Artificial neural networks for predicting the onset of overload instability in Francis turbines

Overload instability is a self-excited phenomenon that occurs in Francis turbines working over the Best Efficiency Point. It provokes huge power swings and pressure fluctuations in the hydraulic circuit. One particular issue is that this phenomenon appears suddenly and just before its onset the machine can operate in a very stable manner. In this study, we show that artificial intelligence techniques such as Neural Networks can be used to evaluate the risk of overload instability several seconds before its appearance. Experimental data, acquired during several overload instability tests in a huge prototype, has been used. The techniques proposed in this paper could be used in advanced condition monitoring systems and could permit a safer operation of the turbine working at high loads.