for the interconnectors, with the addition of uncertainties from key wind generation units in the final work package. The associated probabilities as well as the correlation between the spread values at different nodes of the grid will also be tracked. The goal is to predict the risk of congestion on specific branches of the power grid with a probabilistic approach. This will be done by using power flow models to propagate the probabilities of injections and offtakes at different nodes of the grid and applying them into current scenarios on internal lines of the power grid. "Forecast probabilities of deviations between day ahead scheduled flows and actual energy flow for each interconnector connected to the EUCritical contingency and critical branch pairs identified for analysis within the projectGenerate a sample of scenarios suitable for analysis of the load flow solverDevelop a tool tested on a sample of weeks and provide the congestion risk profile for each pair of contingency and critical branches identified Interface load flow tool with network topological changes, and generation and demand profiles, ensuring solver can run in timescales suitable for operational useCompare results from probabilistic distribution for congestions with existing point forecast, and consider impact on potential operator actionsUpdate overall procedure to include uncertainties from key wind generation units "
Abstract The UK power grid is becoming more interconnected and this, together with increased contribution from renewable energy sources, poses some challenges in the anticipation of energy flows. The volatility inherent to interconnections and renewable energy increases the uncertainty of physical energy flows, complicating the anticipation of internal congestion in the day-ahead market and resulting in more decisions needed within day by the control room to overcome congestions. This project focuses on the probabilistic forecasting of congestion after the clearing of the day-ahead market, assuming that the day-ahead scheduled flows are known at the forecasting time. By quantifying the possible deviations between the scheduled flows and the physical flows, the project will assess the impact of congestion across the network and predict the probabilistic risk of congestion on specific branches of the power grid.
Data

No related datasets

Projects

No related projects

Publications

No related publications

Added to Database 01/11/23