The invention discloses a personalized user short-term load forecasting algorithm based on a decision-making tree. The algorithm is characterized in that the algorithm comprises the following steps: carrying out fluctuating component identification on user load, and judging whether the sawtooth fluctuation thereof is obvious; if so, entering a fluctuating component extraction step, and then, obtaining power consumption mode number; and if not, directly obtaining the power consumption mode number. For the user, of whom the power consumption mode number is larger than 7, a nearest daily load forecasting method is adopted; for the user, of whom the power consumption mode number is within the range of 2-6, a user side short-term load forecasting method based on power consumption mode excavation is adopted; and for the user, of whom the power consumption mode number is only 1, a cluster forecast reduction method is adopted. According to the algorithm, by carrying out data mining on historical load of the user, the power consumption modes of the user are extracted; the personalized user short-term load forecasting algorithm is established according to the number of the power consumption modes; and therefore, accurate user short-period load prediction is realized.