Key Technological Problems Of Flood Water Resources Utilizing In Reservoir Operation

Reservoirs are one of the most efficient measures for the integrated water resources development and management. With the rapid development of social economy and water requirements, the water resources shortage problem is deteriorated and the function of reservoir in terms of floodwater utilization becomes more and more important nowadays in China. This thesis was mainly focused on the determination of the flood control water level (FCWL) in the reservoir operation, which plays a key role in the flood prevention and water resources utilization. According to the Chinese Flood Control Act, the water level of reservoir should be limited below FCWL during flood seasons, so as to offer adequate storage for flood prevention. However, the operation rules based on current FCWL has neglected the real-time flood forecasting information, and gave too much priority to the low probability floods. For floodwater utilization, it's very valuable to explore a more effective methodology to operate the reservoir during flood seasons without enhancing the flood risk. The main contents and innovations of this study were summarized as follows:(1) The literature and research advancement of several concerned topics in the reservoir operation, such as the flood seasonality, the seasonal FCWL, the dynamic FCWL control, the applications of artificial neural network (ANN) and genetic algorithm (GA) in reservoir operation, the simulation and optimization models, and so on, were reviewed extensively both in home and abroad.(2) The flood seasonality that determined the seasonal FCWL was analyzed and discussed. The change point analysis theory was used to divide the flood period into several segmentations through three different sampling methods, which include maximal daily (MD), annual maximal (AM) and the peak-over-threshold (POT). Monte Carlo experiments show that change point analysis with the POT sampling method performs best with the least bias and highest efficiency among of these models. The Three Gorges reservoir (TGR) and the Geheyan reservoir were selected as case studies respectively. It has shown that the change point analysis is an objective method and can obtain rational and accurate flood seasons.(3) The method of risk transformation between different time intervals was conducted. For given design flood standards, the seasonal FCWL isn't unique and should be determined optimally. Based on seasonal design flood routing, an optimal seasonal FCWL model was proposed and developed with the constraints of the sum of seasonal risk less than the annual risk. Case study results indicate that this optimal model can maximize hydropower energy without decreasing design flood standards.(4) With the constraints of the seasonal FCWL risk less than the risk of current annual FCWL, another optimal seasonal FCWL model was also developed based on simulation of the historical data. This optimal model is to maximize the beneficial use and to minimize the flood risk simultaneously. A modified multi-objective GA with mixed integer-programming method was used to optimize the FCWL and refill time for the TGR and yielded a set of Pareto solutions. The fuzzy decision method was adopted to select therelative optimal one among the Pareto solutions. The application results indicate that the optimized model can make an effective tradeoff between the flood control, power generation and navigation, and enhance water resources utilization ratio as well as the economic benefit of the TGR.(5) A reservoir dynamic FCWL control method was developed based on stochastic flood prevention risk operation (FPRO) model. The FPRO model has two constraints, one is that the flood prevention risk should be controlled within the real-time acceptable risk, and the other is that the water level in the end of operation period should be linked with the current FCWL in a stochastic form. Based on the FPRO model, the upper limits of dynamic FCWL can be determined through routing different design floods, in which the reservoir's FCWL can be controlled dynamically without increasing the flood risk both in statistical sense and in deriving different design floods. The TGR was chosen for the case study and the optimal results indicate that the FCWL can change dynamically in the condition of forecasting inflow. It's shown that the proposed model can greatly improve hydropower benefit and decrease flood risk simultaneously.(6) Based on the ANN model, the refill operating rules were derived through three-stage procedures. First, the training data set (reservoir optimal sequences of releases) was searched using the dynamic programming (DP) to solve the deterministic refill operation problem; Second, with the training data set obtained, the ANN model representing the operating rules was trained through the back-propagation (BP) algorithm; Third, the operating rule was refined through simulation-based optimization. That is, by choosing maximization of the hydropower generation as objective function, Simplex method, a nonlinear programming technique, was used to refine the weights of ANN model again. Application results indicate that the proposed model performs well in the TGR refill operation.(7) According to the operation rules of sluice devices, an optimal sluice flood releasing control model was established. Two algorithms, namely DP and GA, were used to solve this integer program problem. It's shown that the GA performs better than DP and the proposed model can reduce the uncertainty of rules of devices during flood control operation.(8) The impacts of climate change and human activity on reservoir operation were also discussed by the method of the time series analysis, which include the trend test, change point detection and Bayesian change point estimation based on general linear model. The case study results of TGR show that the annual hydropower output decrease slowly as the decrease of the annual runoff, and the assurance rate of hydropower generation increase abruptly due to the construction of a number of reservoirs in the upper Yangtze basin. It also shows that the operating rules for each month haven't changed significantly. The requirement length of hydrological data set for the derivation of reservoir operating rules was discussed.