Classification model training method for identifying cavitation phenomenon of water turbine

• Main Authors: HUANG XI, LIANG QUANWEI, WANG ZHIYONG, WANG GANG, WEN JUNJIE, LUO YANCHEN, REN PANQIU, CHEN ZIHAO, WANG GUIHONG
• Format: Patent
• Language: Chinese; English
• Published: 10.03.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a classification model training method for water turbine cavitation phenomenon identification, and belongs to the field of machine learning. Comprises: inputting a training sample; performing binary classification on the input training sample data according to a supervised learning method; solving a maximum spacing hyperplane for a training sample, adding a slack variable, mapping a sample of a low-dimension input space to a high-dimension space by using a dimension raising function to enable the sample to be linearly separable, and searching an optimal classification hyperplane in the feature space; searching an optimal hyper-parameter in the dimension raising function by using an exponential cyclic decline matrix search method; the optimal hyper-parameter is used for training on the whole training set again, and a final classifier is obtained; and storing the trained classification model and applying the classification model to the recognition of the primary cavitation of the water t