A CNN-Based Warning Information Preprocessing Method for Power Grid Fault Diagnosis Analytical Model

摘要

The access of large-scale renewable energy and UHV AC/DC hybrid connection has become the status quo, and the grid structure has become more and more complex. It is urgent to apply smart grid and big data technology to grid fault diagnosis. Existing analytical models used for power grid fault diagnosis, when faced with far more alarm information than protection rules, the information cannot be directly applied to the model, and the alarm information needs to be analyzed manually. The modeling is cumbersome and complicated, which restricts online applications. Therefore, in order to realize the end-to-end alarm information automatically corresponds to the analytical protection rules in the analytical model, a CNN-based alarm information preprocessing method is proposed. First, oriented to the alarm information collection at the time of failure, establish a key fault information extraction model to extract key information related to the analysis rules of the analytical model in the alarm information; then, oriented to the key information in the fault alarm information, establish a key fault information classification model, the key information is classified according to the analysis rules of the analysis model, and the attributes of the alarm information and the analysis model are compared. The effectiveness of this warning information preprocessing method is verified by power grid simulation data.