单纯形一模拟退火混合算法反求水文地质参数及其并行求解

本文利用单纯形法局部搜索速度快和模拟退火算法全局寻优的特点，同时为了克服各自算法的弊病，提出采用单纯形一模拟退火混合算法（SMSA）进行水文地质逆问题的求解。论文详细描述了SMSA算法的具体操作算子的实现，并将该算法应用于一个大型水源地的地下水模拟反演。计算结果表明，SMSA算法在水文地质参数反演计算具有求解速度快，精度高的特点，而且易于实现并行运算。     

应用单纯形-模拟退火混合算法估计河流水质参数

将由单纯形法(SM)与模拟退火(SA)两种算法构成的混合算法(SMSA),应用于求解分析瞬时投放示踪剂情况下的一维河流水团示踪试验数据,估计河流水质参数的函数优化问题。分别就不同的降温指数、内循环次数与新状态产生函数中的扰动系数等算法控制参数对混合算法和改进模拟退火法的收敛速度的影响,进行了数值实验。结果表明,SMSA混合优化算法对于求解估计河流水质参数的函数优化问题是非常有效的。在实验条件下,与单一具有记忆功能的改进模拟退火法的水质参数计算结果相比较,SMSA混合算法具有:①混合算法的收敛速度明显优于改进模拟退火法;②降温指数和内循环次数对SMSA混合算法的收敛速度影响非常微弱;③新状态产生函数中的随机扰动幅度大小对算法收敛速度具有较为明显的影响等特点。     

应用改进并行遗传算法反求水文地质参数研究

利用单纯形法局部搜索速度快和遗传算法全局寻优的特点,同时为了克服两种方法各自的弊病,提出采用混合单纯形技术的并行遗传算法(Hybrid Simplex Parallel Genetic Algorithm,SPGA)进行水文地质逆问题的求解。详细阐述了SPGA算法的具体操作和实现,并将该方法应用于水源地的地下水模拟反演中。计算结果表明,SPGA算法在水文地质参数寻优计算中具有比较好的可靠性和计算效率。     

水文地质参数反演的Hooke-Jeeves粒子群混合算法

水文地质参数寻优结果的好坏会直接影响到地下水数值模拟的精度,而参数寻优结果很大程度上取决于寻优方法的选择。粒子群算法是一种基于群智能的随机全局寻优方法,算法的缺陷是后期搜索效率低劣。基于随机寻优算法的混合策略,引入有效的约束处理手段和粒子群算法惯性因子的动态非线性调整技术,有机融合粒子群算法与Hooke-Jeeves方法,提出一种适用于水文地质参数反演的HJPSO混合算法。应用研究表明,HJPSO混合算法在参数反演计算中求解精度高、收敛速度快、寻优性能强,是一种值得推广的水文地质参数识别方法。     

水文地质参数识别的连续禁忌搜索改进算法

水文地质参数的正确与否是构建地下水数值模型的根本,而参数寻优结果很大程度上取决于优化算法的选择。禁忌搜索算法是一种广泛应用于组合优化问题的启发式全局寻优算法,但在连续函数优化领域应用比较少。基于上述考虑,本文首先引入求解连续函数优化问题的连续禁忌搜索算法并对其进行改进,进而提出一种连续禁忌搜索改进算法(ICTS),最后将其与地下水模型耦合进行水文地质参数识别。算例研究表明,ICTS算法较其他算法(CTS,SGA,Micro-GA,PSO)求解效率提高1.87～4.64倍,求解精度提高1.08～12.86倍。因此ICTS算法在参数反演计算中求解精度高、收敛速度快、寻优性能强,是一种值得推广的水文地质参数识别方法。     

模拟退火算法反演水文地质参数算例研究

基于模拟退火算法(SA)这一全局优化技术,耦合地下水模拟的有限元模型,给出水文地质参数的反演方法.并用水文地质参数反演算例来验证反演模型的可用性和适用性.结果表明,用"变尺度"产生新的渗透系数反演结果明显优于"随机法",当初始温度在500～650、降温系数在0.92～O.95时,模拟退火法在正常情况获得了很好的反演结果,在考虑减少观测孔个数和观测水头值有误差的情况下,也取得了较为满意的反演结果.为了提高运算效率,可考虑与传统反演方法或其它人工智能算法相结合.     

大地电磁测深拟二维模拟退火约束反演

大地电磁测深(MT)资料解释的核心是反演问题,其反演方法可分为两大类:局部方法和全局方法。局部方法基于最小方差原理,依赖于目标函数的导数求解,因而其解依赖于初始模型,容易陷入局部极值,难以得到全局最优解。全局方法如模拟退火算法,它是近年发展起来的全局最优化算法。其主要优点是:不用求目标函数的偏导数及解大型矩阵方程组,即能找到一个全局最优解,而且易于加入约束条件,方法易于移植。目前模拟退火算法已开始用于解决非线性地球物理反演问题,但还存在计算效率低的缺陷。以快速模拟退火算法为基础,对大地电磁测深理论数据及实际资料进行快速模拟退火约束多参量的反演,效果较为理想。对H型地电断面的理论曲线反演和根据某盆地预探井实际资料对中生界厚度及基底埋深的反演作了举例说明。     

复形法在反求水文地质参数中的应用

在众多的水文地质调参方法中,目前,经常使用的是单纯形法。此法虽很有效,但有一较大缺点,即在迭代过程中易出退化现象,从而使迭代难以继续,给计算造成很大的麻烦。复形法却因其顶点数增多,可使退化现象得到改善。在大水矿区水文地质参数反演计算中,尝试利用了复形法。结果表明:复形法稳定性能好,收敛速度快,求解精度高,节省时间,使用方便,所求参数基本上能代表所求区水文地质条件。复形法可求解高维(一般 ≤ 30维)有约束的非线性函数极值问题,它是求解多参数的水文地质反演问题的有效方法之一,适应于水文条件复杂、参数分区多的情况。      

围岩力学参数反演的GSA-BP神经网络模型及应用

针对遗传算法（GA）存在早熟现象和局部寻优能力较差等缺陷,引入具有很强局部搜索能力的模拟退火算法（SA）,组成改进的遗传模拟退火算法（GSA）提高优化问题的能力和求解质量。针对BP神经网络容易陷入局部最小和收敛速度慢等方面的不足,应用改进的遗传模拟退火算法搜索BP神经网络的最优权值和阀值,提高BP神经网络的预测精度,建立了围岩力学参数反分析的GSA-BP神经网络模型。将该模型应用于乌东德水电站右岸地下厂房围岩力学参数的反演分析中,根据监测围岩变形数据反演围岩力学参数,反演所得参数应用到正计算分析中,得出的计算位移与实测值吻合较好,说明该方法的有效性和应用于该工程的可行性。     

基于自适应纯形模拟退火法一维大地电磁测深视电阻率和相位反演研究

针对大地电磁测深反演中线性化方法容易陷入局部极值,而全局优化方法收敛慢等问题,这里采用自适应纯形模拟退火综合优化方法进行大地电磁测深数据反演。该优化方法综合了下降纯形法和模拟退火法各自的优点,已被证明具有全局搜所能力和收敛速度快的特点,并且实现了视电阻率和相位同时反演,减小多解性的同时提高了反演的分辨率。通过H、K和HKH型模型的数值计算,验证了这种综合优化方法的搜索效率和全局收敛性;所有模型合成数据的反演结果都能较好地反映真实模型的结构特征。对于中间层为相对高阻的模型,虽然相对其他层结构恢复差些,但是反演后该层结构基本得到恢复。     

Determining the uniqueness of best-fit trishear models

We show the application of a simulated annealing algorithm to trishear inverse modeling. The algorithm traverses the parameter space in search for best-fit models without being trapped in local minima, and thus sampling for more possible solutions globally. Simulated annealing is a robust and efficient technique to determine the uniqueness of best-fit trishear models; the spread of possible trishear models that can fit a structure. We first apply the algorithm to a decameter-size, contractional fault-propagation fold in west-central Taiwan, for which there is an exceptional exposure of pre-growth and growth strata. Simulated annealing shows that even for this complete fold dataset with low uncertainties, there is a range of models and fault slip/uplift histories that can fit the data, with the consequent implications for the assessment of seismic hazard. We then apply the algorithm to a kilometer-size, extensional fault-propagation fold, the Hadahid monocline, Gulf of Suez Rift, Egypt. In this monocline there is only surface coverage in the footwall anticline areas and the algorithm was used to delimit the range of possible models that can fit the data and their uncertainties, thus avoiding biases in the interpretation. Simulated annealing suggests that the along-strike structural variability of the monocline can result from along-strike variability in fault slip, fault propagation to fault slip ratio and depth of fault nucleation. Both examples illustrate the benefits of searching for a possible range of models rather than a precise best-fit model when modeling fault-propagation folds. In an attempt to understand which parameters control fault development, and also how the spread of possible solutions varies with fold growth, we apply the algorithm to four sequential stages of a published, analog clay model of an extensional forced fold. The inversions of the natural examples and the analog model suggest that the spread of the possible models is a manifestation of the data uncertainties, the suitability of the trishear model, fold evolution, and rock mechanical properties.     

基于模拟退火的Gauss-Newton算法的神经网络在渗流反分析中的应用

提出了模拟退火的Gauss-Newton算法的神经网络,克服了经典BP网络存在的一些缺陷。并以正弦函数的迭代收敛为例,证明了该方法的正确性,有效性和优越性。同时将该方法用于同乐坪大坝的渗流反分析,利用反演出的渗透系数进行渗流场计算。得到的水头预报值与观测值相吻合,可知反演结果是正确的,说明该方法用于实践工程的渗流参数识别是可行的。     

Applicability of the most frequent value method in groundwater modeling

The Most Frequent Value Method (MFV) is applied to groundwater modeling as a robust and effective geostatistical method. The Most Frequent Value method is theoretically derived from the minimization of the information loss called the I-divergence. The MFV algorithm is then coupled with global optimization (Very Fast Simulated Annealing) to provide a powerful method for solving the inverse problems in groundwater modeling. The advantages and applicability of this new approach are illustrated by means of theoretical investigations and case studies. It is demonstrated that the MFV method has certain advantages over the conventional statistical methods derived from the maximum likelihood principle.

---

Resumen El Método del Valor Mas Frecuente (VMF), es aplicado al modelamiento de agua subterránea, como un método geoestadístico simple y efectivo. Este método es derivado teóricamente de la acción de reducir al mínimo la pérdida de información, llamada así divergencia – I. El algoritmo del VMF es entonces acoplado con optimización global(Very Fast Simulated Annealing), para obtener así un método efectivo que resuelva los problemas inversos en el modelamiento de aguas subterráneas. Las ventajas y aplicabilidad de esta aproximación nueva son ilustradas a través de investigaciones teóricas y estudios de caso. Se demuestra que el método VMF tiene ciertas ventajas sobre los métodos estadísticos convencionales derivados del principio de la probabilidad máxima.

---

Résumé On a appliqué la méthode de la valeur la plus fréquente (VPF) comme une méthode géostatistique robuste et efficace pour modéliser les eaux souterraines. Du point de vue théorique, la méthode de VPF part de la minimisation de linformation perdue, dénommée I-divergence. On couple après lalgorithme de la méthode de VPF avec la méthode doptimisation globale affin de réaliser une méthode performante pour résoudre le problème inverse dans le domaine des eaux souterraine. Les avantages et les possibilités dapplication de cette nouvelle approche sont illustrées par des investigations théoriques, ainsi que par des études de cas. On montre que la méthode de VPF présente certains avantages par rapport des méthodes statistiques conventionnelles basées sur le principe de la probabilité maximale.

     

History matching on the Imperial College fault model using parallel tempering

The history-matching inverse problem from petroleum engineering is analysed using the Imperial College fault model. This fault model produces a challenging inverse problem and is designed to show some of the problems which can occur whilst performing history-matching calculations on complicated geologies. It is shown that there can be multiple distinct geologies which match the history data. Furthermore, it is shown that the maximum-a-posteriori estimate does not correspond to the true geology in some cases. Both of these statements are corroborated via numerical examples where the parameter spaces are ?, ?³, ?⁷ and ?¹³. In addition, it is shown that the number of matches which agree with the data increases with dimension for these examples. It is also shown that the different matches can result in different reservoir management decision which, if incorrectly taken, would incur substantial financial penalties. All of these analyses are performed in a systematic manner, where it is shown that the standard algorithms can give a misleading answer. The history-matching problem is written as a minimisation problem, and it is shown that knowledge of all of the local minima is required. This presents significant computational issues as the resulting objective function is highly nonlinear, expensive to evaluate and multimodal. Previously used algorithms have been proved to be inadequate. Parallel tempering is a method which, if run for long enough, can find all the local minima. However, as the objective is expensive, a number of algorithm modifications had to be used to ensure convergence within a reasonable time. This new information is outlined in the paper. The algorithm as implemented produced results and new insights into this problem which were not suspected before. The results produced by this algorithm for the multimodal history-matching problem are superior to all other results of which we are aware. However, a considered amount of computation time was used within this paper, so this result does not infer that the algorithm cannot be improved upon. This algorithm not only produces good results but can be applied to all other history-matching problems. We have shown that this method provides a robust route of finding multiple local optima/solutions to the inverse problem, which is of considerable benefit to the petroleum industry. Furthermore, it is an entirely parallel algorithm which is becoming computationally feasible for other history-matching problems.     

Conditional Simulation of Multi-Type Non Stationary Markov Object Models Respecting Specified Proportions

Although Boolean model simulation has been widely used during the last two decades to simulate sedimentary bodies (especially in fluvio-deltaic environments), a key issue has subsisted. One of the most important parameter for object model simulation, namely the (non stationary) intensity of the underlying object process, is not a parameter provided by the end user but must instead be computed from other input parameters, such as local proportions of lithofacies, erosion rule between objects of different types and interaction between objects. This paper revisits a birth and death algorithm for simulating conditional, non stationary, multi-type objects models with interaction. It provides workable approximations for computing the local intensity of the underlying point process to respect proportion maps. Simulated examples show that this algorithm is able to reproduce the desired proportions. Important issues for implementing this algorithm are discussed.     

改进的遗传算法及其在渗流参数反演中的应用

利用水头实测资料,以裂隙组的渗透系数比例因子为待反演的参数向量,在采用基本遗传算法进行参数反演研究的基础上,针对裂隙岩体无压渗流参数反问题计算量过大,目标参数众多以及参数可能变化范围大等特点,提出了一种混合遗传算法求解此类问题,力求克服简单遗传算法在解决此类问题时存在的局部搜索能力弱、易出现早熟收敛及计算量大等缺陷,并通过典型岩坡渗流算例进行验证,同时给出了基本遗传算法、传统单纯形算法的反演成果。计算结果表明,该方法保持了基本遗传算法优点,并有效地提高了算法的运行效率,从而为求解裂隙岩体无压渗流参数反问题等计算量大的系列问题提供了新的途径。     

基于遗传算法的岩土力学参数反演及其在 ABAQUS中的实现

岩土工程优化反分析是一个典型的复杂非线性函数优化问题,采用全局优化算法是解决这个问题的理想途径。结合ABAQUS有限元软件,提出遗传算法与有限元联合反演法,将有限元程序作为一个单独模块嵌入到遗传算法程序中,以测点的实测值与计算值建立误差函数,编制了遗传算法反演分析程序。并给出应用实例验证了该法的有效性,表明该方法可应用于岩土工程中的反演分析工作。     

Spectral Simulation and Conditioning to Local Continuity Trends in Geologic Modeling

Spectral simulation has gained application in building geologic models due to the advantage of better honoring the spatial continuity of petrophysical properties, such as reservoir porosity and shale volume. Distinct from sequential simulation methods, spectral simulation is a global algorithm in the sense that a global density spectrum is calculated once and the inverse Fourier transform is performed on the Fourier coefficient also only once to generate a simulation realization. The generated realizations honor the spatial continuity structure globally over the whole field instead of only within a search neighborhood, as with sequential simulation algorithms. However, the disadvantage of global spectral simulation is that it traditionally cannot account for the local information such as the local continuity trends, which are often observed in reservoirs and hence are important to be accounted for in geologic models. This disadvantage has limited wider application of spectral simulation in building geologic models. In this paper, we present ways of conditioning geologic models to the relevant local information. To account for the local continuity trends, we first scale different frequency components of the original model with local-amplitude spectrum ratios that are specific to the local trend. The sum of these scaled frequency components renders a new model that displays the desired local continuity trend. The implementation details of this new method are discussed and examples are provided to illustrate the algorithm.