The invention discloses a method and system for carrying out intelligent fault detection on an acquisition operation and maintenance system, and the method comprises the steps: carrying out the feature extraction of the obtained operation state data of each intelligent acquisition device, and obtaining a feature data set; performing dimension reduction processing on main components in the characteristic data set of the intelligent acquisition equipment, and forming a main component set by using the first two main components in the characteristic data set of each intelligent acquisition equipment; performing visualization processing on the running state data; calculating a Gaussian kernel density local outlier factor value of each intelligent acquisition device; and carrying out descendingorder arrangement on the obtained Gaussian kernel density local outlier 1 factor value of each intelligent acquisition device so as to determine the fault probability of each intelligent acquisition device. According to the invention, under the condition that the operation states of all the devices are unknown, the outlier, namely the faulted device, can be found by analyzing the relationship between the devices. The method has higher detection accuracy and a more stable judgment threshold value, and is less influenced by local data distribution.