基于高性能计算的SWAT参数敏感度分析并行框架

随着大规模水文模拟需求的不断提高,如何解决计算需求问题逐渐成为水文研究的一个热点.SWAT(soil and water assessment tool)模型在进行大规模水文模拟时有着良好的适应性与准确度,但其敏感度分析模块由于计算量过高,计算时长往往长达数月之久.为了加快SWAT敏感度分析的运行速度,针对SWAT敏感度分析模块的特点,基于MPI提出了一种高效的主—从式并行计算框架,并在此框架的基础上,通过将正演过程并行化,在敏感度分析的主—从并行框架中引入通信子空间的操作,将并行化的正演与主—从式的外层并行框架相结合,得到一种混合式的敏感度分析并行框架,大大提高了对参数集合的敏感度分析速度,将SWAT敏感度分析模块使用的处理器数量从原始的单核串行一跃提升到百核的数量级.最后通过天山北坡流域的模拟验证了此并行框架的可行性.     

地理国情普查数据在SWAT水文模拟中的适用性研究

针对以往SWAT水文模拟应用的精度较低的土地利用分类数据,论文研究了将高精度地理国情普查数据导入到SWAT模型土地利用分类系统的整合方法,以鹤壁汤河流域为研究区,构建了SWAT水文模型,以Landsat影像解译的土地利用/覆被数据作为参考,评价了地理国情普查数据在SWAT流域水文模型模拟中的适用性。模拟结果表明,地理国情数据模拟结果率定期和验证期相关系数(0.74、0.82)优于Landsat影像解译的土地利用/覆被数据模拟结果(0.71、0.80);基于地理国情数据构建SWAT模型模拟的月径流值与实测径流值吻合较好。     

随机漂移粒子群算法的RZWQM替代模型参数优化

根区水质量模型（Root Zone Water Quality Model,RZWQM）被广泛应用于刻画土壤水文循环过程对作物生长的影响,并通过模型率定模拟指导农业生产管理。然而RZWQM模型的一次率定需要较长时间,在可接受时间范围内找到一组合适的模型参数是一件较困难的工作;同时传统的模型参数试错法依赖于使用者的专业知识和经验,也需要多次尝试才能达到较满意的模拟效果。提出使用稀疏网格方法建立RZWQM模型的近似替代模型,并使用随机漂移粒子群优化算法对替代模型进行自动参数优化,将优化后的参数用于RZWQM模型的实际应用模拟。替代模型近似精度高,率定速度快,大大节省了模型参数优化的计算开销。最后将提出的稀疏网格近似替代模型方法结合随机漂移粒子群优化算法使用美国爱荷华州5年玉米-大豆间中的作物产量、排水流量、[NO-3]-N流失量田间实测数据进行了验证分析。结果显示该方法能够极大地提高模型参数优化效率和节省人力;同时,通过模型性能评价指标PBIAS、NSE和RSR的数值比较也表明该方法优化后的RZWQM模型性能要好于传统试错法的模型性能。     

CASC2D分布式水文模型异构并行算法研究

针对 CASC2D 模型精细化水文模拟时面临的计算耗时长、效率低等问题,在保持产汇流算法和流域拓扑结构的基础上,采用 CPU+GPU 的异构并行算法对 CASC2D 模型程序进行重新设计和优化,模型程序中的降雨、 产流、坡面汇流和河道汇流过程均优化为并行计算,以提高 CASC2D 模型的计算效率。将优化后的 CASC2D 模型应用于前毛庄流域的洪水流量过程模拟,计算结果与原 CASC2D 模型保持一致。在栅格分辨率为 30 m,计算步长为 3 s 时,与原 CPU 串行计算方法相比,并行加速比达到 34 倍以上,并且栅格单元数据精度越高,加速比提升越明显。异构并行算法可在不降低模拟精度的条件下显著提升 CASC2D 模型的计算效率,满足实时水文预报的应用需求。     

基于粒子群优化算法的滤波器优化

给出一种基于粒子群优化算法（PSO）的模拟滤波器优化设计方法。传统的模拟滤波器的精度与效率均较差,引入PSO算法可对滤波器参数进行寻优。将滤波器的设计问题转化为滤波器参数的优化问题,然后利用粒子群优化算法对整个参数空间进行高效搜索以获得最优解;通过变异、重新随机化及采用自适应的惯性权重,提高了算法的搜索效率及收敛性。实例计算表明了算法在该类问题中的有效性和可行性。     

基于粒子群算法的PID参数优化

粒子群优化PSO算法是近几年出现的一种新型演化算法,对连续函数的优化效果良好。通过采用PSO算法对PID参数进行了优化,使用实数编码方法,用局部版粒子群算法取得了良好的优化结果。说明了粒子群算法寻优简单、鲁棒性强、易于并行化,是一种效率很高的寻优方法,是PID参数优化的理想方法。     

一种新型的过程模型参数辨识方法

针对模型参数辨识问题,提出了一种基于菌群优化（BSFO）算法的模型参数辨识方法。通过将辨识参数设置为群体细菌在参数空闸的位置,并模拟细菌群体觅食的动态行为来实现对参数的寻优,有效地提高了参数辨识的精腰和效率。对火电厂热工过程参数辨识的仿真研究验证了本文算法的有效性,结果表明,菌群优化算法能够实现对过程模型参数的有效辨识,仿真结果令人满意。     

一种新型的非线性系统模型参数辨识方法

针对传统模型参数辨识方法和遗传算法用于模型参数辨识时的缺点,提出了一种基于微粒群优化(PSO)算法的模型参数辨识方法,利用PSO算法强大的优化能力,通过对算法的改进,将过程模型的每个参数作为微粒群体中的一个微粒,利用微粒群体在参数空间进行高效并行的搜索来获得过程模型的最佳参数值,可有效提高参数辨识的精度和效率.     

引力搜索优化ELM的企业财务危机预警方法

为提高企业财务危机的预测准确率,提出一种基于引力搜索算法优化核极限学习机(KELM)的并行模型PHGSA-KELM。模型考虑了特征选择机制和参数优化两者对KELM模型起着同等重要的作用,提出改进的引力搜索算法(HGSA)同步实现特征选择机制和KELM参数优化,同时设计的线性加权多目标函数综合考虑了分类精度和特征子集数量,改善了算法的分类性能,并且基于多核平台的多线程并行方式进一步提高了算法的计算效率。通过真实数据集的实验结果表明,提出的模型不仅获得了较少的特征子集个数,找出了与企业财务危机紧密相关的特征,得到了很高的分类准确率,并且计算效率也得到极大提高,是一种有效的企业财务危机预警模型。     

基于遗传算法的PID参数整定与优化

PID参数整定与优化一直是自动控制领域研究的重要问题。采用遗传算法进行PID参数整定与优化是一种寻求全局最优且与初始条件无关的优化方法。在参数整定与优化过程中,考虑了过程控制系统的参数整定特点和寻优精度。     

Design and implementation of a general software library for using NSGA-II with SWAT for multi-objective model calibration

Calibrating watershed-scale hydrologic models remains a critical but challenging step in the modeling process. The Soil and Water Assessment Tool (SWAT) is one example of a widely used watershed-scale hydrologic model that requires calibration. The calibration algorithms currently available to SWAT modelers through freely available and open source software, however, are limited and do not include many multi-objective genetic algorithms (MOGAs). The Non-Dominated Sorting Genetic Algorithm II (NSGA-II) has been shown to be an effective and efficient MOGA calibration algorithm for a wide variety of applications including for SWAT model calibration. Therefore, the objective of this study was to create an open source software library for multi-objective calibration of SWAT models using NSGA-II. The design and implementation of the library are presented, followed by a demonstration of the library through a test case for the Upper Neuse Watershed in North Carolina, USA using six objective functions in the model calibration.     

Multi-objective model auto-calibration and reduced parameterization: Exploiting gradient-based optimization tool for a hydrologic model

Multi-objective model optimization methods have been extensively studied based on evolutionary algorithms, but less on gradient-based algorithms. This study demonstrates a framework for multi-objective model calibration/optimization using gradient-based optimization tools. Model-independent software Parameter ESTimation (PEST) was used to auto-calibrate ISWAT, a modified version of the distributed hydrologic model Soil and Water Assessment Tool (SWAT2005), in the Shenandoah River watershed. The time-series processor TSPROC was used to combine multiple objectives into the auto-calibration process. Two sets of roughness coefficients for main channels, one assigned and calibrated according on soil types and one determined via empirical equations, were examined for stream discharge simulation. Five different weighting alternatives were investigated for their effects on ISWAT calibrations. Results showed that using Manning's roughness coefficients obtained from empirical equations improves simulation results and calibration efficiency. Applying a two-step weighting alternative to different observation groups would provide the best calibration results.     

Moving SWAT model calibration and uncertainty analysis to an enterprise Hadoop-based cloud

With enhanced availability of high spatial resolution data, hydrologic models such as the Soil and Water Assessment Tool (SWAT) are increasingly used to investigate effects of management activities and climate change on water availability and quality. The advantages come at a price of greater computational demand and run time. This becomes challenging to model calibration and uncertainty analysis as these routines involve a large number of model runs. For efficient modelling, a cloud-based Calibration and Uncertainty analysis Tool for SWAT (CUT-SWAT) was implemented using Hadoop, an open source cloud platform, and the Generalized Likelihood Uncertainty Estimation method. Test results on an enterprise cloud showed that CUT-SWAT can significantly speedup the calibration and uncertainty analysis processes with a speedup of 21.7–26.6 depending on model complexity and provides a flexible and fault-tolerant model execution environment (it can gracefully and automatically handle partial failure), thus would be an ideal method to solve computational demand problems in hydrological modelling.     

Identifying model consistency through stepwise calibration to capture streamflow variability

The consistency of calibrated hydrological models (whether the model is internally consistent) is often ignored as model calibration generally only evaluates performance. A correlation matrix is developed in this paper to assess model consistency by comparing different flow dynamics combined with a stepwise calibration approach targeting different flow signals (low, medium and high). The Soil Water Assessment tool (SWAT) was used as an example model and simulations were conducted using Sequential Uncertainty Fitting-2 (SUFI-2) algorithm. Critical evaluation of the method demonstrates that satisfactory model performance does not necessarily lead to a consistent model and this yields poor performance in validation. However, results may vary depending on climatic conditions and temporal scales. By calibrating on disaggregated flow signals and evaluating based on consistency, the proposed method improves model realism which will improve understanding of catchment functioning.     

SWATShare – A web platform for collaborative research and education through online sharing,simulation and visualization of SWAT models

Hydrologic models for a particular watershed or a region are created for addressing a specific research or management problem, and most of the models do not get reused after the project is completed. Similarly, multiple models may exist for a particular geographic location from different researchers or organizations. To avoid the duplication of efforts, and enable model reuse and enhancement through collaborative efforts, a prototype cyberinfrastructure, called SWATShare, is developed for sharing, execution and visualization of Soil and Water Assessment Tool (SWAT). The objective of this paper is to present the software architecture, functional capabilities and implementation of SWATShare as a collaborative environment for hydrology research and education using the models published and shared in the system. Besides the capability of publishing, sharing, discovery and downloading of SWAT models, some of the functions in SWATShare such as model calibration are supported by providing access to high performance computing resources including the XSEDE and cloud. Additionally, SWATShare can create dynamic spatial and temporal plots of model outputs at different scales. SWATShare can also be used as an educational tool within a classroom setting for comparing the hydrologic processes under different geographic and climatic settings. The utility of SWATShare for collaborative research and education is demonstrated by using three case studies. Even though this paper focuses on the SWAT model, the system’s architecture can be replicated for other models as well.     

Toward improved calibration of watershed models: Multisite multiobjective measures of information

This paper presents a computational framework for incorporation of disparate information from observed hydrologic responses at multiple locations into the calibration of watershed models. The framework consists of three components: 1) an a-priori characterization of system behavior; 2) a formal and statistically valid formulation of objective function(s) of model errors; and 3) an optimization engine to determine the Pareto-optimal front for the selected objectives. The proposed framework was applied for calibration of the Soil and Water Assessment Tool (SWAT) in the Eagle Creek Watershed, Indiana, USA using three single objective optimization methods [Shuffled Complex Evolution (SCE), Dynamically Dimensioned Search (DDS), and DiffeRential Evolution Adaptive Metropolis (DREAM)], and one multiobjective optimization method. Solutions were classified into behavioral and non-behavioral using percent bias and Nash–Sutcliffe model efficiency coefficient. The results showed that aggregation of streamflow and NO_x (NO₃-N + NO₂-N) information measured at multiple locations within the watershed into a single measure of weighted errors resulted in faster convergence to a solution with a lower overall objective function value than using multiple measures of information. However, the DREAM method solution was the only one among the three single objective optimization methods considered in this study that satisfied the conditions defined for characterizing system behavior. In particular, aggregation of streamflow and NO_x responses undermined finding “very good” behavioral solutions for NO_x, primarily because of the significantly larger number of observations for streamflow. Aggregation of only NO_x responses into a single measure expedited finding better solutions although aggregation of data from nested sites appeared to be inappropriate because of correlated errors. This study demonstrates the importance of hydrologic and water quality data availability at multiple locations, and also highlights the use of multiobjective approaches for proper calibration of watershed models that are used for pollutant source identification and watershed management.     

基于RMC的蒙特卡罗程序性能优化

蒙特卡罗MC方法是核反应堆设计和分析中重要的粒子输运模拟方法。MC方法能够模拟复杂几何形状且计算结果精度高,缺点是需要耗费大量时间进行上亿规模粒子模拟。如何提高蒙特卡罗程序的性能成为大规模蒙特卡罗数值模拟的挑战。基于堆用蒙特卡罗分析程序RMC,先后开展了基于TCMalloc动态内存分配优化、OpenMP线程调度策略优化、vector内存对齐优化和基于HDF5的并行I/O优化等一系列优化手段,对于200万粒子的算例,使其总体性能提高26.45%以上。     

A quantitative phosphorus loss assessment tool for agricultural fields

In the United States, government sponsored conservation programs are under increasing pressure to quantify the environmental benefits of practices they subsidize. To meet this objective, conservation planners need tools to accurately predict phosphorus (P) loss from agricultural lands. Existing P export coefficient based tools are easy to use, but do not adequately account for local conditions. Hydrologic and water quality models are more accurate, but are prohibitively complex for conservation planners to use. Pasture Phosphorus Management (PPM) Plus was developed as a user-friendly P and sediment loss prediction tool based on the Soil and Water Assessment Tool (SWAT), a popular comprehensive hydrologic and water quality model. PPM Plus is applicable under a wide variety of management options and conservation practices and simple enough for use by conservation planners. SWAT hydrologic components were calibrated to allow application anywhere in the State of Oklahoma. The SWAT model was modified to include soil P algorithm updates and improved representation of conservation practices. This tool was successfully validated using 286 field years of measured data from the southern United States. PPM Plus allows the development of more effective conservation plans by allowing planners to evaluate pollutant losses resulting from a particular management strategy prior to implementation.     

Sequential improvement for robust optimization using an uncertainty measure for radial basis functions

The performance of the sequential metamodel based optimization procedure depends strongly on the chosen building blocks for the algorithm, such as the used metamodeling method and sequential improvement criterion. In this study, the effect of these choices on the efficiency of the robust optimization procedure is investigated. A novel sequential improvement criterion for robust optimization is proposed, as well as an improved implementation of radial basis function interpolation suitable for sequential optimization. The leave-one-out cross-validation measure is used to estimate the uncertainty of the radial basis function metamodel. The metamodeling methods and sequential improvement criteria are compared, based on a test with Gaussian random fields as well as on the optimization of a strip bending process with five design variables and two noise variables. For this process, better results are obtained in the runs with the novel sequential improvement criterion as well as with the novel radial basis function implementation, compared to the runs with conventional sequential improvement criteria and kriging interpolation.