The invention discloses an abnormality detection method for intelligent electrical equipment based on a virtual window, and the method comprises the following steps: S1, obtaining the feature data of the intelligent electrical equipment at a fixed sampling frequency, and carrying out the sequential arrangement of the feature data of the intelligent electrical equipment according to the time sequence; S2, determining the position and size of a feature data virtual window through a manual method or a system default automatic method; S3, judging whether there are the conditions that two continuous feature data values exceeds the boundary in the feature data in the virtual window or not: directly determining that the intelligent electrical equipment is abnormal if there are the conditions that two continuous feature data values exceeds the boundary in the feature data, or else solving a deviation value according to a deviation formula; S4, determining that the intelligent electrical equipment is normal if the deviation value is less than a critical value, or else determining that the intelligent electrical equipment is abnormal. According to the invention, the method achieves the analysis of the intelligent electrical equipment through the virtual window, and achieves the quantitative analysis of the feature data of the intelligent electrical equipment through a statistical method. The method not only can analyze the quantitative relation between the feature data and a reference value, but also can analyze the relation among the feature data, thereby determining the performance of the intelligent electrical equipment more completely, and avoiding a misjudgment condition that may be caused by a conventional threshold comparison method.