Ensemble Based Forecasting and Optimization Framework to Optimize Releases from Water Supply Reservoirs for Flood Control

Water supply reservoir management is based on long-term management policies which depend on customer demands and seasonal hydrologic changes. However, increasing frequency and intensity of precipitation events is necessitating the short-term management of such reservoirs to reduce downstream flooding. Operational management of reservoirs at hourly/daily timescales is challenging due to the uncertainty associated with the inflow forecasts and the volumes in the reservoir. We present an ensemble-based streamflow prediction and optimization framework consisting of a regional scale hydrologic model forced with ensemble precipitation inputs to obtain probabilistic inflows to the reservoir. A multi-objective dynamic programming model was used to obtain optimized release strategies accounting for the inflow uncertainties. The proposed framework was evaluated at a water supply reservoir in the Hackensack River basin in New Jersey during Hurricanes Irene and Sandy. Hurricane Irene resulted in the overtopping of the dam despite releases made in anticipation of the event and resulted in severe downstream flooding. Hurricane Sandy was characterized by low rainfall, however, raised significant concerns of flooding given the nature of the event. The improvement in NSE for the Hurricane Irene inflows from 0.5 to 0.76 and reduction of the spread of PBIAS with decreasing lead times resulted in improvements in the forecast informed releases. This study provides perspectives on the benefits of the proposed forecasting and optimization framework in reducing the decision making burden on the operator by providing the uncertainties associated with the inflows, releases and the water levels in the reservoir.

     

Multiplicative seasonal Arima model for longterm forecasting of inflows

This paper deals with stochastic modelling of monthly inflows into a reservoir system in the monsoon climatic coditions using a multiplicative seasonal ARIMA model based on 25 years of data with logarithmic transformation. The developed model was applied to forecast the monthly inflows for 27 years. The comparison of these forecasted flows with the actual flows reveals that the ARIMA family models are adequate for longterm forecasting of inflows. The parameter uncertainity was also evaluated and found to be minimal thus avoiding the frequent updating of the model for forecasting. The use of the model in evolving optimal cropping patterns and optimal operational policies is also highlighted.     

Reservoir Operation Using a Dynamic Programming Fuzzy Rule–Based Approach

A dynamic programming fuzzy rule–based (DPFRB) model for optimal operation of reservoirs system is presented in this paper. In the first step, a deterministic dynamic programming (DP) model is used to develop the optimal set of inflows, storage volumes, and reservoir releases. These optimal values are then used as inputs to a fuzzy rule–based (FRB) model to establish the general operating policies in the second step. Subsequently, the operating policies are evaluated in a simulation model. During the simulation step, the parameters of the FRB model are optimized after which the algorithm gets back to the second step in a feedback loop to establish the new set of operating rules using the optimized parameters. This iterative approach improves the value of the performance function of the simulation model and continues until the satisfaction of predetermined stopping criteria. This method results in deriving the operating policies, which are robust against the uncertainty of inflows. These policies are derived by using long-term synthetic inflows and an objective function that minimizes its variance. The DPFRB performance is tested and compared to a model, which uses the commonly used multiple regression–based operating rules. Results show that the DPFRB performs well in terms of satisfying the system target performances and computational requirements.     

Reservoir flood control operation model incorporating multiple uncontrolled water flows

This study presents a weighted pre‐emptive goal programming model formulation for coordinated reservoir operation, with easy inclusion of uncontrolled water flows. The model is combined with a multiple water inflows forecasting model, and can be used for real time reservoir operation. Water flow routing from various upstream sites is accounted by with a single compact equation. Integration of controlled and uncontrolled water flows in the optimization model simplifies the operation model, resulting in accurate computation of the downstream water flow. Multiple objectives with water storage and flow variables are used to derive optimal regulation for a reservoir system under flood conditions. For real time operations, the model can be used to determine optimal water release rates for a current period, on the basis of an optimal water release schedule for an operating horizon (T). The model is applied to the flood control operation of reservoirs in the Narmada River Basin (India), with three controlled and three uncontrolled water flows affecting the downstream flow at Hoshangabad. Reservoir water storage and downstream control point flows are zoned, with prioritized objectives used to derive the optimal water release rates. Model applications to the 1999 flood event in the Narmada River Basin with observed and forecasted inflows illustrates that, if water inflows were known through a forecasting technique well in advance, the coordinated operation of the reservoirs could substantially reduce the peak water flows at the control points. The study also indicates that uncontrolled channel flows at the damage site were sufficiently high to cause flooding at the damage site.     

分布式水文模型结合气象预报方法初步探讨

为了尽量消除因流域空间非均一性引起的水文模拟不确定性,采用基于GBHM分布式水文模型以及具有明确物理意义的模型参数,利用三峡区间2011年5~6月期间的气象预报信息,探讨该区域实时洪水预报方法,以及不同预见期的洪水预报精度。结果表明,分布式水文模型与气象预报数据结合,能够较好地模拟该区间的洪水过程。该方法在一定预见期内能够对实时洪水过程进行预报,预报精度很大程度上取决于降水预报的准确性。     

Reservoir Operation Incorporating Hedging Rules and Operational Inflow Forecasts

Reservoir operation incorporating a naïve hedging strategy and operational inflow forecasting is studied in this paper. Gridded precipitation forecasts from climate model, ECHAM4.5, are used as potential predictors for reservoir inflow forecasting. In building a statistical predicting model, principal component analysis (PCA) is used to reduce the dimension of the regression model. Performance evaluation indices, including water supply satisfaction ratio, environmental flow satisfaction ratio, end-of-month storage satisfaction ratio and flood prevention capacity index, are defined. Three scenarios where a naïve hedging operation rule under different set of reservoir inflow are investigated. These are evaluated for a water supply reservoir, Falls Lake Reservoir, at Neuse River in the southeast United State. Reservoir simulation with monthly average inflow serves as a benchmark. The utility of operational inflow forecasts is quantified by the improvements of performance indices. Results show that reservoir operation under perfect inflow forecasting has the highest values for most indices. Compared to climatology, operational inflow forecasts result in higher index values. Among all the performance indices, end-of-month storage satisfaction ratio is the most sensitive index to inflow information. Limitation of this study and further work is also discussed.     

Two-step daily reservoir inflow prediction using ARIMA-machine learning and ensemble models

The reservoirs play a crucial role in the development of civilisation as they facilitate the storage of water for multiple purposes like hydroelectric power generation, flood control, irrigation, and drinking water etc. In order to effectively meet these multiple purposes, the knowledge of the inflow in the reservoir is essential. Apart from the historical data, future prediction of the inflows is also necessary specially in context of climate change. A two-step algorithm for the prediction of reservoir inflow to enable meticulous planning and execution of daily reservoir operation keeping the historical variation of inflow in account has been proposed. The developed algorithm takes into account the patterns in the historic inflow data using the time series analysis along with the variability in the climatic patterns using the different predictors in the machine learning model. The first step uses time series model, ARIMA method to forecast the monthly inflows, which are then used as the targets in the second step for the month-wise daily forecasting of the inflows using the two types of ensemble models, namely, averaging and boosting models in machine learning. The test results show that for both the monthly models and daily models the NRMSE and NMAE values were low for the monsoon periods compared to the non-monsoon periods. The averaging ensemble models were found to perform better than the boosting ensemble models for maximum number of months. The yearly results show an error of less than 5% between actual and predicted values for all the test cases, showing the precision in the developed algorithm. Further, the uncertainty analysis shows that the prediction done using the weighted average of the different inflow scenarios performs better than the prediction against the single inflow scenario.     

水库调度性能风险评价方法研究

运用水库常规调度和优化调度模型,分别确定水库调度策略,从水电站发电和下游生态需水的可靠性、可恢复性、脆弱性和防洪调度权转移风险出发,建立基于综合利用水库调度模型的调度性能风险评价指标体系。以新安江水库调度为例,对1960—2009年旬径流系列的常规与优化调度结果进行风险评价,结果表明:水库优化调度方法比常规调度方法的发电效益、可靠性、可恢复性更高,但其发电脆弱性和防洪调度权转移风险更高。水库调度性能指标全面地评价了不同调度方法对水库调度结果的影响,该研究为综合利用水库在防洪安全、供水安全、生态与环境安全等方面的决策管理提供了一种新的评价思路。     

长洲水利枢纽洪水预报精度的研究

长洲水利枢纽汇流面积大,复杂的来水条件影响了洪水预报的精度。通过对2009年以来的多场洪水进行分析,分析了部分场次洪水预报精度及准确性不足的原因,结合洪水预报系统提出了提高预报精度及准确性的解决方法,对做好长洲水利枢纽水库经济运行调度、水库防洪安全调度有着重要意义。     

Enhancing Inflow Forecasting Model at Aswan High Dam Utilizing Radial Basis Neural Network and Upstream Monitoring Stations Measurements

The Nile River is considered the main life artery for so many African countries especially Egypt. Therefore, it is of the essence to preserve its water and utilize it very efficiently. Developing inflow-forecasting model is considered the technical way to effectively achieve such preservation. The hydrological system of the Nile River under consideration has several dams and barrages that are equipped with control gates. The improvement of these hydraulic structures’ criteria for operation can be assessed if reliable forecasts of inflows to the reservoir are available. Recently, the authors developed a forecasting model for the natural inflow at Aswan High Dam (AHD) based on Artificial Intelligence (AI). This model was developed based on the historical inflow data of the AHD and successfully provided accurate inflow forecasts with error less than 10%. However, having several forecasting models based on different types of data increase the level of confidences of the water resources planners and AHD operators. In this study, two forecasting model approach based on Radial Basis Function Neural Network (RBFNN) method for the natural inflow at AHD utilizing the stream flow data of the monitoring stations upstream the AHD is developed. Natural inflow data collected over the last 30 years at four monitoring stations upstream AHD were used to develop the model and examine its performance. Inclusive data analysis through examining cross-correlation sequences, water traveling time, and physical characteristics of the stream flow data have been developed to help reach the most suitable RBFNN model architecture. The Forecasting Error (FE) value of the error and the distribution of the error are the two statistical performance indices used to evaluate the model accuracy. In addition, comprehensive comparison analysis is carried out to evaluate the performance of the proposed model over those recently developed for forecasting the inflow at AHD. The results of the current study showed that the proposed model improved the forecasting accuracy by 50% for the low inflow season, while keep the forecasting accuracy in the same range for the high inflow season.     

The Utility of Land-Surface Model Simulations to Provide Drought Information in a Water Management Context Using Global and Local Forcing Datasets

Drought diagnosis and forecasting are fundamental issues regarding hydrological management in Spain, where recurrent water scarcity periods are normal. Land-surface models (LSMs) could provide relevant information for water managers on how drought conditions evolve. Here, we explore the usefulness of LSMs driven by atmospheric analyses with different resolutions and accuracies in simulating drought and its propagation to precipitation, soil moisture and streamflow through the system. We perform simulations for the 1980-2014 period with SASER (5 km resolution) and LEAFHYDRO (2.5 km resolution), which are forced by the Spanish SAFRAN dataset (at 5km and 30km resolutions), and the global eartH2Observe datasets at 0.25 degrees (including the MSWEP precipitation dataset). We produce standardized indices for precipitation (SPI), soil moisture (SSMI) and streamflow (SSI). The results show that the model structure uncertainty remains an important issue in current generation large-scale hydrological simulations based on LSMs. This is true for both the SSMI and SSI. The differences between the simulated SSMI and SSI are large, and the propagation scales for drought regarding both soil moisture and streamflow are overly dependent on the model structure. Forcing datasets have an impact on the uncertainty of the results but, in general, this impact is not as large as the uncertainty due to model formulation. Concerning the global products, the precipitation product that includes satellite observations (MSWEP) represents a large improvement compared with the product that does not.

     

实时洪水预报中的雨量短期预报

为提高实时洪水预报的精度，延长实时预报预见期，针对目前预见期内降雨不作处理的状况，提出了对预见期内的雨量进行短期预报的思路．经分析比较发现，采用相邻时段雨量相关关系预报预见期内的雨量，对延长洪水预报预见期有一定的作用．在针对亭下水库流域进行试验时，经对不同预见期的预报结果分析后证明，此方法相对于不预报雨量的方法有更高的精度．且延长了预报的预见期．     

A seasonal forecasting procedure for reservoir inflows in Central Asia

This study develops a procedure for seasonal forecasting river discharge from headwaters above strategically important hydropower plants in Kyrgyzstan and Tajikistan. The El Niño Southern Oscillation, North Atlantic Oscillation, and Indian Ocean Dipole indices were used as inputs. Predictability was evaluated for average summer inflows conditional on the tercile of the preceding winter climate mode. We find that the winter Niño 3.4 index was significantly positively correlated with following summer inflows to Nurek, Andijan, and Toktogul reservoirs during the period 1941–1980. Kruskal–Wallis and Kolmogorov–Smirnov tests show significant differences in the distributions of summer inflows depending on previous winter Niño 3.4 for all three reservoirs. At Nurek, summer inflows were on average 19% greater following a winter El Niño. During 1941–2016, mean summer inflows to Nurek reservoir linked to previous November–December Niño 3.4 achieved a Heidke Hit Proportion of 51–59% (compared with 33% expected by chance). Acceptable predictions of summer inflow volume were made 44% of the time. Higher inflows are explained by a south‐westerly moisture flux that brings above average precipitation to Central Asia during winter El Niño conditions. Our procedure requires limited data, technical or computing resources—all considerations in data sparse, low capacity regions. Given planned developments of other large, headwater impoundments in Central Asia, early outlooks of discharge could contribute to improved dam safety, economic performance, and transboundary water sharing around such projects.     

Value of Medium-range Precipitation Forecasts in Inflow Prediction and Hydropower Optimization

This paper presents an inflow-forecasting model and a Piecewise Stochastic Dynamic Programming model (PSDP) to investigate the value of the Quantitative Precipitation Forecasts (QPFs) comprehensively. Recently medium-range quantitative precipitation forecasts are addressed to improve inflow forecasts accuracy. Revising the Ertan operation, a simple hydrological model is proposed to predict 10-day average inflow into the Ertan dam using GFS-QPFs of 10-day total precipitation during wet season firstly. Results show that the reduction of average absolute errors (ABE) is of the order of 15% and the improvement in other statistics is similar, compared with those from the currently used AR model. Then an improved PSDP is proposed to generate monthly or 10-day operating policies to incorporate forecasts with various lead-times as hydrologic state variables. Finally performance of the PSDP is compared with alternative SDP models to evaluate the value of the GFS-QPFs in hydropower generation. The simulation results demonstrate that including the GFS-QPFs is beneficial to the Ertan reservoir inflow forecasting and hydropower generation dispatch.     

基于数值天气预报模式的流域中期径流预报

研究耦合天气预报模式的径流预报对提高预报预见期及流域防洪减灾具有重要意义。以金溪池潭水库流域为例,通过尺度转换和气象要素联结实现GEM和GFS两种数值天气模式与新安江模型的单向耦合,进行流域水文模拟以及中期径流预报。日径流过程和次洪过程模拟结果发现耦合数值天气预报模式的流域中期径流预报能够较好地预估一段时间内的径流总量,而对洪峰以及洪水过程预报能力稍有不足。预报误差来源有水文模型误差和降水预报误差两种,且降水预报的误差在水文模型中会有放大的效应,这增加了中期径流预报的不确定性。     

雨量预报方法的评价模型

建立的雨量预报方法预报水平评价模型,包括预报方法的准确性评价模型和考虑公众感受的评分模型。准确性评价模型包括雨量预报方法的预报水平的整体评价指标,以及针对不同预报时段、预报等级和站点的分项评价指标,公众感受评分模型基于公众对预报偏差的心理感受特点对预报水平给出评分。通过降雨空间插值方法由网格点的雨量预报值得到观测站点的预报值,应用所建立的指标对给定区域的91个站点41 d的雨量实测值和两种预报方法的预报值进行了分析和评价。     

兰州市1951—2015年降水量变化特征研究

为了解兰州市降水量的变化特征,利用兰州气象站1951—2015年月降水量序列数据,采取线性倾向估计法、滑动平均法、累积距平法、Mann-Kendall非参数检验法及Morlet小波分析等方法分析兰州市近65年来年降水量及各季降水量的变化特征。研究结果表明：近65年来兰州市年降水量总体呈波动式下降趋势,倾向率为-8.20 mm/10年。4季中除冬季降水量呈略微上升趋势外,其余3季降水量均与年降水量变化趋势相同,呈波动式下降趋势。年降水量于1979年发生由增多至减少的突变,各季降水量变化同样存在若干突变点。兰州市降水量变化存在多时间尺度效应,其中年降水量变化的第一、第二及第三主周期分别为46年、32年及14年。目前处于第一主周期降水偏丰阶段,且在未来5~7年内仍将处于偏丰状态。研究成果可为兰州市农业生产活动提供一定的科学依据。     

Multiple Model Predictive Flood Control in Regulated River Systems with Uncertain Inflows

This paper presents a novel approach to real time automatic flood control in a managed river network that is subject to uncertain inflows. The proposed approach uses multiple models to represent inflows ranging from low to high flow. Optimal model selection is achieved in a minimum mean square error sense using a bank of Kalman filters to identify the most likely inflow characteristic. There are no a-priori probabilities assigned to the individual models. Model Predictive Control is used for water level controller design. Our Adaptive Multi Model Predictive Control (AMMPC) method is proposed as an alternative to existing techniques that also use multiple inflow models but with a-priori inflow model probabilities, either weighted or equally likely. The performance of the approach is demonstrated using a simulated river-reservoir model as well as using data collected at the Wivenhoe Dam during the 2011 floods in Queensland, Australia.     

Middle- and Long-Term Streamflow Forecasting and Uncertainty Analysis Using Lasso-DBN-Bootstrap Model

Middle-term and long-term streamflow forecasting is of great significance for water resources planning and management, cascade reservoirs optimal operation, agriculture and hydro-power generation. In this work, a framework was proposed which integrates least absolute shrinkage and selection operator (lasso), DBN and bootstrap to improve the performance and the stability of streamflow forecasting with the lead-time of one month. Lasso helps to screen the appropriate predictors for the DBN model, and the DBN model simulates the complex relationship between the selection predictors and streamflow, and then bootstrap with the DBN model contributes to evaluate the uncertainty. The Three-River Headwaters Region (TRHR) was taken as a case study. The results indicated that lasso-DBN-bootstrap model produced significantly more accurate forecasting results than the other three models and provides reliable information on the forecasting uncertainty, which will be valuable for water resources management and planning.

     

A Stochastic Approach for Seasonal Water-Shortage Probability Forecasting Based on Seasonal Weather Outlook

This study proposed a stochastic approach to forecast water-shortage probabilities for the coming three months in central Taiwan. Monte Carlo method is used to repeat random sampling from the seasonal weather outlook. For each Monte Carlo trial, the monthly rainfalls and monthly mean temperatures for one to three months ahead in eleven upstream catchments of central Taiwan can be obtained. Further, the disaggregation model is used to convert the monthly values into daily rainfall and temperature series. The HBV-based hydrological model uses the daily series to simulate daily inflows for each catchment as the input of system dynamic model for simulating the water budget of water resources system. After all the Monte Carlo trails, the monthly water-shortage probabilities for one to three months ahead can be calculated. The results reveal that the proposed approach can reasonably forecast the water-shortage conditions for one to three months ahead, which are beneficial for regional drought warning and decision support of drought-disaster prevention.