基于粒子群优化最小二乘支持向量机的裂缝及缝洞充填物识别

PSO-LSSVM-BASED IDENTIFYING METHOD FOR THE FRACTURE-VUG FILLINGS

碳酸盐岩储层中发育的缝洞是油气的主要储集空间和渗流通道,缝洞充填物影响着储层的渗流能力和储集能力,碳酸盐岩储层中裂缝及缝洞充填物的识别对于此类储层的勘探开发具有重要意义。基于常规测井资料对裂缝和缝洞充填物的响应特性,提出综合应用粒子群算法(PSO)和最小二乘支持向量机(LSSVM)的裂缝及缝洞充填物识别方法 (PSO-LSSVM),并将该方法和BP神经网络分别应用于滨里海盆地东缘石炭系碳酸盐岩储层裂缝及缝洞充填物的识别。对比分析两种方法的识别结果,PSO-LSSVM的识别效果比BP神经网络好。利用PSO-LSSVM方法得到的识别结果与FMI电成像测井图像及岩心资料得到的结果有较好的一致性。

Fracture-vugs seepage channels for the well-developed in carbonate reservoirs are the main part of the accumulating space and oil and gas, but the fillings in the fracture-vugs influence the storage and seepage capacities of the reservoirs, so the identifications of the fractures and development of this kind of the logging data to the fractures and fillings play an important roles in the exploration and carbonate reservoir. Based on the response characteristic of the conventional well fillings , the PSO-LSSVM identifying method was presented for the fractures and fillings by integrating the Particle Swarm Optimization （ PSO ） and the Least Squares Support Vector Machines （ LSSVM ）. And moreover the method and BP neural network method were applied to identify the fractures and fill- ings of the Carboniferous carbonate reservoirs in the eastern margin of Pre-Caspian Basin. The comparison of the identified results of these two methods shows that the PSO-LSSVM method is better than the other. And furthermore the result of the PSO-LSSVM method is well coincident with those of FMI electrical images and core data.