基于近似支持向量机的裂缝分类方法

对广泛存在于各类岩层中的裂缝带进行精细刻画与综合预测是裂缝型油气藏勘探的关键。为了避免多解性问题，学者们通常采用多属性对其进行综合预测，但如何有效地利用众多地震属性与裂缝带发育程度之间的非线性关系对裂缝带发育状况进行准确分类仍是一大难题。将近似支持向量机算法引入裂缝带的分类识别中，建立了3种刻画储层裂缝带的地震属性与井中裂缝发育信息之间的非线性模型，得出了反映裂缝带特征的最佳判别规则，利用该规则对多个属性进行综合判别，克服了单属性的多解性，提高了储层裂缝带的分类精度。实例应用表明，该算法削弱了依靠单一因素识别储层裂缝带的局限性，为储层内裂缝带发育状况的准确分类提供了新的研究思路。

Fracture classification method based on proximal support vector machine

Fractured oil and gas reservoirs are widely found in all kinds of rock formations,how to make fine characterization and comprehensive prediction of fracture zone is the key to the exploration of fractured oil and gas reservoirs.In order to avoid the problem of multiple solutions,many scholars usually use multi-attribute to predict it synthetically.However,there are many complex nonlinear relationships between the development degree of cracks and some seismic attributes,the effective use of the correspondences for nonlinear prediction is also a difficult problem in the comprehensive prediction of fracture zones.The proximal support vector machine algorithm was introduced into the classification of fracture zone,and the nonlinear model between three kinds of seismic properties depicting reservoir fracture zone and crack development information in well were established,and the best discriminating rule reflecting the characteristics of fracture zone was obtained.Based on the rule,multiple seismic attributes can be comprehensively discriminated,the multi-solution of single attribute was overcome,and the classification accuracy of reservoir fracture zone was improved.Example showed that the algorithm weakened the limitation of identifying reservoir fracture zones by single factor classification,and it can provide a new research idea for the accurate classification of the development of fracture zones in rock formations.