智能优化算法在概率积分参数反演中的比较

发布时间：2018-08-05 20:45

【摘要】：概率积分法是我国矿山开采沉陷预计的主要方法,其预计的精度直接取决于参数准确性。采用智能优化算法对实测地表沉陷数据反演是获取概率积分法参数的主要方法。为研究优化算法在开采沉陷概率积分参数反演中的应用效果,采用VB语言编程实现了模矢法、遗传算法、粒子群算法、模拟退火算法等常见概率积分参数反演算法,通过构造理论数据分析和比较了这4种算法参数反演的效果,并从运行时间、求参稳定性、搜索性能、抗局部解能力等方面对4种算法进行综合评价。研究结果表明:4种算法参数反演结果精度较高,参数相对误差小于2%,且对观测站中的观测值随机误差、粗差问题具有较强的抗干扰能力。模矢法运行效率高但容易陷入局部解,粒子群算法效率较低,遗传算法和退火算法全局能力强但后期收敛能力较弱。
[Abstract]:Probabilistic integration method is the main method for predicting mining subsidence in China, and its accuracy is directly dependent on the accuracy of parameters. It is the main method to obtain the parameters of probabilistic integration method by using intelligent optimization algorithm to retrieve the measured ground subsidence data. In order to study the application effect of optimization algorithm in mining subsidence probabilistic integral parameter inversion, the common probability integral parameter inversion algorithms, such as modular vector method, genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm and so on, are programmed by VB language. The effect of parameter inversion of these four algorithms is analyzed and compared by constructing theoretical data, and the four algorithms are evaluated synthetically from the aspects of running time, parameter stability, search performance, anti-local solution ability and so on. The results show that the precision of the inversion results of the four algorithms is high, the relative error of the parameters is less than 2, and the gross error problem has strong anti-interference ability to the random error of observation value in the observation station. Modular vector method has high efficiency but is easy to fall into local solution, particle swarm optimization algorithm is inefficient, genetic algorithm and annealing algorithm have strong global ability, but late convergence ability is weak.
【作者单位】：中国矿业大学环境与测绘学院;国土环境与灾害监测国家测绘地理信息局重点实验室;河南省地矿局测绘地理信息院;
【分类号】：TP18;TD327

【参考文献】