Abstract: Natural gas hydrate has the advantages of high gas storage rate, low pollution and large reserves, which has good prospects for development. In addition, natural gas hydrates formed in the process of natural gas processing and transportation will lead to the problem of serious such as pipeline blockage. Therefore, it is of practical significance to analyze and predict the formation of natural gas hydrate. In order to predict the formation of natural gas hydrate, aiming at the merits and demerits of previous research on the prediction methods of natural gas hydrate formation, the combination model of GM (1,1) theory and Back Propagation neural network is established by using MATLAB computer language, which is based on the artificial neural network with the ability to solve complex system problem.Considering the improvement of prediction accuracy, the difference combination method is used to combine the two methods. The GM (1,1), back propagation neural network and the combined model are used to predict and compare the pressure data obtained in the experiment. In order to further verify the accuracy of the combined model, Markov chain model is selected for the prediction test. The results show that the combined model of GM（1,1）and BP neural network has higher canprecision and this method can be widely used in many directions, which can provide a theoretical basis for the development and utilization of NGH in the future.

Key words: Natural gas hydrate, The grey forecast, The back propagation neural network, Difference combination method, Markov chain model

摘要： 天然气水合物具有储气率高、污染低、储量大等优点，具有良好的发展前景，但是在天然气加工和运输过程中形成的天然气水合物会造成管道堵塞等严重状况，因此，分析和预测天然气水合物的生成具有实际意义。为了预测天然气水合物的生成情况，针对前人研究天然气水合物生成预测方法的优缺点，引用了具有解决复杂系统问题能力的人工神经网络，运用MATLAB语言编程建立了灰色理论（Grey Forecast）理论和BP神经网络（Back Propagation Network, BP）的组合模型。为了提高预测精度，选用了差值结合法将两种方法结合，分别运用GM（1,1）、BP神经网络以及此组合模型对实验中得到的压力数据进行预测并加以比较；为了进一步验证组合模型的精准度，选用了马尔科夫链模型进行预测检验。结果表明，GM（1,1）和BP神经网络组合模型具有较高的精准度，且此方法可以广泛运用到较多方向，可为今后的天然气水合物开发利用提供理论依据。

关键词: 天然气水合物, 灰色理论, BP神经网络, 差值结合法, 马尔科夫链模型