Anomaly Monitoring of Hydropower Units under Complex Conditions

Hydropower has now become a fundamental industry and infrastructure for the social and economic development of our country. While hydropower brings convenience and benefits, it also comes with significant safety hazards. Therefore, real-time condition monitoring of hydropower units is necessary. This paper addressed the complex operational conditions of hydropower units by first dividing the data into mixed datasets containing both normal and abnormal data, and independent datasets where normal and abnormal data were separated. For the mixed datasets with both normal and abnormal data, the upper and lower bounds for anomaly detection were obtained using the standard deviation criterion based on normalized weighting and Monte Carlo simulation. For the independent datasets that follow a normal distribution, the dynamic upper and lower bounds for anomaly detection were determined based on the standard deviation criterion. For the independent datasets that do not follow a normal distribution, anomaly monitoring and calculation were performed based on multi-factor indicators. Experimental results show that when the proposed methods were applied to actual data from hydropower units under complex conditions, the accuracy of anomaly detection was relatively high, resulting in better monitoring of abnormal conditions.