Runoff prediction is an important task of water conservancy departments. In order to improve the reservoir long-term runoff forecasting accuracy, adaptive regulation ant colony system algorithm (ARACS) is proposed. The forecast model is set up by using an adaptive regulation ant colony system algorithm and the radial basis function (RBF) neural network combined to form ARACS-RBF hybrid algorithm, and then training the neural network by using the ARACS algorithm. It can automatically determine the parameters of the neural network from the sample data and form the reservoir long-term runoff forecast model based on the hybrid algorithm. Then the reservoir long-term runoff forecast was carried out by using the method and history runoff data. In long-term runoff forecasting such factors impacting long-term runoff as meteorology, weather, rainfall and season are comprehensively considered. The results indicate that the method can reflect the general trend of the stream flow truly, which provides a new method to estimate the no linearity of time series. The result shows the convergence of method is faster and forecast accuracy is more accurate than that of the traditional ant colony system algorithm-RBF neural network and RBF neural network. The method improves forecast accuracy and improves the RBF neural network generalization capacity; it has a high computational precision, and in 98% of confidence level the average percentage error is no more than 6.5%. The hybrid algorithm can be used efficiaciously in long-term runoff forecasting of the reservoir and river.