基于PSO-WRF的渭河中下游年径流预测模型

针对传统随机森林( RF) 模型决策树因投票权重相同而导致预测精度不高的问题,采用加权随机森林( WRF) 模型全面考虑各决策树分类能力的差异性,建立决策树加权投票机制; 同时运用粒子群算法( PSO) 进行参数全局寻优,避免依据经验选取参数的不科学性,通过二者耦合最终构建PSO-WRF 模型。利用渭河中下游咸阳站与华县站 1960—2009 年的径流系列对 RF、WRF、PSO-WRF三种模型进行训练及测试,结果表明,PSO-WRF 在咸阳站与华县站的平均相对误差绝对值( MRE) 分别为 7. 05%和 9. 41%,且均方根误差( RMSE) 、平均绝对误差( MAE) 等指标均降至优化前的 30% ～50%,各年预测误差最低可降低至优化前的 1 /3 ～ 1 /6。PSO-WRF 模型优化效果显著,表现出良好的预测精度和泛化能力,能够为相关径流预测研究提供一定参考。     

大型河工模型分布式表面流场测量系统研制及应用

在河工模型试验中, 粒子图像表面流场测量方法得到了广泛应用。研制了一种新型分布式表面流场测量系统,该系统采用局域网组网与光纤传输相结合,通过POE千兆交换机与高清智能一体化工业摄像机相连,显著降低了布线复杂度,具有系统传输距离远、布设简单、集成度高、可扩展性强等优点。系统具备可视化全自动采集、可视化错误矢量剔除、导出多种数据格式,生成流场等值线图、流线等功能。在系统研制基础上,提出了一种对粒子图像表面流场测量系统进行精度检测的新方法,通过精确控制匀速旋转平台模拟水流运动,将表面流场测量系统实测数据与旋转平台上各点精确数据进行对比检测,检测结果表明,研制的表面流场测量系统测量误差小于5%,已在长江河口模型等多个大型河工模型中得到成功应用。     

一种高效的SWAT模型参数自动率定方法

本文分析了SWAT模型和PSO算法的原理,将PSO算法引入SWAT模型中,构建了新的SWAT模型参数自动率定模块,通过在天津蓟运河流域实例研究,发现该方法率定精度较高,收敛速度快,运行结果稳定,整体率定效果优于模型自带的参数率定模块。如果用改进后的模块在Linux平台开展自动率定,可以使模型自动率定效率提高到当前水平的7倍,适用于大型流域或长时间系列模拟。而PSO算法作为一种通用的优化算法,可广泛用于各种水文模型的参数率定。     

Multi-Criteria Process-Based Calibration Using Functional Data Analysis to Improve Hydrological Model Realism

It has been argued that rainfall-runoff model calibration based solely on streamflow is not sufficient to evaluate the realism of a hydrological model to represent the internal fluxes. Therefore, model calibration has evolved to evaluating model performance using a number of hydrological signatures that link the model to the underlying processes. However, this approach uses goodness-of-fit measures, unable to describe the entire dynamic of time series, to evaluate model consistency and to simulate hydrological signatures. The present paper develops a stepwise multicriteria calibration using hydrograph partitioning and calibration criteria defined on the basis of Functional Data Analysis (FDA), a statistical tool that conserves all important features of the hydrograph by approximating times series as a single function. The aim of this approach is to improve model realism by scrutinizing model components and by evaluating its ability to reproduce the entire flow dynamic. The proposed approach is compared to a calibration against daily streamflow only. The stepwise calibration improved the estimation of the flood curve, the annual peak volume as well as the performance of the model at sites other than the calibration station.     

遗传程序设计在南水北调中线流量计率定中的应用

针对传统流量计率定中相关线选择的随意性及拟合精度不高的问题,提出了基于遗传程序设计流量计率定方法。采用二叉树形式存储函数表达式,并分别将流量计测所得流量和相应的流速仪得到的流量作为训练的输入和输出样本,经过各种演化计算,寻得最优模型结构。将遗传程序设计应用于南水北调中线流量计的率定中,并同传统率定方法进行了详细比较。实例表明,该方法能够通过演化计算自动寻找最满意的描述流量关系的函数,比起传统统计方法具有较大的灵活性和智能性,避免了预先假定具体数学表达式的不足,且拟合精度高,为流量计率定提供了新的有效方法。     

Surface wave measurements with IoT image processing

This study develops two different approaches to perform temporal and spatial measurements of surface wave profile for experimental studies in transparent wave flumes. Both are based on image acquisition and processing with an Internet of Things (IoT) system consisting of three sets of GoPro camera cum Raspberry Pi connected wirelessly together in a local LAN. The first approach uses advanced edge algorithms with perspective transformation of the multiple cameras for the detection, while the second approach adopts Convolutional Neural Network (CNN) algorithms instead with training of the processed image data using information from additional discrete probes installed. Their accuracy is assessed under a range of experimental conditions of regular and irregular waves with different wave heights and periods, based on metrics that consist of the average errors of the predicted water surface profile as well as position errors for wave crests and troughs. The effects on the measurement accuracy due to the image acquisition frequency, camera resolution and camera location are also investigated. The results show that higher wave steepnesses generally lead to larger detection errors, and measurements for irregular waves are also more challenging. In addition, positioning the cameras closer to the wave flume sidewalls yields better detection results as expected, particularly in resolving wave crests and troughs, although the field of view narrows at the same time. However, higher video frequencies and camera resolutions might not necessarily improve the accuracy contrary to common expectation due to jaggedness in the image processing. Overall, both approaches are shown to be viable for the measurement of wave profile in the laboratory. The first approach is more straight forward in terms of implementation, and it performs well for regular wave conditions. The second approach requires more complex training of the neural networks, but its accuracy is significantly higher particularly for irregular waves.     

A study of the stochastic evolution of hydraulic geometry relationships

Uncertainty resulting from climatic and environmental changes creates barriers to the accurate acquisition of information about rivers. In this study, a set of stable stochastic differential equations is developed to simulate the dynamic probability distributions of typical hydraulic geometry variables represented by slope, width, depth, and velocity with variations in bankfull discharge over time for a river system. Random parts of the equations are modelled based on single Gaussian white noise and further on combined Gaussian/fractional white noise with Poisson noise. Consistent estimates of the equation parameters are made using a composite nonparametric maximum likelihood estimation (MLE) method. The proposed models are examined through a Monte Carlo simulation of the lower Yellow River, and the results successfully reveal the potential responses of hydraulic geometries to stochastic disturbance and that average trends largely synchronize with the measured values. Comparisons of the three different models confirm the advantages of fractional jump‐diffusion model, and according to further discussion, stream power on the basis of such a model is concluded to serve as the better systematic measure of river dynamics. The proposed stochastic approach is new to the field of fluvial relationships, and its application could help to design and monitor river systems with specified accuracy requirements.     

Reservoir Inflow Modeling Using Temporal Neural Networks with Forgetting Factor Approach

In this paper, a recursive training procedure with forgetting factor is proposed for on-line calibration of temporal neural networks. The forgetting factor discounts old measurements through an on-line model calibration. The forgetting factor approach enables the recursive algorithm to reduce the effect of the older error data by multiplying the error data by a discounting factor. The proposed procedure is used to calibrate a temporal neural network for reservoir inflow modeling. The mean monthly inflow of the Karoon-III reservoir dam in the south-western part of Iran is used to test the performance of the proposed approach. An autoregressive moving average (ARMA) model is also applied to the same data. The temporal neural network, which is trained with the proposed approach, has shown a significant improvement in the forecast accuracy in comparison with the network trained by the conventional method. It is also demonstrated that the neural network trained with forgetting factor results in better forecasts compared to the statistical ARMA model, which has been calibrated through this approach.     

POS优化算法在SWAT模型参数优化中的应用

当前,分布式水文模型SWAT模型在国内水文模拟中应用较为广泛,但模型参数较多,人工经验设定参数存在工作量较为繁杂,且模拟精度不高的缺陷,为此本文引入POS优化算法,对SWAT模型参数进行批量优化,并以汤河西支流域为研究区域,结合区域实测水文数据,对比分析参数优化前后,对SWAT模型模拟精度的影响。研究结果表明： POS优化算法可实现SWAT模型参数的批量优化,相比参数优化前,参数优化后SWAT模型模拟径流深相对误差减少5.7%,流量过程拟合系数提高0.118。研究成果对于水文模型参数优化和自动率定提供参考价值。     

Design of a retention tank: comparison of stormwater quality models with various levels of complexity.

Stormwater quality simulation models are useful tools for the design and management of sewer systems. Modelling results are highly sensitive to experimental data used for calibration. This sensitivity is examined for three modelling approaches of various complexities (site mean concentration approach, event mean concentration approach and build-up, washoff and transport modelling approach) applied to a typical case study (design of a dry detention tank), accounting for the variability of calibration data and their effect on simulation results. Calibrated models with different calibration data sets were used to simulate 3 years of rainfall with different retention tank specific volumes. Annual pollutant load interception efficiencies were determined. Simulations results revealed (i) that there is no advantage in using the EMC model compared to the SMC model and (ii) that the BWT model resulted in higher design ratios than those given by the SMC/hydraulic approach. For both EMC and BWT models, using an increasing number n of events for calibration leads to narrower confidence intervals for the design ratios. It is crucial for design ratios to account for successive storm events in chronological order and to account for the maximum allowable flow to be transferred to the downstream WWTP.     

Potential Improvement of the Parameter Identifiability in Ungauged Catchments

The objective of this study was to reduce the parameter uncertainty which has an effect on the identification of the relationship between the catchment characteristics and the catchment response dynamics in ungauged catchments. Model deficiencies influencing on the identification of the regional relationships were identified through analysing the non-stationarity nature under different climate conditions. An advanced calibration approach was proposed to improve the identification of the regional relationships, according to the deficiencies on model structure suitability for the different flow regime. This study demonstrated the refined calibration strategy can improve the identification of the relationships between the catchment characteristics and the calibrated model parameters for the dry period. In the assessment of model structure suitability to represent the non-stationary catchment response characteristics, there was a flow-dependent bias in the runoff simulations. In particular, over-prediction of the streamflow was dominant for the dry period. The poor model performance during the dry period was associated with the largely different impulse response estimates for the entire period and the dry period. Based on assessment of model deficiencies, the rainfall–runoff models were separately calibrated to different parts of the flow regime, and the calibrated models for the separated time series were used to establish the regional models of relevant parts of the flow regime (i.e. wet and dry periods). The effectiveness of the parameter values for the refined approach in regionalisation was evaluated through investigating the accuracy of predictions of the regional models. The regional models from the refined calibration approach clearly enhanced the hydrological behaviour for the dry period by improving the identification of the relationships between the catchment attributes and the catchment response dynamics representing the time constants in fitting recession parts of hydrograph (i.e. improving the parameter identifiability representing the different behaviour of the catchment) in regionalisation.     

Improvement of numerical simulation of bimolecular reactive transport using time-dependent parameters

The overestimation of the contaminant concentration is a main issue in simulating the reactive transport using the common advection-dispersion-reaction equation(ADRE). To solve this problem, a new modeling method is developed. The parameters of the model are identified based on experimental data. The results of the model are compared with previous results in the literature, and the sensitivity of the model is analyzed by examining the model's response to changes of the model parameters. Main conclusions are as follows:(1) The numerical modeling approach is feasible with an improved simulating accuracy. The predicted values are in agreement with the experimental measurements. The relative errors of the peak concentration between the simulated and experimental values are less than 2.5%. The errors are greatly reduced as compared with the results(about 67.8% as the maximum) based on a traditional ADRE in the literature.(2) There are three parameters( m, 0 and D) which can be calibrated on the basis of experimental data in the model. The reactive product concentrations are mainly influenced by the parameters involved in the reactive ratio such as m and 0. The hydrodynamic dispersion coefficient D has almost no influence on the reactive product transport.(3) Our model does not provide better fitting curves for the "early arrival" and the "long tail". The "early arrival" and the "long tail" are associated with low values of the product AB. Under these conditions, the reaction rate is close to 0, and the model of the ADRE reduces to the advection-dispersion equation(ADE). Further mechanism study is needed in the future.     

融冰洪水演进的马斯京根模型

为解决参数率定过程复杂的问题,将河段内的融冰产生的流量视为河段下断面出流的一部分,构建出适用于文开河融冰时期洪水演进的马斯京根模型,并将优化算法应用到模型参数率定的过程中。以黄河宁蒙河段为例,采用试错法、非线性规划法和智能算法中的遗传算法这3种方法对所构建的模型参数进行率定。模拟结果表明:3种方法所模拟的出流过程线均合格;整体上,非线性规划法模拟的精度最高,其洪水的确定性系数D_c为0.980,过程平均相对误差RE为4.840%;而试错法模拟的洪峰流量更为准确,且冰期模拟精度高于无冰期。本研究为融冰洪水演进的模拟提供了一种新方法。     

基于改进降水输入模块的融雪径流模拟:以拉萨河为例

降水是自然界物质循环和水循环的重要组成部分,是高寒地区径流的重要来源,水文模型中降水数据的输入精度对提高高寒地区融雪径流模拟效果具有十分重要的作用。青藏高原地区气象站点较少,站点数据无法全面反映流域内降水时空分布的真实情况,传统的融雪径流模型在地形、风向和水汽等要素对降水垂直分布的影响考虑不够全面,制约了模型在山区融雪模拟以及预测中的应用,因此有必要对模型的降水输入项进行改进,以期提高半干旱高寒地区融雪径流模拟效果。本文基于改进的遥感卫星数据校正理论,开发了适用于半干旱高寒地区的降水输入模块,将其与度日因子模型进行耦合,利用高程分带将降水组合成半网格半站点的降水输入数据驱动模型,并在拉萨河流域进行试验研究。结果表明:降水输入模块能够显著提高降水卫星反演地面降水精度,改进后的融雪模型在率定期和验证期的NSE(Nash-Sutcliffe efficiency coefficient)分别为0.741和0.770,高于原融雪模型的模拟效果,表明改进后的模型能够在流域各个分区获得较为精确的降水数值,融雪径流模拟精度比原模型精度得到提高。总之,耦合降水输入模块的融雪模型可以有效提高降水输入精度,对缺资料半干旱高寒地区融雪模拟具有重要的参考价值。     

Real-time Neural-network-based Ensemble Typhoon Flood Forecasting Model with Self-organizing Map Cluster Analysis: A Case Study on the Wu River Basin in Taiwan

Accurate hourly real-time flood forecasting is necessary for early flood warning systems, especially during typhoon periods. Artificial intelligence methods have been increasingly used for real-time flood forecasting. This study developed a real-time flood forecasting model by using back-propagation networks (BPNs) with a self-organizing map (SOM) to create ensemble forecasts. Random weights and biases were set for the BPNs to learn the characteristics of a catchment system. An unsupervised SOM network with a classification function was then used to cluster representative BPN weights and biases; clusters of BPNs with high accuracy were selected to act as experts for the ensemble models to forecast flow rates. The model was applied to flood events in the Wu River Basin of Taiwan. Most observed values were within the forecasting intervals of the BPN clusters in the calibration and validation phases, indicating that the models had acceptable accuracy. For the large flood events of typhoons Saola in the calibration phase and Soulik in the validation phase, the mean average error of the ensemble mean model for the cluster A was 143.1 and 327.4 m³/s, respectively; these values were lower than those for the best individual model within the cluster (194.3 and 917.9 m³/s). The ensemble model thus outperformed the individual models and can accurately forecast flood values and intervals. Therefore, the model can be used to accurately forecast floods.

     

GA-ILP Method for Optimization of Water Distribution Networks

Optimization of water distribution networks has been of central importance for recent decades. Genetic Algorithms (GA) are the most famous metaheuristics widely used for this purpose with great success. However, the fact that GA basically requires a large number of computations, has led to investigate for faster solvers. In this research, a new approach is proposed in which a simple GA is linked with the Integer-Linear Programming (ILP) method resulting in a hybrid optimization scheme. Using the mathematical method of ILP, the search space is significantly reduced thereby a limited number of evaluations are required to achieve a good solution. The approach is applied to two benchmark pipe-networks in order to show its ability in terms of accuracy and speed. The results are then compared with the previous works. The obtained results indicate that the proposed model is computationally efficient, like classic methods, while is still very promising in finding the global optimum like the nature-inspired metaheuristics.     

Test of newly developed conceptual hydrological model for simulation of rain-on-snow events in forested watershed

A conceptual hydrological model that links the Xin’anjiang hydrological model and a physically based snow energy and mass balance model, described as the XINSNOBAL model, was developed in this study for simulating rain-on-snow events that commonly occur in the Pacific Northwest of the United States. The resultant model was applied to the Lookout Creek Watershed in the H. J. Andrews Experimental Forest in the western Cascade Mountains of Oregon, and its ability to simulate streamflow was evaluated. The simulation was conducted at 24-hour and one-hour time scales for the period of 1996 to 2005. The results indicated that runoff and peak discharge could be underestimated if snowpack accumulation and snowmelt under rain-on-snow conditions were not taken into account. The average deterministic coefficient of the hourly model in streamflow simulation in the calibration stage was 0.837, which was significantly improved over the value of 0.762 when the Xin’anjiang model was used alone. Good simulation performance of the XINSNOBAL model in the WS10 catchment, using the calibrated parameter of the Lookout Creek Watershed for proxy-basin testing, demonstrates that transplanting model parameters between similar watersheds can provide a useful tool for discharge forecasting in ungauged basins.     

Application of New Hybrid Optimization Technique for Parameter Estimation of New Improved Version of Muskingum Model

The Muskingum method is one of the most utilized lumped flood routing model in which calibration of its parameters provides an active area of research in water resources engineering. Although various techniques and versions of Muskingum model have been presented to estimate the parameters of different versions of Muskingum model, more rigorous approaches and models are still required to improve the computational precision of calibration process. In this study, a new hybrid technique was proposed for Muskingum parameter estimation which combines the Modified Honey Bee Mating Optimization (MHBMO) and Generalized Reduced Gradient (GRG) algorithms. According to the conducted literature-review on the improvement of Muskingum flood routing models, a new six-parameter Muskingum model was proposed. The hybrid technique was successfully applied for parameter estimation of this new version of Muskingum model for three case studies selected from literature. The obtained results were compared with those of other methods using several common performance evaluation criteria. The new hybrid method with the new proposed Muskingum model perform the best among all the considered approaches based on most of utilized criteria. The new Muskingum model significantly reduces the SSQ value for the double-peak case study. Finally, the achieved results demonstrate that not only the hybrid MHBMO-GRG algorithm overcomes the shortcomings of both phenomenon-mimicking and mathematical optimization techniques, but also the presented Muskingum model is appeared to be the most reliable version of Muskingum model comparing with other considered models in this research.     

数值集合降雨预测的校正后处理方法研究

利用2017年4月1日-9月30日全球集合预测系统的降雨预测数据和雅砻江流域气象站点的降雨观测数据,采用基于左删失广义极值分布的集合模式输出统计方法对流域内降雨预测进行校正,对比分析该方法两种建模形式在校正结果上的差异。结果表明：采用集合成员均值校正的方式可以有效改善原始预测对于降雨过分高估的问题,其预测结果明显优于采用集合成员校正方式的预测结果,后者由于模型参数增加而出现过度拟合问题,限制了其在雅砻江流域中的应用。另外采用集合成员均值校正方式的预测结果的准确性在不同流域范围存在明显差异并倾向低估流域内较大降雨量,因此在后续的研究中需要进一步针对该方法无法对极值降雨量进行准确预测的问题进行改进。