Data-Driven Turbine Modeling and Load Rejection Analysis

Abstract

The full characteristic model of hydraulic turbine must be considered in the research of control and transition process calculation of hydraulic turbine generator unit. In order to obtain the full characteristic model of hydraulic turbine, it is indispensable to reasonably extend the high-efficiency working condition characteristic area to the low-efficiency area according to the comprehensive characteristic curve. In this paper, BP(back propagation) neural network is used to process the comprehensive characteristic curve of hydraulic turbine to obtain the full characteristic model. In view of the defects of traditional BP neural network, such as slow convergence speed, long training time and easy oscillation in the learning process, BP neural network is improved. The improved BP neural network is used to train the model of flow and torque characteristics, and then the full characteristic model of hydraulic turbine is trained by using the value point to extend the flow and torque characteristic data. Experiments demonstrate that the full characteristic model established by the improved BP neural network algorithm has higher accuracy. Finally, the full characteristic model is used to calculate the load rejection transition process. The results demonstrate that the full characteristic model is suitable for the calculation of hydraulic turbine transition process and meets the requirements of engineering application.