Abstract: The seasonal load of city gas networks is one of the most fundamental design parameters in the peak-shaving of the underground gas reservoirs. According to the specialty of gas load, a forecast model based on the artificial neural networks toolboxes has been formed. It can express both characteristics of increment trend of time series and fluctuation from indefinite factors. The results through actual example show that the built model is of better convergence on computation and forecast preciseness, better applicability and flexibility than those of other models.