基于混合模拟退火算法的阵列侧向测井实时反演研究

测井资料现场实时反演是判断测井质量及进行精细评价的基础。为在保证反演速度的同时，进一步提高反演精度，结合模拟退火算法和马奎特算法，提出了混合模拟退火算法，用以对阵列侧向测井资料进行3参数（地层电阻率、冲洗带半径和冲洗带电阻率）混合反演。研究发现，初始值的选择对混合反演速度具有明显的影响，引入基于电阻率幅度差信息的冲洗带半径初始值选取策略，能够避免反演结果陷入局部最小值,加快反演速度。实例计算结果表明，混合反演算法的反演速度满足实时反演要求，且反演精度相较传统马奎特算法有明显提高；反演结果与试油结果一致，验证了混合反演算法的适用性。混合模拟退火算法为现场阵列侧向测井资料的反演处理提供了新的技术。 

Research on Array Lateral Logging Real-Time Inversions Based on Hybrid Simulated Annealing Algorithms

The real-time inversion of logging data is the basis for judging the field logging quality on-site and the subsequent detailed evaluations. By using the simulated annealing algorithm and Marquette algorithm, the hybrid inversion of three parameters (formation resistivity, radius of flushing zone and flushing zone resistivity) was carried out with the array lateral logging data, so as to further improve the inversion accuracy while ensuring the inversion speed of the conventional real-time inversion methods. The research showed that the selection of initial value had a clear effect on the results of hybrid inversion. A strategy for selecting the initial value of the radius of flushing zone can avoid the inversion result falling into the local minimum based on the resistivity amplitude difference. Further, it can speed up the inversion. The example application results indicated that the inversion speed of hybrid inversion algorithm could meet the real-time inversion requirements. The research also found that the inversion accuracy was higher than that of the traditional Marquette algorithm significantly. Remarkably, the inversion results were consistent with the test results, which verified the applicability of hybrid inversion algorithm. The hybrid simulated annealing algorithm has provided new idea and technical support for the inversion processing of field array lateral logging data.