Identification of weather patterns and transitions likely to cause power outages in the United Kingdom

The study addresses how large-scale atmospheric circulation patterns relate to weather-induced power outages in the UK, aiming to improve predictability and preparedness. Adverse weather is a leading cause of power system faults, producing both high-impact/low-probability and low-impact/high-probability events, as exemplified by the August 2019 UK lightning-triggered outage. Prior approaches often model component fragility to weather or examine reliability impacts but do not explicitly link failures to specific circulation regimes (“weather patterns”). Weather patterns, being more predictable than their specific surface impacts, can be used to infer likely consequences. This work leverages 30 Met Office daily weather patterns to identify which patterns and transitions elevate outage risk by cause and season, thereby informing operational preparedness and long-term resilience planning.

Previous research estimated outage likelihood via exposure and fragility of components and analyzed impacts of extreme weather on reliability and failure rates. Some studies used representative extreme scenarios to assess component status and planned for renewables considering spatiotemporal weather variability. However, explicit associations between power system failures and large-scale weather patterns were lacking. Weather pattern classifications have been widely used to isolate predictable circulation components and relate them to impacts including coastal and fluvial flooding, lightning activity, and volcanic ash risk. They show useful predictability on sub-seasonal timescales. In the UK, studies linked wind gusts to wind/gale faults and explored spatiotemporal clustering and potential climate change impacts on faults, yet without pattern–fault linkage. Prior work also proposed probabilistic impact assessment of lightning using weather patterns and examined future changes in pattern occurrence and resultant precipitation and coastal flooding changes.

Data: Power system failures were obtained from the National Fault Interruption Reporting Scheme (NaFIRS) for Great Britain from 2010–2019, categorized by direct weather-related causes (e.g., lightning, wind/gale, snow/ice, rain, flooding). Weather patterns were the 30 daily Met Office circulation types derived via k-means clustering of MSLP fields (1850–2003), extended using ERA5 reanalysis to 2020 (further extended to 2022 for this study). Seasonal definitions follow astronomical seasons. Core computations: For each cause x, pattern w, and season s, the frequency of outages F_xws equals the sum of daily failures during pattern w (Eq. 1). The contribution of each cause to seasonal totals is assessed via R_x (Eq. 2). Pattern transition probabilities P(w_k|w_{k-1}) are taken from historical daily classifications (1950–2020) to compute the probability of subsequent patterns (Eq. 3), extended to multi-day chains for longer-duration phenomena (Eq. 4). Trend strength is quantified as S(w_k|w_{k-1}) = P(w_k|w_{k-1}) P(w_{k-1}) for one-day transitions or products over chains for multi-day sequences (Eqs. 5–6). For immediate-effect hazards (wind/gale, lightning, rain), Eq. (3) is emphasized; for cumulative hazards (snow/ice, freezing fog/frost, solar heat, flooding), Eq. (4) and lagged analyses are emphasized. Methods can be shifted to analyze outages occurring days after a given pattern. Scope: The main analysis focuses on the top three winter causes by frequency to ensure robust trend detection. To reduce potential duplication bias in the GB NaFIRS dataset (multiple locations reporting the same event), two regional NaFIRS datasets were analyzed for corroboration: Northeast England (2004–2021) and Southern Scotland (2017–2022). Visualizations include 2D histograms of pattern frequencies up to 14 days preceding outages and Sankey diagrams of pattern transitions/persistence associated with outages.

• Scale and seasonality: Nearly 70,000 weather-induced failures were analyzed (GB, 2010–2019). Seasonal total counts (approximate from Table 1): Winter 24,854; Spring 12,076; Summer 14,748; Fall 16,788. Seasonal cause shares: Winter—Wind and gale 74.62%, Lightning 12.15%, Snow and ice 8.84%; Spring—Wind and gale 38.94%, Lightning 34.45%, Snow and ice 20.45%; Summer—Lightning 67.55%, Wind and gale 24.27%; Fall—Wind and gale 69.91%, Lightning 19.16%, Snow and ice 4.74%.
• Winter top causes and counts considered: Wind and gale (18,546), Lightning (3,020), Snow and ice (2,198).
• High-risk patterns in winter:
  • Wind and gale: Over 50% of outages occurred during or shortly after patterns 26 (very cyclonic north-westerly/south-westerly depending on context) and 30 (very cyclonic westerly).
  • Lightning: Over 50% of outages occurred during or the day after patterns 20 (cyclonic westerly), 23 (unbiased westerly), 26 (very cyclonic north-westerly), and 30 (very cyclonic westerly).
  • Snow and ice: Nearly one-third of outages occurred a few days after pattern 27 (anticyclonic easterly); additional relevant cold, unsettled patterns include 24, 28, and 29.
• Transition/persistence structures (winter):
  • Wind and gale: Dominant pathways are 26→26 (persisting stormy south-westerly/north-westerly) and 30→30 (persisting stormy westerly); important transitions also include 20→20, 20→26, 20→30, 30→26, 23→20, 23→23, 23→26.
  • Lightning: Dominant persistent westerly regimes 20→20, 23→23, 26→26, 30→30, reflecting unstable westerlies with strong winds.
  • Snow and ice: Dominant cold/blocked transitions include 24→24, 20→26, 27→27, 27→28, 28→28, and 29→29, representing northerly to easterly flows and cold pools over the UK.
• Notable climatological transition: In winter, a common transition with 17.65% probability is 20→26, indicating progression of a deep low to the north of the UK; high-impact types tend to persist or transition among themselves, supporting predictability.
• Regional corroboration: Northeast England and Southern Scotland analyses show broadly consistent high-risk patterns for wind/gale and lightning (greater roles for patterns 20 and 23 in NE; pattern 26 prominent for lightning in SS) and snow/ice (pattern 27 consistently important; contributions from 26 and 28 vary regionally).

Linking outages to large-scale circulation types reveals that high-wind and high-precipitation westerly regimes (patterns 20, 23, 26, 30) dominate wind/gale and lightning-related failures in winter, while colder, snowy regimes (24, 26, 27, 28, 29) underpin snow/ice outages. The seasonality of pattern occurrences explains both the predominance of failures in winter and the exclusion of some patterns from key impact sets due to lower seasonal frequency. The persistence of high-impact patterns and their tendency to transition among similar high-impact regimes enhance the feasibility of a pattern-conditioned forecasting approach, which can leverage the superior predictability of patterns at sub-seasonal lead times to anticipate elevated outage risk. These insights support operational preparedness (mobilization of resources and preventive strategies weeks ahead) and strategic planning (investment targeting resilience against the most impactful regimes and considering projected changes in pattern frequencies).

The study identifies specific Met Office weather patterns and transitions that are most likely to lead to weather-induced power outages in the UK, especially in winter. High-risk westerly cyclonic regimes drive wind/gale and lightning outages, while cold blocked regimes are associated with snow/ice outages. Because these patterns and their transitions are forecastable with useful skill up to several weeks in advance, the findings enable development of a pattern-conditioned outage risk forecasting system and improved DNO preparedness. The methodology also translates readily to climate change projections by mapping future changes in pattern frequencies to changing outage risks. Future work includes operationalizing a forecasting tool (e.g., building on the Met Office Decider product), extending analyses to other seasons and causes with severity metrics, and integrating updated/expanded NaFIRS datasets to enhance robustness.

• Data reporting: GB NaFIRS may include multiple reports for the same meteorological event across locations, potentially inflating connections; regional datasets were used to mitigate and document uncertainty.
• Data access: NaFIRS data are proprietary and not publicly available; regional datasets cover differing time spans, introducing heterogeneity.
• Focus on frequency: The analysis targets frequent (often lower-impact) events to detect robust trends; it does not directly quantify customer impact severity in the main results.
• Seasonality and sampling: Some patterns are rare in certain seasons, limiting statistical power for those cases; high-number patterns occur more often in winter, contributing to seasonal imbalance.