本文介绍了滚动神经网络模型，引入了有序导数的概念，讨论了此网络的实现过程。提出了迭代滚动递推ＢＰ神经网络的天然气产量预测模型，建立了气藏稳产期和递减期天然气生产能力预测的迭代递推滚动ＢＰ网络的数学模型通过验证表明，该模型计算结果达到了数值模拟的精度，对新区的天然气生产能力可做到较为精确的预测。 This paper introduces the rolling neural network model, introduces the concept of ordered derivatives, and discusses the realization of this network. The prediction model of natural gas production based on iterative rolling recursive BP neural network is proposed and the mathematic model of iterative recursive rolling BP neural network is established to predict the production capacity of natural gas in stable and declining gas reservoirs. The results show that the numerical results of this model reach the numerical simulation The accuracy of the new area of ​​natural gas production capacity can be more accurate prediction.