This project supports the drive to be an even more reliable network operator through improving the resilience during severe events. The initiative is split across two work streams:
Work stream 1 – Lightning into PowerON:
This work stream will trial a proof of concept where UK Power Networks’ Network Management System (PowerOn) will receive lightning strikes locations in real time via an API developed by an international weather consultancy. These locations will be linked to poles and to the network diagram using time and distance parameters to enable verification of the reason behind circuit-breaker opening and disconnecting the circuit. After this mapping is established, a dedicated alarm will be created in PowerOn to notify control engineers that the faulted circuit was likely struck by lightning. This could reduce CIs and CMLs associated with lightning strikes.
Work stream 2 – Resource estimation:
We will gain access to advanced weather forecast parameters from stations across the UK Power Networks licence areas. This work stream will trial the concept of using predictive analytics to combine historic fault data to weather parameters. This will build on and enhance our existing capabilities to forecast the impact of severe weather. The novelty will be within the development of the link between high frequency sampled weather data and the distribution network. This will ultimately drive a numerical prediction of weather related fault volumes and locations. These forecasts will provide a probabilistic view of the storm impact and quantify the expected level of risk each weather event presents.
As part of the project, the resources required to adequately manage the severe weather will be forecasted as well, by leveraging the valuable experience of our emergency planning team.
Ultimately, this will trial a system capable to forecasting, planning and responding to severe weather events more effectively and optimise resource requirements – thus reducing the impact of outages to customers and controlling costs.
Objectives
This project aims to:
1. Develop and trial solutions which support the readiness for severe weather through:
a. Work stream 1 – Linking near real time lightning data to the network; and
b.Work stream 2 – Analysing weather parameters and fault data to provide resource estimation and fault volume forecasts.
2. Determine how to present the outputs in a consistent, verifiable and easy to use manner for:
a. Work stream 1 – confirmation that a fault is a result of a lightning strike; and
b. Work stream 2 – probabilistic forecast of fault volumes and recommended resources.
Ultimately, the main project objective, consistent across both work streams, is to improve the network reliability during severe weather days and minimise the disruption to our customers.
Learnings
Outcomes
Work Stream 1: Lightning Fault Prediction. The project successfully built a module within PowerOn that links near-live time lightning data to network assets. If there has been a fault within a short period of time of lightning being registered, the system will now alarm to notify the control engineer that a fault was likely caused by lightning. However, within the duration of the trial, a smaller number of lightning strike events were coincident with network faults occurred than intended. The total number of predicted network faults from lightning was 37. At this time, no changes in network response to lightning-suspected faults are proposed. Before any change to network procedures is proposed, it is the intention to continue collecting lightning and fault data until a library of 200 events is compiled. It is believed that this will be sufficient for a detailed assessment of the technology and development of a safety case.
Work Stream 2: Resource Forecast. The project has completed initial statistical analysis for severe weather events from the year August 2020 to July 2021. The algorithm is able to signal when a high fault volume event is going to happen and in EPN and SPN. The majority of the time it is able to provide a reasonably accurate magnitude. It should be noted that this is a probabilistic forecasting algorithm and it is not designed to accurately forecast the exact number of faults in a given region but it should be able to provide useful signals to the Emergency Planning team and predict the magnitude of faults more accurately than an experienced human can, this will allow the emergency planning team to escalate the company’s response when appropriate and ensure resource allocation is optimised.
Lessons Learnt
The fault forecasting capability of the resource estimation tool is limited by the data it has access to. The tool is only looking at the relationship between the weather and the number of faults. In particular, the project team believes the accuracy of the tool could be improved with access to vegetation data as this is a major cause of faults when wind speeds are high.
The fault forecasting algorithm developed for Storm Resilience cannot be used for LPN as this licence area is almost exclusively served by underground cables. In the project, there was no visible relation between observed daily outages and daily accumulated precipitation in the LPN area. Future work could be completed to integrate the fault forecasting capabilities of multiple NIA projects, for example ‘Underground fault predictive model and earthing assessments’ aimed to predict the relationship between rainfall and underground faults. Some of these points can be revalidated during the next period of high rainfall.
For the EPN and SPN regions, there was a positive relation between the observed daily low voltage faults and the daily accumulated precipitation. At high precipitation levels, there was a slight overforecast of faults in SPN and a slight underforecast of faults in EPN. This was a result of the relatively small sample utilised in this project. Further monitoring is required to validate this point.
The project engaged with the customer services team at UK Power Networks to understand how they predict call volumes during a weather event. It was agreed that trying to predict the resource requirements in customer services was out of scope for Storm Resilience but a future project could look at using the fault forecasts to predict the number of calls customer services would receive. The call volume prediction could be used to help the customer services team optimise call handler resource allocation
