Adverse weather significantly impacts power systems globally, causing failures and widespread blackouts. The UK experienced a notable outage in August 2019, costing over £15M and leaving millions without power for days. Power systems are vulnerable to a range of weather conditions, highlighting the need to understand weather-induced outages to improve preparedness. Previous research focused on individual component failure probabilities based on weather exposure and fragility modeling, or considered extreme weather scenarios to determine component status. However, these studies didn't consider the influence of large-scale atmospheric circulation patterns (weather patterns) on outage occurrences. Weather patterns offer a valuable tool, as they are more predictable than their surface impacts. While some efforts utilized weather forecasts or examined meteorological conditions leading to major outages, none explicitly investigated the relationship between weather patterns and power system failures to improve outage prediction and preparedness. This study uses 30 Met Office daily weather pattern definitions to analyze the relationship between these patterns and UK power outages (2010-2019) from the National Fault Interruption Reporting Scheme (NaFIRS) data, aiming to identify trends to mitigate the impact of weather incidents on power systems and to inform future reliability and resilience investment decisions based on predicted changes in outage frequency under future climate scenarios.

Existing literature explored the probability of power outages caused by extreme weather using exposure and fragility modeling of individual system components. This approach models failure probability as a function of component exposure to weather variables over time, applied to different extreme weather events. Some studies used representative extreme weather scenarios to determine the operating and damage statuses of specific power system components. Other research investigated the effects of adverse weather on power system reliability and failure rates, and some considered weather variability for cost-effective long-term power system planning with high renewable energy shares. However, none of these considered the relationship between different types of weather-induced power system failures and specific large-scale atmospheric circulation patterns (weather patterns). Weather patterns have been widely used to isolate the most predictable atmospheric circulation components and link them to surface impacts, exploiting their higher predictability compared to impacts themselves. Successful applications have been demonstrated in sub-seasonal predictions. Some previous works used weather forecasts or identified meteorological conditions leading to major outages to provide data for distribution network operators (DNOs), but none examined the relationship between weather patterns and power system failures to predict and prepare for outages.

This study uses daily data from January 1, 2010, to December 31, 2019, from the National Fault Interruption Reporting Scheme (NaFIRS) in Great Britain and weather patterns provided by the Met Office during the same period. The analysis computes the frequency of weather-induced outages by weather phenomenon, weather pattern, and season. The frequency of weather-induced power system failures (*F_xws*) is calculated as the sum of failures (*f_xwt*) attributed to weather phenomenon *x* during weather pattern *w* and time interval *t*, within season *s*. The contribution of *F_xws* to the total number of outages (*F_r*) is denoted by *R_x*. The probability of a weather pattern's occurrence (*P(w_k)*) is calculated recursively, conditioning on the previous day's pattern (*P(w_k|w_k-1)*). For phenomena with prolonged effects (e.g., snow and ice), a generalized equation (*P(w_k)*) is used, considering the *N* preceding days. The strength of the trend relating a weather-induced failure to a weather pattern transition or persistence is calculated using equations (5) and (6), which consider the probabilities of pattern transitions and persistence. The analysis is extended to consider periods after specific weather patterns to account for cumulative effects of some weather phenomena. To mitigate the potential inflation of connections between patterns and faults due to multiple reports of the same event, the analysis incorporates shorter NaFIRS datasets from specific regions (Northeast England, 2004-2021; Southern Scotland, 2017-2022).

The analysis reveals strong relationships between specific weather patterns and weather-induced power outages. In winter, weather patterns 26 (very cyclonic north-westerly) and 30 (very cyclonic westerly) are strongly associated with wind and gale outages, accounting for over 50% of occurrences. Lightning strike outages show similar associations with patterns 20 (cyclonic westerly), 23 (unbiased westerly), 26, and 30. Snow and ice outages are mainly linked to weather pattern 27 (anticyclonic easterly), with nearly a third of outages occurring a few days after this pattern. These trends are consistent across Great Britain, Northeast England, and Southern Scotland, although with varying contributions from specific patterns. Sankey diagrams illustrate weather pattern transitions and persistence associated with the most common outage causes. For example, for wind and gale outages, persistence of patterns 26 and 30 are prominent trends, as are transitions between patterns representing unsettled westerly, north-westerly, or south-westerly flows. Lightning strikes show persistence of unstable westerly patterns, while snow and ice show trends related to cold, unsettled northerly flows. These patterns align with their characteristic high wind speeds, precipitation, and snowfall. Seasonal variation in pattern frequencies partly explains discrepancies between pattern characteristics and outage trends; high-numbered patterns (more frequent in winter) are associated with more outages.

The findings demonstrate a strong link between specific weather patterns and power outages, addressing the research question by identifying high-risk patterns and transitions. This is significant for improving prediction and preparedness. The use of weather pattern forecasts, which have lead times of up to several weeks, offers a substantial improvement over current methods relying on short-term weather warnings. This allows for proactive resource mobilization (emergency generators, mobile substations) and preventive strategies (grid reconfiguration, generation re-dispatch). The results support the development of a pattern-conditioned fault forecasting system, leveraging the predictability of weather patterns and their persistence or transition to high-risk patterns. This could also incorporate future climate projections to assess climate change impacts on outage risks. The study highlights the value of NaFIRS data for such analyses and advocates for increased data sharing to improve grid safety, asset management, and resilience enhancement.

This research identifies high-risk weather patterns and transitions associated with power outages in the UK, offering a significant improvement in outage prediction and grid preparedness. The methodology, leveraging the predictability of weather patterns, facilitates proactive resource management and preventive strategies. Future work could focus on developing a pattern-conditioned fault forecasting system and incorporating climate change projections to assess future risks. Increased data sharing from UK DNOs would further enhance the accuracy and impact of such analyses.

The study is limited by the availability of NaFIRS data; data sharing from DNOs is voluntary, resulting in variations in dataset time spans across regions. Multiple reports of the same event within the GB NaFIRS dataset could lead to inflated connections between patterns and faults, although this was addressed using regional data. The analysis focuses on the most frequent outage causes, potentially overlooking less frequent but high-impact events. The methodology may require further refinement for regions with different weather patterns and power grid characteristics.