The invention relates to a short-term power consumption prediction method based on I-LSTM, and belongs to the field of power system prediction. The method comprises the following steps: S1, collecting historical data of a power system, and processing incomplete data and abnormal values; S2, dividing the data into a training set, a verification set and a test set according to time; S3, constructing an (Improved Long Short Term Memory) I-LSTM network model, and inputting the training set into the I-LSTM network model for training; S4, setting a network loss function, an optimization algorithm, a learning rate and a batchsize of the I-LSTM network model; and S5, predicting the test set, and obtaining a prediction result of the test set according to a model obtained according to the accurate change condition of the verification set. According to the method, the key information in the user historical data sequence and the characteristic relationship between the user power consumption data can be better mined, and the user power consumption prediction precision and stability are effectively improved.