Nonlinear Francis Hydroturbine Generator Set Neural Network Model Predict Control

Due to the difficulty in describing the nonlinear characteristic of Francis turbine, this paper takes advantage of the powerful nonlinear approximate ability of the feed forward neural network to put up the Francis turbine neural network model (FTNNM) and the neural network identification model (NNIM) for the nonlinear Francis hydroturbine generator set (TGS) including FTNNM. The neural network model predicts control (NNMPC) uses NNIM to predict future response to potential control signals of the FTGS. An optimization algorithm then computes the control signals that optimize future FTGS performance. The Levenberg-Marquardt algorithm is used to train the FTNNM and the NNIM. The convergence speed of the offline training is fast and the accuracy of the model is high. The neural network model FTNNM reflect nonlinear characteristic of the Francis turbine truly and the neural network identification model NNIM reflect the nonlinear relationship between the inputs and outputs of the FTGS. Simulation results show that NNMPC is an effective tool for the nonlinear FTGS including FTNNM