The present invention relates to an improved fuzzy neural network-based short-term power load prediction method. The method includes the following steps that: power system parameters are collected; data samples are obtained; the number of the hidden layers of the method is obtained; normalization processing is performed on the input parameters of a neural network; and an excitation function-adjustable BP neural network learning algorithm is obtained, and a prediction result is obtained. With the method of the invention adopted, a power supply unit can accurately predict power load, especiallyrealize the ultra-short-term prediction and short-term prediction of the power load, can improve the impact of the load on a power system, and play an important role in making more rational power generation plans for a wind power farm, reducing costs and enabling the wind power farm to participate in power generation competition. The method is improved in the prediction accuracy of the power loadand the convergence rate of the prediction method, is of great practical significance and has a bright application prospect.