Neural network model for the condition assessment of hydro turbines

Abstract

During regular operation, hydro turbine elements have complex operating conditions. In order to prevent major failures, among other common preventive measures, it is necessary to make a useful model of hydro turbine. Development of this model was focused on real elements and data set of real operating history. To solve this, an integral diagnostic approach is used. Based on the real diagnostics data of the condition and data history of the hydro turbine shaft and the designed life expectancy, a multi-layer perceptron (MLP) based artificial neural network (ANN) is built. The idea with this model was to enable to life assessment of the turbine shaft in any particular moment during regular operation. This paper describes a model for estimating the condition of the shafts of turbines of the current generator in Hydropower plant Derdap 2. The significance of this approach is that real experimental data support development of ML ANN (number of neurons and layers) topology, which is optimal for this model, training and testing. Results obtained from this model are used for decision-making about the planning time schedule of maintenance actions, as well as reducing overhaul time and direct losses.