Analysis of Optimal Parameters for Discriminating Cavitation Types by SSAE-RF

Abstract

Hydro power has many advantages, such as pollution-free, relatively mature technology and high security. Hydro turbine is the core component of hydro power station. Cavitation has always been one of the main threats to the safe operation of hydraulic turbine units. In order to improve the overall classification and recognition accuracy of cavitation noise signal features of hydro turbine, a deep learning algorithm model based on stack sparse coding combined with random forest is proposed, and the optimal parameter selection of the algorithm model is analyzed in detail, so as to enhance the efficiency of deep learning algorithm. A group of cavitation noise signals with different intensities are selected. On the premise that other parameters of the algorithm remain unchanged, and choose a different parameter each time of training, and the optimal parameters of the algorithm are finally found; The default parameter algorithm model and support vector machine algorithm are used for comparison. The results show that the deep learning algorithm is superior to the traditional classification and recognition algorithm, and the overall classification and recognition accuracy of the deep learning algorithm using the optimal parameters is effectively improved. The research results can be used as an important reference for the deep learning algorithm to extract, classify and identify the cavitation noise intensity characteristics of hydro turbines.