The invention discloses a washing machine running non-intrusive identification method based on active power fluctuation. The identification method comprises the steps of sampling the voltage and currents of an inlet wire of a total power supply within a certain sampling frequency range, forming a voltage signal sampling sequence u and a current signal sampling sequence i, and calculating an average power sequence P; constructing a large window W for the average power sequence P, the large window being capable of being divided into m uniform small windows wk and each small window containing n discrete active power points; solving for the difference between the maximum value and the minimum value of the small windows wk in the large window, defining the difference as a range Dk, giving a threshold D0, and conducting statistics on the number of small windows M satisfying Dk>D0 in the large window W; if M>m/2, defining the window as a fluctuation window; and conducting statistics on three consecutive large windows, and if two of the large windows are fluctuation windows, then determining the washing machine is running. According to the invention, the identification and accuracy of the washing machine is greatly improved.