基于T-S模糊神经网络的烟道气驱建模

Flue gas flooding modeling based on T-S fuzzy neural network

对于油藏注采的优化,通常采用油藏数值模拟软件(CMG、VIP和Eclipse)对分别给定的不同开采方案进行油藏数值模拟,然后选择相对较好的开发方案。针对烟道气驱机理模型复杂、涉及许多高阶耦合偏微分方程组、难于差分求解和优化的问题,提出了一种基于混合螺旋优化算法(HSO)和T-S模型的模糊神经网络建模方法。该方法通过混合螺旋优化算法优化神经网络的权重参数以及高斯隶属度函数的中心和宽度参数,提高了模型精度。利用该方法建立烟道气驱的辨识模型,通过数值仿真计算验证了该建模方法在建立烟道气驱模型上的准确性和可靠性。

The flue gas flooding mec hanism model is complex and involves many high order coupled partial differential equations,which are difficult to be solved and optimized. This paper proposes a fuzzy neural network modeling method based on hybrid spiral optimization algorithm（HSO） and T-S model. The method combines the advantages of neural network,fuzzy system and intelligent algorithm,and improves the accuracy of the model by optimizing the weight parameters of neural network and the center and width parameters of Gaussian membership function. The identification model of flue gas flooding is established by this method,and numerical simulation results verify the accuracy and reliability of the modeling method.