Predicting Overhead Line Failures Using AI and Weather Data in Norway

Organisation type

Public administration at national level

Use case area

Energy

Geographic scope

Norway

Domain

Economy
Environment

Challenge:

Statnett, the Norwegian electricity transmission system operator, faces significant challenges in maintaining the reliability of overhead power lines, as about 90% of all temporary failures are weather-related. Traditional methods for predicting these failures rely on static reliability criteria, which struggle to accommodate the increasing uncertainty brought by irregular energy sources and evolving eater patterns. With a focus on lightning, one of the major causes of failures, there is a pressing need for a more accurate and dynamic prediction model that can balance system reliability with operational costs.

Approach:

Statnett undertook a project to improve lightning-induced failure predictions with data-driven models. First, the historical weather and failure data were used to create fragility curves for overhead lines. These curves represent the relationship between weather events, such as wind and lightning, and the probability of failure. Using these fragility curves, a model was developed to estimate the probability of failure for each overhead line, initially applied for long-term planning through Monte Carlo simulations. The model was then adapted to use current weather forecasts to predict failure probabilities. Weather data was sourced from the Norwegian Meteorological Institute through an API. This weather forecast data was processed to map weather variables, such as wind speed, to the geographical locations of power line towers. The processed weather data was combined with the fragility curves to update the expected failure probabilities for each overhead line, i.e. calculating the probability of failure based on current weather conditions. A forecast service was developed to automate this process, running continuously and updating forecasts four times a day. The updated failure probabilities were visualised using an interactive dashboard displaying the probability of failure for each overhead line, aiding system operators in preparing for severe weather events.

Outcomes:

The data-driven model significantly improved the accuracy of failure predictions, allowing for better planning and resource allocation in the maintenance of overhead lines. The model successfully captured the seasonal variations and identified the nature of lightning-related failures, providing Statnett with a more reliable tool for long-term planning. The predictive capability meant that Statnett could anticipate and prepare for adverse weather conditions, reducing the likelihood of unexpected outages.

Data Sources:

In this use case, data from historical failure statistics, weather patterns, and lightning indices were meticulously analysed to derive insightful conclusions. The integration of these diverse data sources allowed for a holistic understanding of the conditions influencing failures, enabling more accurate predictions and informed decision-making.

Additional Information:

Point of Contact:

https://datascience.statnett.no/contact/