Hydroelectric generating set multichannel degradation trend prediction and model training method

The invention discloses a hydroelectric generating set multichannel degradation trend prediction and model training method, and belongs to the technical field of hydroelectric generating set state prediction. According to the scheme, firstly, a multi-channel health model is constructed based on working condition data and vibration data of each channel of a hydroelectric generating set shaft system; calculating through a multi-channel health model to obtain a theoretical health value, and comparing the theoretical health value with the vibration data monitoring value to obtain a degradation degree; calculating the correlation between the channels through the deterioration degree of each channel; and finally, constructing a multi-channel degradation trend prediction model based on the time graph convolutional network, taking the degradation degree trend sequence of each channel and the correlation between the corresponding channels as a training set, taking the prediction label value as a target variable, and c