The invention discloses a method for detecting power consumption data abnormality. In view of the characteristics of time-correlation and high dimension of the current power consumption data, a powerconsumption data abnormality detection model is improved, when the power consumption data abnormality detection model is trained, on the basis of historical power consumption data, the data association information of the historical power consumption data is analyzed by using an LSTM network so as to train the power consumption data abnormality detection model after performing dimension reduction processing on the historical power consumption data, so that the power consumption data abnormality detection model adapting to the time correlation characteristics and high dimensionality characteristics of the power consumption data, the input power consumption data to be detected are detected by the power consumption data abnormality detection model to obtain a detection result, and compared with the traditional detection model, the detection accuracy is greatly improved, the false alarm rate is reduced, and the power consumption data abnormality detection links are simplified. The inventionfurther provides an apparatus and device for detecting power consumption data abnormality, and a readable storage medium, which have the above advantageous effects.