摘要
针对天然气水合物相平衡问题,文中提出用基于带动量因子的BP神经网络进行计算和预测。首先用遗传算法优化确定BP神经网络的结构和参数,得到最优化结构的神经网络;其次结合Levenberg-Marquart优化算法,建立天然气水合物相平衡计算及预测的神经网络模型;最后以实验测定的(CH4+CO2+H2S)三元酸性天然气水合物体系的平衡数据为训练和预测样本进行了计算。计算表明,预测结果与实验数据有良好的一致性,而且由于BP神经网络作为所谓的“纯粹”的算法不需要热力学模型,这对于相平衡计算是非常方便的,所以是研究天然气水合物相平衡计算及预测的一种新的有效方法。
The BP Artificial Neural Network （BP-ANN） with motivation rate is proposed to calculate and predict the phase equilibrium conditions. Firstly, the genetic algorithm （GA） is used to optimize and determine the structure and parameters of the BP-ANN, thus a network with optimized structure is obtained; Secondly, combined with Levenberg-Marquart algorith, the BP-ANN model for calculating and predicting phase equilibrium of natural gas hydrate is established; Finally, equilibrium data of （CH4＋CO2＋H2S） ternary sour natural gas hydrate are calculated. It is found that the calculated results agree well with experimental data, moreover, as a pure algorithm, the BP-ANN does not need any thermodynamic model, and the method is convenient to implement, which show that the presented method is efficient for calculating and predicting phase equilibrium of natural gas hydrate.
出处
《海洋技术》
2007年第2期54-56,75,共4页
Ocean Technology
基金
国家自然科学基金资助项目(20376078)
关键词
BP神经网络
天然气水合物
相平衡
算法
预测
BP Artificial Neural Network
natural gas hydrate
phase equilibrium
algorithm
prediction
