Research on running state recognition method of hydro-turbine based on FOA-PNN

摘要

To effectively monitor the operating state of hydro-turbine, a diagnosis strategy based on the operating conditions and pressure pulsation of the turbine is proposed. The improved Hilbert-Huang Transform (HHT) method is used to study the characteristics of pressure pulsation under different operating conditions. The physical parameters of pressure pulsation are extracted through the mutual information theory. Procedures include optimizing the smoothing factor sigma of the Probabilistic neural network (PNN) network through the Fruit fly optimization algorithm (FOA), constructing the FOA-PNN network model, classifying the unit operating status. The result shows that when sigma = 0.23, the prediction accuracy of the FOA-PNN network is 100%, and the training time is 0.336372 s. It is proven that the FOA-PNN can predict the running state of the turbine in a short time and monitor the running malfunction in real time.