Self-Adaptive Anomaly Detection Method for Hydropower Unit Vibration Based on Radial Basis Function (RBF) Neural Network

摘要

In order to improve the adaptability and effectiveness of anomaly condition recognition for hydropower unit, an adaptive anomaly detection method of hydropower unit vibration is presented based on radial basis function (RBF) neural network. The optimal value of vibration parameters in real-time condition is dynamically computed by using RBF neural network in this method. The relative distance between vibration real data and optimal value is calculated as the anomaly. This index can describe the changes of vibration parameters and identify anomalies of hydropower unit condition. The obtained results of abnormal alarm can meet the actual demands by using the proposed method in vibration monitoring of hydro-power unit. This method can well describe the slow process of deterioration for vibration parameters and identify abnormal vibration in a sensitive manner. This method will be practical as to the operation guarantee of hydropower unit.