River Flow Modelling Using Fuzzy Decision Trees

A modern real time flood forecasting system requires itsmathematical model(s) to handle highly complex rainfall runoffprocesses. Uncertainty in real time flood forecasting willinvolve a variety of components such as measurement noise fromtelemetry systems, inadequacy of the models, insufficiency ofcatchment conditions, etc. Probabilistic forecasting is becomingmore and more important in this field. This article describes a novel attempt to use a Fuzzy Logic approach for river flow modelling based on fuzzy decision trees. These trees are learntfrom data using the MA-ID3 algorithm. This is an extension of Quinlan's ID3 and is based on mass assignments. MA-ID3 allows for the incorporation of fuzzy attribute and class values intodecision trees aiding generalisation and providing a framework for representing linguistic rules. The article showed that with only five fuzzy labels, the FDT model performed reasonably welland a comparison with a Neural Network model (Back Propagation)was carried out. Furthermore, the FDT model indicated that therainfall values of four or five days before the prediction time are regarded as more informative to the prediction than the morerecent ones. Although its performance is not as good as the neural network model in the test case, its glass box nature couldprovide some useful insight about the hydrological processes.     

一维非恒定水动力学模型的实时洪水预报

本文建立了一维非恒定水动力学模型与最小二乘法耦合的实时洪水预报模型。运用最小二乘的递推算法，引入时变遗忘因子实时跟踪模型中时变参数的变化，并动态修正参数，实时改变一维非恒定水动力学模型结构，对洪水预报进行实时修正。通过长江干流宜昌～螺山段洪水实时预报应用．改进后模型预报精度比原一维非恒定水动力学模型计算结果有很大提高，表明所建立的模型用于洪水预报是行之有效的，在洪水预报中具有推广应用价值。     

REAL-TIME FLOOD FORECASTING METHOD WITH 1-D UNSTEADY FLOW MODEL

A real-time forecasting method coupled with the 1-D unsteady flow model with the recursive least-square method was developed. The 1-D unsteady flow model was modified by using the time-variant parameter and revising it dynamically through introducing a variable weighted forgetting factor,such that the output of the model could be adjusted for the real time forecasting of floods. The application of the new real time forecasting model in the reach from Yichang to Luoshan of the Yangtze River was demonstrated. Computational result shows that the forecasting accuracy of the new model is much higher than that of the original 1-D unsteady flow model. The method developed is effective for flood forecasting,and can be used for practical operation in the flood forecasting.     

数值天气预报的成功经验对洪水预报的启示

鉴于近30年来数值天气预报有了长足的进步,而在洪水预报中试图推行流域水文模型却进展缓慢这一现状,试图寻找洪水预报精度与数值天气预报精度的差距原因。通过对大气模式(型)和流域水文模型的比较,围绕预报方法及预见期、初始场和边值场、数据同化和融合等问题,论述了数值天气预报的成功经验和流域水文模型在洪水预报应用中存在的问题,期望洪水预报能从中得到启示,在不远的将来实现流域水文模型与数值天气预报的无缝衔接,以进一步提高洪水预报精度,增长洪水预报预见期。     

基于最大Lyapunov指数的混沌预测在洪水实时预报中的应用

为提高洪水预报精度,从而提高水库防洪调度可信度,采用混沌预测模型,研究了复杂洪水动力特征和非线性特征并分析洪水实时预报问题。考虑到时间延迟与嵌入维数的相关性,利用C-C法计算相空间重构参数,判别了洪水混沌特性。为避免由嵌入维数m的选取引起的最大Lyapunov指数的明显波动,使用了改进的小数据量方法计算最大Lyapunov指数。构造了基于最大Lyapunov指数的混沌洪水实时预报模型,并将其应用到湖南五强溪水库的洪水预报,计算结果表明该模型具有较高的预报精度。     

基于气象模式的汉江流域洪水预报系统

针对现行洪水预报方法的不足与汉江流域的特点,建立了汉江流域MM5气象预报模式和VIC分布式水文模型,耦合集成3种预报模式并开发了基于气象模式的汉江流域洪水预报系统。应用2005~2007年汛期日水文气象资料进行验证,结果表明所建VIC分布式水文模型具有较高的模拟与预报精度。基于气象模式的汉江流域洪水预报系统可为汉江流域的中长期水文预报和水资源综合管理提供技术支撑和决策参考。     

Flood Control Operations Based on the Theory of Variable Fuzzy Sets

Flood control decisions are often involved with quantitative and qualitative criteria. In this paper, a decision model is presented for flood control operations based on the theory of variable fuzzy sets. Using dual comparison, two models computing relative membership grades with qualitative and quantitative criteria are established, respectively. A method integrating subjective preference and iterative weights is proposed for weight-assessment. First, an initial solution of criteria weights is obtained by using proposed fuzzy optimal iteration model. Then, according to their knowledge related to real time flood operations, operators may modify the initial weights if necessary. When the relative membership grades of alternatives belonging to all rankings are fixed by using multi-criterion variable fuzzy model proposed, the decision alternative is chosen according to the ranking characteristic value computed using a defuzzification equation. The case study of Fengman Reservoir flood operation (in China) is provided to illustrate the application of the proposed method. With the incorporation of operator’s knowledge related to flood operations, the proposed model is flexible and practical.     

门限人工神经网络模型及其在洪水预报中的应用

结合门限自回归模型与人工神经网络模型的建模思想，首次提出这两种方法的耦合模型，即门限人工神经网络模型，新模型的实质是一种分段非线性化的处理方法，是对现有门限模型分段线性化的很好改进。实例计算结果说明，新模型在洪水的预报中是有效的，在各种非线性时序动态预测中具有普遍意义和广泛的实用价值。     

岩溶地区径流预报中BP网络的改进

岩溶地区下垫面复杂,各种岩溶管道、裂隙、溶洞发育使得流域不闭合,地下暗河存在水量交换,而地下水库的调蓄作用,使得流域出口断面总流量与降雨量不成绝对的线性关系。为了克服上述问题带来的岩溶地区降雨径流预报精度低问题,提出了改进的BP网络方法,并通过实例验证了此方法的可行性。以六冲河七星关站断面以上流域的平均日降水量、平均日蒸发量、前期流量作为影响因子,建立了2种预报模型：①传统BP网络模型;②运用SPASS软件筛选BP的影响因子数和调整输入层初始权值,并对逐日径流量资料进行对数处理建立改进的BP网络模型。通过实例分析发现改进的BP网络模型预报效果更好,可以有效地提高大洪峰和小洪峰的预报精度。     

Improving Flood Forecasting in a Developing Country: A Comparative Study of Stepwise Multiple Linear Regression and Artificial Neural Network

Due to limited data sources, practical situations in most developing countries favor black-box models in real time operations. In a simple and robust approach, this study examines performances of stepwise multiple linear regression (SMLR) and artificial neural network (ANN) models, as tools for multi-step forecasting Chindwin River floods in northern Myanmar. Future river stages are modeled using past water levels and rainfall at the forecasting station as well as at the hydrologically connected upstream station. The developed models are calibrated with flood season data from 1990 to 2007 and validated with data from 2008 to 2011. Model performances are compared for 1- to 5-day ahead forecasts. With a high accuracy, both candidate models performed well for forecasting the full range of flood levels. The ANN models were superior to the SMLR models, particularly in predicting the extreme floods. Correlation analysis was found to be useful for determining the initial input variables. Contribution of upstream data to both models could improve the forecasting performance with higher R ² values and lower errors. Considering the commonly available data in the region as primary predictors, the results would be useful for real time flood forecasting, avoiding the complexity of physical processes.     

KALMAN FILTERING CORRECTION IN REAL-TIME FORECASTING WITH HYDRODYNAMIC MODEL

Accurate and reliable flood forecast is crucial for efficient real-time river management, including flood control, flood warning, reservoir operation and river regulation. In order to improve the estimate of the initial state of the forecasting system and to reduce the errors in the forecast period a data assimilation procedure was often need. The Kalman filter was proven to be an efficient method to adjust real-time flood series and improve the initial conditions before the forecast. A new model integrating the hydraulic model with the Kalman filter for real-time correction of flood forecast was developed and applied in the Three Gorges interzone of the Yangtze River. The method was calibrated and verified against the observed flood stage and discharge during Three Gorges Dam construction periods （2004）. The results demonstrate that the new model incorporates an accurate and fast updating technique can improve the reliability of the flood forecast.     

Rainfall-runoff simulation and flood forecasting for Huaihe Basin

The main purpose of this study was to forecast the inflow to Hongze Lake using the Xin＇anjiang rainfall-runoff model. The upper area of Hongze Lake in the Huaihe Basin was divided into 23 sub-basins, including the surface of Hongze Lake. The influence of reservoirs and gates on flood forecasting was considered in a practical and simple way. With a one-day time step, the linear and non-linear Muskingum method was used for channel flood routing, and the least-square regression model was used for real-time correction in flood forecasting. Representative historical data were collected for the model calibration. The hydrological model parameters for each sub-basin were calibrated individually, so the parameters of the Xin＇anjiang model were different for different sub-basins. This flood forecasting system was used in the real-time simulation of the large flood in 2005 and the results are satisfactory when compared with measured data from the flood.     

三水源模型在滦河流域的应用分析

流域洪水预报模型大致可分蓄满产流和超渗产流2种方式,针对滦河流域的地形、地貌及雨洪特性,采用三水源模型,属于蓄满产流方式。介绍三水源模型实时洪水预报系统在滦河流域的应用。利用模型,根据计算误差来实时修正估计结果、模型参数等,使得实时洪水预报系统具有较好的实时修正和动态跟踪能力。实际应用过程中,由于流域实际产流过程是复杂、多样的,蓄满与超渗2种产流方式在同一区域可能同时存在,预报结果与实测数据存在一定误差,但通过实例分析,多数误差在规定的范围内,正常情况下的预报精度较高,基本上满足了水库防汛要求。     

长江三峡洪水预报系统简介

本文介绍了利用水文数字模型制定三峡洪水作业预报系统的尝试，对一些有关问题进行了讨论，水利部水文利调度中心自１９８５年起已将该系统用于作业预报，１９８７年移植于葛洲坝电厂做入库流量预报，该系统从历史水文资料输入、模型率定、实时资料处理到预报结果输出，以及历史资料的显示，过程线对比等都实现了计算机自动化，预报精度完全满足生产要求，是一个比较完整的实时洪水预报系统。     

流域洪水预报系统及其关键问题研究

论述了流域洪水预报系统的主要任务和核心内容,介绍了流域洪水预报系统的结构和主要功能,分析了系统的主要功能模块,将智能算法与传统的水文预报方法相结合,在水文预报模型参数自动率定、实时洪水校正两方面进行了尝试,并建立了相应的模型,经过流域长系列水文资料检验和应用,获得了满意的结果,为实时洪水预报开辟了新的途径。     

长江中下游洪水联机实时预报系统

长期以来，长江洪水预报一直采取手工作业方式，为了提高洪水预报的实时性、准确性和可靠性，长委水文局预报处与其他单位共同研制出长江中下游洪水联机实时预报系统，于１９９５年投入运用。该系统由预报作业、水情检索和系统管理三部分构成，融合了先进实用的模型理论、计算机信息技术和专家经验。实际运用中，采用预报实时校正的方法，由预报员对预报值进行比较分析后确认。系统试用结果表明：由于将计算机系统与预报员作业连为一体，提高了成果质量和工作效率，将实时跟踪能力强的ＣＲＦＰＤＰ模型投入作业运行，提供了分析预报结果合理性的有力工具；利用预报员的经验，注重对模型预报值的可靠性和可能的偏差方面分析是非常必要的。另外针对系统数据库管理问题，提出了研制新的洪水预报系统的设想。     

基于PDM的飞来峡水库实时水文预报模型

水文预报模型是实时预报预警系统的核心。基于新型的概率水文预报PDM(The Probability Distributed Model),构建了北江流域飞来峡水库概率水文预报模型。将整个流域划分为18个子流域,对每个子流域分别建立PDM,并采用历史降雨流量数据序列对各子流域和流域PDM进行参数率定;以横石站为参考站,采用52场次实测历史洪水进行模型验证和精度评定,洪峰预报合格率为99%,精度等级为甲等,模拟精度高,说明了PDM结构和参数取值的合理性,以及模型在该流域的适用性。     

A fuzzy inference method based on association rule analysis with application to river flood forecasting

In this paper, a computationally efficient version of the widely used Takagi-Sugeno (T-S) fuzzy reasoning method is proposed, and applied to river flood forecasting. It is well known that the number of fuzzy rules of traditional fuzzy reasoning methods exponentially increases as the number of input parameters increases, often causing prohibitive computational burden. The proposed method greatly reduces the number of fuzzy rules by making use of the association rule analysis on historical data, and therefore achieves computational efficiency for the cases of a large number of input parameters. In the end, we apply this new method to a case study of river flood forecasting, which demonstrates that the proposed fuzzy reasoning engine can achieve better prediction accuracy than the widely used Muskingum-Cunge scheme.     

MSKLOSS河道洪水演算模型参数敏感性分析

模型参数的敏感性分析是洪水预报的基础性工作。以中国洪水预报系统为平台,采用局部分析法,对MSKLOSS河道洪水演算模型参数进行敏感性分析,以提高洪水预报的精度。为提高MSKLOSS河道洪水演算模型的应用效率,考虑模型适用性、模拟分析方便性,选取海河南系河道下垫面接近现状条件的滹沱河黄壁庄-北中山、南运河岳城-蔡小庄"96·8"和大清河东茨村-新盖房"12·7"三场洪水资料对模型参数进行分析。分析表明:河道分段数、初始下渗率、下渗曲线指数和湿周等参数的敏感性较高,稳定下渗率的敏感性在洪水量级较小时相对较高,在参数优选和实时作业预报时需注意其初始范围的设定。     

Improving the real-time probabilistic channel flood forecasting by incorporating the uncertainty of inflow using the particle filter

An accurate and reliable real-time flood forecast is crucial for mitigating flood disasters. The errors associated with the inflow boundary forcing data are considered as an important source of uncertainties in hydraulic model. In this paper, a real-time probabilistic channel flood forecasting model is developed with a novel function to incorporate the uncertainty of the forcing inflow. This new approach couples a hydraulic model with the particle filter(PF) data assimilation algorithm, a sequential Bayesian Monte Carlo method. The stage observations at hydrological stations are assimilated at each time step to update the model states in order to improve the next time step's forecasting. This new approach is tested against a real flood event occurred in the upper Yangtze River. As compared with the open loop simulations, the evaluations of model performance with several deterministic and probabilistic metrics indicate that the accuracy of the ensemble mean prediction and the reliability of the uncertainty quantification are improved pronouncedly as a result of the PF assimilation. Further assessment of the prediction results at different lead times shows that the improvement of model performance deteriorates with the increase of the lead time due to the gradual diminishing of the updating effect for the initial conditions. Based on the analyses of the number of particles and the assimilation frequency, we find that the optimal number of particles can be determined by balancing the model performance and the computation cost, while a high assimilation frequency is preferred to incorporate the emerging observations to update the model states to match the real conditions.