In order to improve the accuracy of short-term electric load forecasting, a cycle forecasting method for short-term electric load forecasting is proposed based on a radial basis function network-style coefficients autoregressive model with an exogenous variable(RBF-ARX)model. First, the short-term electric load forecasting was regarded as a nonlinear time series prediction problem, and an autoregressive model(ARX model)of electric load forecasting was established based on historical load data. Then, the ARX model parameters were approximated with the RBF neural network and were estimated with an off-line structured nonlinear parameter optimization method(SNPOM). Finally, based on this, a cycle forecasting method for short-term electric load forecasting was established. The proposed method was used to predict the short-time electric load in a certain city of Hunan Province. The predicted results were compared with the actual load values. The results show that the proposed method has high accuracy, reliability, and practicability.