The invention provides a short-term traffic flow forecasting method based on a gray ELM neural network. The method comprises the steps that collected data are grouped to acquire equal dimension innovation sequences, and then accumulating is carried out to acquire an accumulated equal dimension innovation sequence; the accumulated equal dimension innovation sequence is processed to acquire the input matrix and the target output matrix of a network; the weight and threshold of the network are randomly generated; the network parameters are set; the generated input matrix set and target output matrix set of the network are input into the neural network, and the network is trained; test data are input to acquire the forecasted output result of the network; and the network forecast result minus the cumulative value of the equal dimensional innovation sequences to acquire an actual forecast result to complete the forecast. The method provided by the invention has the advantages that the input data are processed by a gray model; the difference is smaller; and the forecasting accuracy of the gray ELM neural network is greatly improved.