|Title: ||Flood Control Combining Optimization Techniques with Hydrologic-Hydraulic Modelling|
|Other Titles: ||Overstromingsbeheersing d.m.v. optimalisatie en hydrologische-hydraulische modellering|
|Authors: ||Chiang, Po-Kuan; S0183597|
|Issue Date: ||1-Apr-2015 |
|Abstract: ||Flood is one of the natural disasters. It frequently causes vast amount of economic losses and other consequences (social, ecological), even losses of human lives. Due to these severe injuries, how to perform an effective flood control is always a huge challenge for governments and water authorities. In Belgium, the river Demer has its history to be viewed as a definite case for discussing flood problems. In the past, flooding along this river could not be prevented during several periods of heavy rainfall events. In order to alleviate such flood disasters, the Flemish Environment Agency (Vlaamse Milieumaatschappij, VMM) installed several hydraulic structures (e.g. movable gated weirs and flood reservoirs) and formulated operating rules to regulate them. After that, the VMM water authority is able to reduce or limit the flood hazard. It can, however, not fully avoid the most extreme flood events. A complete real-time flood control scheme has the ability to integrate weather prediction, flood simulation and optimization models. Due to these characteristics, it is expected to be more advantageous than current logical operating rules (three-position controller) to handle severe rainfall events. Therefore, this dissertation investigates the applicability of an advanced control strategy by means of Model Predictive Control (MPC), and discusses its potential performance for the flood mitigation of the Demer river system.|
Developing a suitable river flood model is a must for this research. Firstly, a detailed full hydrodynamic model, InfoWorks-RS model, was applied to simulate the detailed physically-based hydrodynamic processes for open channels, floodplains, embankments and hydraulic structures of the river networks. However, one of the main problems to date is that such detailed full hydrodynamic model has very long computational times. Especially the model-supported real-time flood control requires effective and efficient hydraulic computations as large numbers of iterations are to be executed in optimization procedures. A conceptual model can resolve this problem. By means of the simulation results of the detailed full hydrodynamic model, an identification and calibration procedure was developed in this study for the purpose of having a conceptual model built up and calibrated based on a limited number of simulations with a more detailed full hydrodynamic model. The conceptual river model aimed to concisely describe the system dynamics and responses during different flow conditions with high accuracy (consistent and close to the detailed full hydrodynamic InfoWorks-RS model) and at high speed. It is found that the conceptual river model is capable of accomplishing simulation of historical flood events with high statistical performances (Nash-Sutcliffe Efficiency values higher than 0.90) within few seconds. It is much faster than the detailed full hydrodynamic model, which enables the conceptual model to be applied for real-time flood control.
The upstream and lateral inflow discharges of both the detailed hydrodynamic model are calculated by a rainfall-runoff model, more specifically the Probability Distributed Model (PDM). PDM was implemented for the different subcatchments of the river Demer basin.
For the final goal of this research, improved gate-operation policies (gate crest levels as optimization variables) are searched for flood events through large number of iterations, which are run by an optimization model. The approach was principally to develop a procedure for the real-time optimal control of the hydraulic structures. This required comprehending the whole procedure for such control, making use of the technique of MPC and learning how to link it with the conceptual model. In addition to the MPC algorithm, an adequate MPC-based optimization technique had to be selected, which can cope with the strongly dynamic and nonlinear dynamics of the process model in real time at each sampling instant (e.g. 5 minutes). A Genetic Algorithm (GA) was selected as optimization method for this study, and combined with MPC that enables to cope with the nonlinear dynamics. The MPCGA model searches for better control actions by minimizing the cost function while at the same time avoiding violation of the defined constraints. The optimization results testify that MPCGA is capable of improving of the current regulation strategy that is based on fixed regulation rules and three-point controllers.
In summary, this research concludes that real-time flood control can be carried out successfully by connecting the conceptual hydrologic-hydraulic model with the nonlinear MPC optimized by a GA. A concrete procedure for real-time flood control is built up and demonstrated/evaluated for the river Demer case by comparing with the current regulation strategy. It is shown for the extreme September 1998 flood that the better control policies found by the MPCGA controller succeed to keep all twenty selected/monitored water levels beneath their corresponding flood levels, and solve the flood damages that occur when the three-position controller is in place at 3 locations. This study hence has shown that the advanced real-time control procedure developed in this research succeeds to mitigate negative/harmful flood conditions along a river system.
|Publication status: ||published|
|KU Leuven publication type: ||TH|
|Appears in Collections:||Hydraulics Section|