基于改进的粒子群算法测井数据处理

Logging data processing based on improved particle swarm optimization

在铀矿伽马测井中,基于我国铀矿类型复杂,伽马测井所测得的数据会受到仪器以及岩性等影响,基于实际环境在测量中需要采用套管,但是套管会对测量数据产生影响,导致数据出现异常或者丢失,为解决这一问题常规处理措施是重复测井,但这一方式会导致测量成本翻倍、效率降低,文中提出使用计算机技术中的粒子群算法对铀矿伽马测井数据进行修正,降低套管对测井数据的影响。粒子群算法因其通用性强、收敛性快以及对硬件要求低等优点常被用来修正数据,但是在运行的同时,会因为陷入局部极值而导致修正结果不理想,并且在运行的后期同其他优化手段相比收敛速度慢、精度也较低。为避免此类缺点,提高修正精度,文中提出一种改善分组策略以及优化惯性权重选择方法的改进粒子群(IPSO)算法。改进方式主要针对适应度的值将所有粒子划分为优解组以及劣解组,针对优解组进行交换工作,针对劣解组进行变异操作,最后验证了优化结果,所提算法提高了异常数据的修正精度。

In gamma logging of uranium deposits in China,due to the complex types of uranium deposits,the data obtained from gamma logging may be affected by instruments and lithology.Based on the actual environment,it is necessary to use casing in the measurement.However,casing can have an impact on the measurement data,resulting in anomalies or loss of data.To solve this problem,the conventional treatment is to repeat logging,but this method can double the measurement cost and reduce efficiency,using particle swarm optimization algorithm in computer technology to modify gamma logging data of uranium mines is proposed to reduce the impact of casing on logging data.Particle swarm optimization(PSO)algorithm is often used to correct data due to its strong universality,fast convergence,and low hardware requirements.However,during operation,it may fall into local extremum,resulting in unsatisfactory correction results.In the later stage of operation,the convergence speed is slower and the accuracy is lower compared to other optimization methods.To avoid such shortcomings and improve correction accuracy,an improved particle swarm optimization algorithm(also known as IPSO algorithm)that improves grouping strategies and optimizes inertia weight selection methods is proposed.The improvement method can mainly divide all particles into superior solution group and inferior solution group according to the value of fitness.Exchange work is carried out for the superior solution group and mutation operation is carried out for the inferior solution group.The optimization results and the improvement of correction accuracy for abnormal data are verified.