The invention belongs to the technical field of power system load prediction, and particularly relates to a power system short-term load prediction method based on an IEEMD-BPNN (improved ensemble empirical mode decomposition-back propagation neutral network). The power system short-term load prediction method includes: step 1, using IEEMD to decompose non-steady-state and nonlinear historical power load data to acquire a series of stable intrinsic mode function (IMF) components and a tendency allowance; step 2, removing first-order IMF components (IMF1) acquired in step 1; step 3, using a BPNN to predict IMF components and tendency allowances, at each order, acquired in step 1 and step 2 respectively; step 4, using the BPNN for nonlinear combining of the IMF components and the tendency allowances, at each order, acquired in step 3 to acquire a final power load prediction result.