Random forest-based ultra-high voltage converter valve fault detection method and system

The invention discloses an ultra-high voltage converter valve fault detection method based on a random forest. The method comprises the steps of obtaining historical operation data of a converter valve and processing the historical operation data to obtain a training data set; constructing a converter valve fault detection initial model; calculating optimal parameters of the converter valve fault detection initial model; based on the optimal parameters, training the initial model by adopting the training data set to obtain a converter valve fault detection model; and adopting the converter valve fault detection model to carry out fault detection on the actual converter valve. The invention also discloses a system for realizing the fault detection method for the ultra-high voltage converter valve based on the random forest. According to the method, firstly, data preprocessing is carried out, redundant features are removed, features with high correlation are reserved, then the particle swarm optimization is com