基于快速、高精度遗传算法神经网络的薄互储层参数预测

Thin interbedded reservoir parameters predicting based on high speed and precise genetic algorithm neural network

针对传统遗传算法 (GA)和人工神经网络 BP算法各自存在的不足 ,引入自适应机制的浮点数编码的遗传算法 ,并将其与 BP网中的梯度下降法相结合 ,进行混合交互运算 ,形成 GA- BP混合算法。该算法使网络具有较快的收敛速度和较高的逼近精度 ,能较好地解决综合多种地震信息进行薄互储层参数预测的精度和收敛速度问题 ,并通过实例验证了此方法的正确性和实用性。

Ageneticalgorithmwithadaptiveandfloating pointcodeis proposed to overcome disadvantages of the genetic algorithm and BP algorithm. This algorithm is combined with BP to give GA BP mixed algorithm which has higher accuracy and faster convergence speed. The new algorithm also provides improved predict accuracy of thin interbeded reservoir parameters. An example shows the validity and feasibility of this algorithm.