Deep learning technique enhances lightning risk prediction for power grids

Lightning is one of the primary causes of transmission line trips, posing a significant threat to the safety of power grids. However, due to the complexity and sporadic nature of lightning, achieving accurate forecasts has always been a challenge.

 

Recently, researchers at the China National Energy Key Laboratory of Lightning Disaster Detection, Early Warning and Safety Protection，as well as the Laboratory of Lightning Monitoring and Protection Technology of State Grid Corporation of China, have made significant breakthroughs in lightning prediction. By developing a deep learning–based nowcasting model, they can effectively predict the location and frequency trends of organized thunderstorms, providing robust support for predicting lightning risks to power grids. This research has been published in Atmospheric and Oceanic Science Letters.

 

The research team utilized wide-area lightning monitoring data from the State Grid Corporation of China and geostationary satellite imagery, combined with Convolutional Gated Recurrent Unit (Conv-GRU) networks and attention mechanism modules, to develop the lightning nowcasting model.

 

“Our model not only accurately predicts where lightning will occur, but also forecasts its frequency. It has shown excellent performance in predicting a winter thunderstorm in Central China and a spring tornadic thunderstorm in South China,” says Dr Fengquan Li, the first author of the paper.

 

Dr Jian Li, the academic leader of the laboratory, points out that, “In the future, we plan to enhance the accuracy of our lightning prediction model by integrating more data sources related to lightning formation, and further optimizing the model framework. This will better support the prediction of, and protection against, lightning disasters affecting power grids.”

---

---