基于BP神经网络方法的分层注水量计算

对静态地层系数法和动态方程法等分层注水量计算方法分析表明:现有方法考虑注水量影响因素较少,计算误差较大,适用性较差。应用BP神经网络方法计算分层注量,以砂岩厚度、有效厚度、渗透率和沉积相影响系数等16个影响因素作为模型输入参数,单层吸水量作为模型输出。实例计算结果表明:BP神经网络法计算分层注水量与实测值的最大误差为6.51%,平均误差为3.21%,准确性较好,说明BP神经网络方法在分层注水量计算方面具有较好的应用前景。

The calculation of separated layer water-injected volume based on BP neural network

Through the analysis of the method of separated layer water-injected volume of static state formation capacity and dynamic equation,it noted that:the current method seldom considering affecting factor of injection flow rate is less,the calculation error is bigger,and the applicability is poorer.The separated layer water-injected volume by using BP neural network was calculated and defined sand thickness,effective thickness,permeability,sedimentary facies influence coefficient and other 16 factors as model input parameters,individual layer water-injected volume was defined as output parameter.Through the calculated results of example showed that: the calculation result error in measured value returned by maxim of 6.51%,and by an average of 3.21%,the calculated result compared with actual injection was accurately.So the method of BP neural network used in splitting calculation of water-injected volume could have excellent application prospect.