The invention discloses a photovoltaic array fault detection method based on Grubbs criterion and outlier. The photovoltaic array fault detection method based on Grubbs criterion and outlier includes the steps: A, acquiring the real-time current and voltage of each string of a photovoltaic array and the irradiation and temperature of the photovoltaic array every 5 seconds; B, establishing a photovoltaic array simulation model, brining the acquired irradiation and temperature into the model to obtain the reference current and voltage; C, making a difference between the practical current and the reference current, forming one array through combination of the difference of each string of the photovoltaic array, detecting the abnormal data point through application of the Grubbs criterion, and recording the fault characteristic value of the abnormal data as 1, or recording the fault characteristic value of the abnormal data as 0; D, forming a one-dimensional array through combination of the current difference every 20 seconds in order, obtaining LOF of each current difference through application of an outlier algorithm, and distributing the LOF factor to each string according to the time; and E, according to the results of the step C and the step D, comprehensively determining whether a fault occurs. The photovoltaic array fault detection method based on Grubbs criterion and outlier can detect the fault of a photovoltaic assembly in real time, especially the early fault.