Project Summary
P4R aims to “significantly improve network planning, modelling and forecasting capabilities” to “deliver the next generation of user driven digital products […] across transmission and distribution”. Through an advanced indication of where inclement weather will affect the network and a better prediction of expected fault numbers, P4R will enable resources (engineers, mobile generation, welfare provisions, customer liaison staff, mobile catering for consumers etc) to be proactively placed in those areas most likely to be impacted, something that is especially important in island locations where travel distances are significant. This is expected to bring forward the travel time to those faults where an onsite presence is required, enabling power supply to be restored sooner than is currently possible. This creates a more resilient network, minimises disruption for customers and brings about financial, social and environmental benefits.
The Discovery Phase demonstrated the potential benefits of the solution. In Alpha, the project utilised data science capabilities and experience in design and software development to further refine the prototype, resulting in a Fault Forecast Engine which uses cutting-edge statistical methods and an Interface that closely meets the user needs.
In Alpha, a prototype fault forecasting model and accompanying data infrastructure was implemented which verified the feasibility of the P4R’s innovations. These statistical models have shown forecasts are highly accurate in days 1-7, a key timescale for operational planning for the Control Room. It was also found that the model successfully predicted severe weather events resulting in large numbers of faults.
There was also a strong focus on engagement with users and User Experience (UX) experts to co-design the solution’s UX to ensure that it meets user requirements, such as the display of the entire fault probability distribution instead of the RAG status to aid decision-making.
Following the successful demonstration of the prototype’s efficacy, Beta will evolve the prototype into a commercial solution that can be rolled out across GB DNOs and beyond, improving resilience with data-driven fault forecasting and decision-support.
Our perception of the problem has not changed. While this technology has been demonstrated in Alpha, there is still further work to refine the solution to ensure its interoperability so that it can be used beyond SPEN / SHEPD. In Beta, we are therefore turning our attention to industrialising the solution and improving the fault forecasting, utilising a broader range of data from both other networks as well as new weather sources to provide a higher resolution reanalysis.
Previous user engagement and literature reviews have demonstrated that an improvement in fault forecasting capability is achievable with further iterations and this broader data input. By running trials with multiple DNOs in parallel with the modelling enhancement, the project team can integrate novel solutions and validate any findings during the trial. The ambition is not only a robust software solution that is interoperable with each DNOs IT architecture but also the evolution of a range of additional, complimentary novel features before the solution is adopted into BAU.
SPEN and SHEPD – will play a key role relating to the control room requirements, ensuring the required features are correctly developed into a viable solution. SMEs within the businesses will input into the functionality testing of the prototype and review any project outputs. They will continually capture learnings from the trials to understand how P4R is driving business decisions and what benefits it is yielding. They will be the end solution users in BAU.
Sia Partners – they bring the relevant data science and software development expertise and experience required for this project. They will build the infrastructure supporting the solution, the software itself, the industrialisation of the models and the user interface. They will also prepare the transition to BAU and set up the organisation for further rollout
University of Glasgow – has world-leading expertise in the application of forecasting in the energy sector and statistical methods required to develop the fault forecasting capability envisioned in P4R. Their deep multidisciplinary expertise spanning energy systems, meteorology and statistics make them an invaluable partner in this project.
The ambition and expectation is that the end software solution will be fit for all GB and international DNOs, as well as any adjacent sectors who suffer weather-related interruptions.
Following some recent high-profile storms, DNOs must be now seen to make a step-change in how they prepare and react to severe weather. Additionally, there is pressure for DNOs to exceed previous performance with network resilience a growing priority from both the regulator and consumers who are increasingly dependable on their power supply and renewable generators who need a reliant connection to the network.
By providing Control Room operatives short-term predictions regarding the expected level of faults in each district across the licence area, DNOs can better prepare for a storm and restore power supply sooner than is currently possible and minimise disruption for customers.
Innovation Justification
DNOs must minimise outage periods and the associated CMLs. With currently available information and BAU processes, estimating the impact of severe or extreme weather on maintenance and fault volumes relies on user judgment. P4R is developing a fault forecasting system that will provide short-term fault forecasts and early warning of fault volumes up to one week ahead. P4R combines multiple complex data sources with novel statistical learning techniques to produce accurate fault forecasts. Furthermore, by quantifying forecast uncertainty, P4R empowers users to act and manage risk proactively for the first time, aided by a user interface developed in collaboration with control room engineers. Beta phase SIF funding is required to operationalise and enhance the prototype developed in Alpha and trial the system in a live environment.
Most relevant state of the art product: Existing commercial offerings are unproven with limited features, such as lacking fault-specific forecasts and detailed uncertainty quantification. No past innovation projects have considered probabilistic fault prediction and related decision-support, leaving a significant gap in DNOs' predictive capability. The one related fault forecasting product on the market today has been developed for the US for very different network design, weather conditions and user needs, and so far has a limited client base. The key technical feature of P4R, accurate fault prediction for GB’s electricity networks (and data) and unique weather conditions, has not been demonstrated by any other product.
Compared to the most relevant state of the art product, the fault forecasting capability being developed by P4R represents a step-change in forecast quality and utility. Engagement with end-users in Discovery and Alpha has produced a detailed specification for forecast information and communication which is not met by any existing product or service. Users require predictions not just of the most likely number of faults, but the probability of exceeding key thresholds that impact service levels and decision-making. P4R forecasts the probability distribution of the number of faults at multiple time points over the next seven days and communicates this via traffic-light system for key thresholds. Our testing of the Alpha prototype correctly predicted the risk of “red” fault levels (the most severe) 85% of the time at one-day-ahead and provided early-warning of this risk 5-7 days ahead in the majority of cases.
Beyond incremental innovation: P4R represents a major shift in practice away from exclusively human interpretation of basic weather forecasts to sophisticated data-driven analytics. End user assessment has verified that P4R forecasts are actionable and expected to lead to multiple immediate benefits, which will be enhanced by continuous improvement, opportunities for which have been identified in Alpha (see appendix).
Integration and Commercial Readiness: The current IRL is 4 (quality and assurance of integration) and estimated to be IRL-7 (verified and validated) at the end of Beta. The current CRL is 6 (product optimisation) and estimated to be CRL-8 (market introduction) at the end of Beta.
Scale, SIF objectives and the relevant Innovation Challenge: The proposed scale of P4R’s Beta phase represents the multi-party effort required to industrialise the code base, data pipelines and user interface prototyped in Alpha, add further innovative enhancements, and run live trials to verify the performance and utility of the system in an operational setting with the view to scaling the solution to meet the needs of other sectors and stakeholders in due course. Once implemented, P4R will bring immediate benefits to network customers though enhanced resilience and robustness. This project directly addresses the Data and Digitalisation Innovation Challenge by making novel use of data and digital platforms to significantly improve “network planning, modelling and forecasting capabilities”. The P4R forecasting system is a “user driven next-generation digital product” designed to improve weather resilience. P4R is presently focused on electricity distribution, but engaging with electricity transmission (NGET, SPT) and the rail sector (Network Rail), and potentially others in the future.
This project cannot be funded elsewhere or considered as part of BAU activities because the solution is as yet unproven and requires innovation support to both increase the TRL and enable BAU adoption, something not permitted within price control funding.
P4R does not undermine the development of competitive markets. It will bring a new product to an existing albeit nascent market, and far from undermining the development of this market, it will stimulate the market by injecting fresh competition.
Counterfactual approaches have been explored throughout Discovery and Alpha, using multiple candidate weather data products, statistical modelling approaches, and user-interface designs, amongst others. Quantitative (accuracy metrics) and qualitative (user feedback) have been employed to identify and develop the most appropriate solutions. Alpha concluded (Deliverable 3) that multi-model ensemble weather forecasts and non-parametric modelling produce the most accurate fault predictions, and that a “traffic light” system with options to drill-down into forecast data provides users with optimal user experience and access to information, thus facilitating decision-making.
Benefits
By accurately predicting how many and where network faults are likely to occur up to 7 days in advance, P4R will have resources onsite earlier, enabling power supply to be restored sooner than is currently possible and minimising disruption for customers to bring about financial, social and environmental benefits.
Having established those faults where P4R would enable an earlier restoration time, the size / scale of that time saving was applied to potential benefits. The impact of the new restoration time was modelled for the relevant faults and the new value compared against the counterfactual.
The modelling utilised actual fault data from the last 5 years and the values for actual CML costs have been taken directly from SPEN's Reporting System for Exceptional Event Claims. Should the number of extreme weather events vary significantly from these 5 years then the benefits realised would also vary.
Following this modelling, the following benefits are anticipated with Year 1 taken as 2026 when BAU begins, and the discount starting in 2021. The values below reflect the preferred option of 2 DNOs generating benefits from P4R.
‘Cost savings to consumers (£m)’
A loss of power supply increasingly inconveniences individuals and businesses, especially the vulnerable and with working from home becoming more prevalent. Living without modern conveniences can cause anxiety, with appliances’ battery-life diminishing, particularly for the medically dependent. Places of work may be closed, causing lower-income families further stress because of lost wages. P4R’s social benefits have been calculated by multiplying the expected CML savings from P4R by the societal CML value in the CBA.
Year 1: £583,008
Year 3: £3,254,791
Year 5: £6,011,850
Year 10: £12,128,932
Cost reductions in operating the networks and wider energy system (£m)
CML Savings: DNOs are set targets for the number of unplanned CMLs on their networks (the outage duration multiplied by the number of customers affected). Performance against these targets is linked to financial rewards and penalties. Having resources strategically positioned ahead of time in those areas most likely to be impacted will lead to a reduction in outages which in turn creates a financial CML payment saving to the DNO. These figures do not include extreme weather events which are excluded from the incentive mechanism.
Year 1: £66,995
Year 3: £374,016
Year 5: £690,837
Year 10: £1,393,766
Guaranteed Standards of Performance (GSP) savings: GSP sets out how quickly DNOs must restore power following an interruption in supply. Should DNOs not meet these standards, customers are entitled to statutory compensation. With quicker restoration, some customers’ outage duration will therefore be brought under a GSP time threshold, creating a financial saving to DNOs.
Year 1: £19,868
Year 3: £110,918
Year 5: £204,873
Year 10: £413,333
Storm Support: As part of DNO’s Storm Support, they provide meal vouchers, alternative accommodation (eg hotels) and warm packs (eg hats, gloves, blankets, torches etc.) to customers who are off supply for extended periods. P4R will drive a reduction in the need to provide Storm Support to affected customers, creating a financial saving to SPEN.
Year 1: £47,580
Year 3: £265,630
Year 5: £490,638
Year 10: £989,865
Fuel Savings: When power supply restoration is anticipated to be longer than usual, DNOs will provide onsite diesel generators to supply back-up power to its customers while that fault is repaired. P4R will result in both a reduction in the need to provide a generator as well as shortening any time they are required. A reduction fuel consumption creates a financial saving to SPEN.
Year 1: £7,730
Year 3: £43,157
Year 5: £79,715
Year 10: £160,825
Carbon reductions – direct or indirect (MTCO2e)
Renewable generators need a reliant connection to the network to operate and sell their electricity. Long network outages therefore prevent renewable generation accessing the grid with potential carbon impacts.
Additionally, as part the Storm Support, when any power restoration is anticipated to be longer than usual, DNOs aim to provide onsite diesel generators to supply back-up power to its customers while that fault is being repaired (the assumption was taken that diesel generators were provided to 1 in 5 faults when the outage went beyond 8 hours). P4R is expected to deliver a potential improvement to those restoration times resulting in both a reduction in the need to provide a generator as well as shortening any time they are required. This reduction diesel has an associated carbon emission reduction benefit.
Year 1: £3,416
Year 3: £19,435
Year 5: £36,442
Year 10: £76,230
Predict4Resilience also anticipates other, wider benefits which are captured within the Benefits Map. These include the associated benefit for other agencies. In a storm scenario, DNOS work with a wide range of partners and some vulnerable residents are relocated.
