| 基于量子行为粒子群算法的航空瞬变电磁拟二维反演技术 |
| |
| 引用本文： | 何一鸣,,,薛国强,,,,赵炀,,.基于量子行为粒子群算法的航空瞬变电磁拟二维反演技术[J].延边大学理工学报,2020,0(6):722-730. |
| |
| 作者姓名： | 何一鸣 薛国强 赵炀 |
| |
| 作者单位： | 1. 中国科学院地质与地球物理研究所 中国科学院矿产资源研究重点实验室,北京 100029; 2. 中国科学院
地球科学研究院,北京 100029; 3. 中国科学院大学 地球与行星科学学院,北京 100049; 4. 长安大学
地球物理场多参数综合模拟实验室(中国地球物理学会重点实验室),陕西 西安 710054 |
| |
| 摘 要： | 传统的确定性反演算法严重依赖初始模型,易陷入局部极小值中,导致最终反演结果偏离真实模型。粒子群(PSO)算法作为一种随机性反演算法,具有较强的跳出局部极小值的能力,但是仍存在早熟收敛和收敛速度慢等问题,限制了该算法在二、三维电磁反演中的发展。针对上述问题,首先提出采用量子行为粒子群(QPSO)算法代替传统粒子群算法,将量子在势阱中运动规律引入到粒子群算法中,使得粒子可以出现在势阱内任何存在概率分布的位置上,有效地克服了由于群体的聚集性所导致的早熟收敛问题。此外,采用拟二维反演算法代替传统二维反演算法,使得反演模型参数维度下降,寻优过程中局部极小值个数将大幅度减少,显著提高粒子群算法的收敛速度,但是在粒子群中开展拟二维反演时,传统的正则化参数的寻优过程将浪费大量计算资源。结合量子行为粒子群算法中各测点的全局最优粒子在粒子群进化过程中的重要地位,采用α-Trimmed方法开展相邻点间全局最优粒子模型参数光滑约束,实现粒子群算法快速横向约束反演。最后将量子行为粒子群算法拟二维反演技术应用到含噪全航空瞬变电磁仿真数据处理中,反演结果与原始模型具有较好的一致性。
|
| 关 键 词： | 航空瞬变电磁法 粒子群算法 群理论 量子理论 拟二维反演 α-Trimmed方法 Occam反演 聚焦反演 |
Quasi-2D Stochastic Inversion of Airbone Transient Eletromagnetic Data
Based on Quantum-behaved Particle Swarm Optimization Algorithm |
| |
| HE Yi-ming,,' target="_blank" rel="external">,XUE Guo-qiang,,,' target="_blank" rel="external">,ZHAO Yang,,' target="_blank" rel="external">.Quasi-2D Stochastic Inversion of Airbone Transient Eletromagnetic Data
Based on Quantum-behaved Particle Swarm Optimization Algorithm[J].Journal of Yanbian University (Natural Science),2020,0(6):722-730. |
| |
| Authors: | HE Yi-ming " target="_blank">' target="_blank" rel="external"> XUE Guo-qiang " target="_blank">' target="_blank" rel="external"> ZHAO Yang " target="_blank">' target="_blank" rel="external"> |
| |
| Affiliation: | 1. Key Laboratory of Mineral Resources, Institute of Geology and Geophysics, Chinese Academy of Sciences,
Beijing 100029, China; 2. Innovation Academy of Earth Science, Chinese Academy of Sciences, Beijing 100029,
China; 3. College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049,
China; 4. Integrated Geophysical Simulation Lab(Key Laboratory of Chinese Geophysical Society), Chang'an
University, Xi'an 710054, Shaanxi, China |
| |
| Abstract: | The traditional deterministic inversion algorithm relies heavily on the initial model, which is easy to fall into the local minimum and leads to the final inversion results seriously deviate from the real model. As a relatively new stochastic inversion algorithm, particle swarm optimization(PSO)algorithm has a strong ability to jump out of local minimum. However, there are still some problems such as premature convergence and slow convergence speed, which limit the development of algorithm for 2- or 3-dimensional electromagnetic inversion. In view of the above problems, quantum-behaved particle swarm optimization was proposed to replace the traditional particle swarm optimization, and the law of quantum motion in the potential well was introduced into the particle swarm optimization algorithm, so that the particles could appear in any position with probability distribution in the potential well, which could effectively overcome the premature convergence problem caused by the aggregation of population. In addition, the quasi-2D inversion algorithm was used to replace the traditional 2-dimensional inversion algorithm. As the parameter dimension of inversion model decreases, the number of local minima was greatly reduced, which significantly improves the convergence speed of particle swarm optimization. However, if the traditional regularization process is still used in the quasi-2D inversion of particle swarm optimization, the optimization of regularization parameters will waste a lot of computing resources. Combining with the important position of global optimal particles at each measuring point in quantum-behaved particle swarm optimization in the evolution of particle swarm, the α-Trimmed method was introduced to realize fast lateral constraint inversion by parameter smoothing constraint of global optimal particle model between adjacent points. Finally, the quasi-2D inversion technique based on quantum-behaved particle swarm optimization algorithm was applied to the simulation data of airborne transient electromagnetic method with noise, and the final inversion results were in good agreement with the forward model. |
| |
| Keywords: | airborne transient eletromagnetic method particle swarm optimization algorithm group theory quantum theory quasi-2D inversion α-Trimmed method Occam inversion focusing inversion |
|
| 点击此处可从《延边大学理工学报》浏览原始摘要信息 |
|
点击此处可从《延边大学理工学报》下载全文 |
|
