Grid Monitoring Alarm Event Recognition Method Integrating Expert Rules and Deep Learning

SCADA collects a large amount of alarm information during the power system operation. If the controller cannot quickly judge the monitoring alarm events, it may cause the expansion of the accident. Therefore, we propose a grid monitoring alarm event recognition method combining expert rules and deep learning. The method includes two stages: expert knowledge base determination and deep learning recognition model determination. In the first stage, the characteristics and rules related to monitoring alarm events are sorted, the expert knowledge base is constructed, and the alarm events are initially identified. In the second stage, the preliminary judgment results and the alarm information are combined into the convolutional neural network (CNN) recognition model, thus the final recognition result is obtained.

I. Introduction

As the scale of the power system continues to expand and becomes more complex, real-time alarm information collected by SCADA to each substation continues to increase. What’s worse, during alarm events, situations such as the equipment rejection and misplace, data loss or distortion while uploading information, etc. are far beyond the judgment of the dispatcher. It becomes challengeable for the dispatcher to identify the alarm events quickly and accurately. Therefore, the intelligent grid monitoring alarm event recognition method has become an important research content to improve the level of grid monitoring and operation.