Abstract: Under the current context of new renewable energy power stations with high proportions of renewable energy integration, the spatiotemporal coupling characteristics between meteorological factors and renewable power generation will further amplify the fragility of grid operations. Therefore, this paper proposes an RPA-based automated data collection design for renewable energy power stations. First, a power output fluctuation load model for renewable energy power stations is constructed, followed by the design of RPA-based automated data collection for these stations. Additionally, a deep learning-based risk early warning system for renewable energy power stations is proposed. The study demonstrates that when inputting line power flow characteristics and weather data features, the model's predicted power station line conditions exhibit high consistency with actual conditions, achieving a prediction accuracy rate of 90%.