Energy system power outage sensitive user identification model based on CNN+LSTM

In the context of deep integration of CPSS (Cyber Physical Social Systems), energy system data presents multi-source, complexity, and dynamic interactivity. To solve the problem of identifying power outage sensitive users, we propose a power outage sensitive user analysis and identification method based on CNN+LSTM. Firstly, perform preprocessing such as cleaning and structuring of power load data to ensure data quality; Next, conduct correlation analysis to explore the intrinsic relationship between the factors and characteristics affecting power load and the sensitivity to power outages; Then, the coefficient correction method is used to extract the user load curve and optimize the feature weights to enhance the adaptability of the model; The final design is a power outage sensitive user recognition model based on CNN+LSTM, which integrates time series and spatial features to achieve accurate recognition of power outage sensitive users. The experimental results show that in multiple experiments covering multidimensional data such as household electricity consumption and energy consumption, this method effectively improves the accuracy of anomaly detection, with an average power outage sensitive user recognition rate of 95.93%. It performs well in key indicators such as recall rate and F1 score, providing strong support for energy system optimization management and user service.