在进行储层预测和评价时 ,通常使用与储层预测有关的各种地震属性。以各种方法提取的一系列地震属性包含着丰富的地质信息 ,但有些属性可能彼此相关 ,这就造成信息的重复和冗余。由此可见 ,属性的无限增加也会给储层预测带来不利的影响。针对具体问题 ,从全体地震属性中挑选出最佳的地震属性子集是非常必要的 ,此即地震属性优化问题。其目的就是从众多地震属性中挑选出与研究目标关系最密切、反应最敏感的少数属性 ,再利用优化后的地震属性进行目标层储层参数 (如孔隙率、泥质含量和储层厚度等 )反演。本文主要讨论地震属性优化的遗传算法 (GA)与 BP神经网络相结合的 GA - BP方法 ,通过对大港探区 L JF区块三维地震资料的实际应用 ,取得了良好的地质效果 Various seismic attributes related to reservoir prediction are commonly used in reservoir prediction and evaluation. A series of seismic attributes extracted by various methods contain abundant geological information, but some of the attributes may be related to each other, which results in duplication and redundancy of information. It can be seen that the infinite increase of attributes will also adversely affect reservoir prediction. In view of the specific problems, it is very necessary to select the best subset of seismic attributes from the total seismic attributes, which is the seismic attribute optimization problem. Its purpose is to pick out the few attributes that are most closely related to the research target and the most sensitive ones from the many seismic attributes and then use the optimized seismic attributes to perform the target layer reservoir parameters (such as porosity, shale content and reservoir thickness) ) Inversion. This paper mainly discusses the GA - BP method combining genetic algorithm (GA) with BP neural network which is optimized by seismic attributes. Through the practical application of 3D seismic data of LJF block in Dagang exploration area, good geological effects have been achieved