Risk Assessment and Protection Technology Exploration of Collision Between Unmanned Aerial Vehicles and Power Grid Facilities

Preventing unmanned aerial vehicle (UAV) collisions with power grid facilities provides protection for UAV and power grid facilities. To prevent collisions, how to assess collision risk is critical. In this study, a novel collision risk assessment model for UAVs to power grid facilities was established by employing a recurrent neural network (RNN). The model used the time modeling capacity of the RNN model and the modeling power for complex relationships to model and forecast the UAV's flight paths and power grid facility data. After imperfection training with the prepared data and optimizing the parameters, the precision of collision risk assessment has increased. Our experimental results showed that the accuracy of the model's collision risk prediction was 94%–99%, and the RNN-based collision risk assessment model of UAV to power grid facilities based on the precision rate. In terms of the recall rate, the model significantly outperformed traditional support vector machine (SVM) model, and can better capture the collision situation more accurately and improve the effectiveness of collision risk assessment.