> ## Content Index
> Fetch the complete content index at: https://www.stirup.si/llms.txt
> Use this file to discover other available public pages before exploring further.

# Forecasting space weather risks on power grids
- URL: https://www.stirup.si/forecasting-space-weather-risks-on-power-grids-63e9a822/
- Published: 2026-10-12T02:42:36.000Z
- Updated: 2026-10-12T02:42:36.000Z
- Description: A machine learning system developed by Microsoft Research can predict space weather risks for power grids up to 30-60 minutes in advance, using solar-wind data and geological factors to estimate risk at 66,935 U.S. substations.
- Author: Satish Das
- Tags: AI, Stirup Research

## Introduction to Space Weather Forecasting for Power Grids

Space weather events, such as geomagnetic storms, can have significant impacts on Earth's power grids, GPS systems, and satellite operations. These events are driven by solar activity, which can alter the Earth's magnetic field and induce currents in power transmission networks. Predicting the effects of such events is critical for grid operators to take preventive measures and minimize damage. A new machine learning system developed by Microsoft Research aims to provide early warnings of potential risks by analyzing solar-wind data and local geological conditions.

## Machine Learning Pipeline for Risk Estimation

The system uses an end-to-end machine learning pipeline that integrates solar-wind observations, forecasts of the Auroral Electrojet (AE) and Disturbance Storm Time (Dst) indices, and local geological data. These inputs are combined to estimate the rate of magnetic-field change (dB/dt), which is a key indicator of geomagnetically induced currents (GICs) that can damage power infrastructure. The pipeline generates location-specific risk estimates for 66,935 substations across the continental United States, providing grid operators with actionable insights up to 30-60 minutes before a storm arrives.

## Key Components of the System

The system operates in three main stages. First, solar-wind measurements from the L1 Lagrange point are used to generate forecasts of the AE and Dst indices. Second, a gradient-boosting model combines these forecasts with local geological and grid infrastructure data to estimate dB/dt. Finally, the predictions are converted into location-specific risk estimates and aggregated into a continental risk assessment. This approach allows for a more nuanced understanding of how different regions may be affected by space weather events.

## Evaluation and Performance

The system was evaluated using data from the 2020-2026 period, during which it demonstrated strong performance in predicting geomagnetic activity. The AE predictor outperformed several empirical solar-wind-based approaches, capturing extreme-event activity used by the GIC forecasting system. The Dst predictor also provided a valuable signal for large-scale geomagnetic storm strength. The GIC risk stage achieved detection rates of 76.5% for major events, 81.2% for severe events, and 64.1% for extreme events, though false-alarm rates increased with storm severity.

## Implications and Future Work

This work highlights the potential of physics-grounded machine learning in improving the accuracy and timeliness of space weather risk assessments. By providing location-specific information, the system can help utilities prioritize engineering reviews and implement targeted protective actions. However, further validation with operational data and grid operators is needed before the system can be deployed in real-world grid operations. Future work includes extending forecast horizons, scaling the system internationally, and integrating it more deeply with grid operations.

**Source:** [Microsoft Research](https://www.microsoft.com/en-us/research/blog/forecasting-space-weather-risks-on-power-grids/?ref=stirup.si). Analysis based on publicly available information; no hands-on testing is claimed.