基于遗传神经网络的储层敏感性预测方法研究

储层敏感性是储层与外来流体发生各种物化作用,使储层孔隙结构和渗透性发生变化的特性,这种变化会不同程度地损害油层,从而导致产能损失或产量下降.如果能在施工之前对储层的敏感性做出预测,那么在进行施工的过程中,就可以采取相应的措施,减少储层的损害.在储层敏感性进行预测的方法中,BP神经网络是应用最广泛的方法之一,可以对储层的各种敏感性进行预测.但这也存在着一些问题,比如局部寻优、收敛速度慢等,所以在神经网络的基础上,加入了遗传算法,可以对神经网络进行优化,使其能最大程度地快速找到全局最优.实践证明,这种方法能够满足目前储层敏感性预测的实际需求.

PREDICTION AND RESEARCH ON FORMATION SENSITIVITY BASED ON GENETIC ALGORITHM AND NEURAL NETWORK MODEL

Formation sensitivity is referred to the characteristic that pore structure and permeability will be changed when many physicochemical interactions between the reservoir and fluid happen.This change will cause damage to reservoir in different degree with capacity loss or the fall in output.If formation sensitivity can be predict before the engineering,corresponding measures will be taken to reduce the damage to the reservoir.Among the methods of the prediction on formation sensitivity,BP neural network is one of the method which is applied most extensively.It can predict various sensitivities,but it remained some questions such as local-optimization and poor convergence and so on.So on the basis of neural network,genetic algorithm is added to optimize neural network,and the global minimizer could be found quickly to the highest degree.Practice has shown that this method can meet the actual need of prediction on formation sensitivity nowadays.