The invention discloses a GA-BP neural network-based switch cabinet insulation defect detection method and system. The switch cabinet is an air switch cabinet, and the method comprises the following steps: building a training data set which comprises the type and content data of gas in the air switch cabinet during normal operation and the type and content data of gas in the air switch cabinet during partial discharge under different insulation defect types; constructing a GA-BP neural network model; training the GA-BP neural network model by the training data set to obtain an identification model with a function of identifying the insulation defect of the air switch cabinet; periodically measuring the components and content of gas in the air switch cabinet to be detected; and identifyingthe measured gas components and content by using the identification model, and outputting an identification result. According to the method provided by the invention, external interference can be avoided, so that the accuracy is improved.