“Sliding” Power System Risk Accumulation Model Identification Based on GAF and Improved CNN

In the context of challenges such as large-scale integration of renewable energy and dynamic complex changes, risk assessment of power systems is crucial to ensure the safe operation of the electric grid. As the grid becomes increasingly complex, traditional risk assessment methods struggle to handle large volumes of data and adapt to dynamic changes. In grid operations, incidents are often the result of gradual accumulation of multiple risk factors, necessitating timely monitoring and evaluation of these dynamic changes. Modeling and identifying the "Sliding" mode of power system risk accumulation captures the progressive nature of risks, allowing for early detection of potential hazards, preventing emergencies, and ensuring grid stability and reliability. This study introduces a method combining the Gramian Angular Field (GAF) and an enhanced convolutional neural network (CNN) to effectively identify the "Sliding" mode of power system risk accumulation. Initially, the grid's health is analyzed comprehensively from both macro and micro perspectives, creating an integrated risk index to quantify overall system risks. The power grid risk accumulation model introduces the definition and characteristics of the "Sliding" mode for studying risks during grid operation. Using the GAF method, time series data from the "Sliding" mode, simulated using IEEE 9 and IEEE 39 systems, are transformed into images to extract deep features. An improved CNN model then identifies this risk mode. Comparative analysis demonstrates that the combination of GAF and the improved CNN efficiently and accurately identifies the "Sliding" risks, providing robust support for grid risk assessment and emergency response.