| Flood disasters are one of the most frequent natural disasters that threaten human production and life.Reservoir forecast flood control operation is the most important non-engineering measure for flood control,disaster reduction and flood resource utilization.It plays a key role in ensuring the steady growth of the national economy,the safety of people’s property and social stability.Therefore,how to control reservoirs to effectively exert the benefits of reservoirs has high research value.Based on the purpose of increasing the efficiency of reservoirs,this paper focuses on two key issues in flood forecast and forecast control operation:the difficulty of flood forecasting in data-sparse basins and the need for further improvement of reservoir forecast control operation.This paper takes Nierji Reservoir as the case study and carries out systematic research to provide a new method for reservoir forecast flood control operation.The main research contents and achievements are as follows:(1)For the basin with complex terrain and sparse data,the accuracy of previous forecasting models is not high.The paper proposes a real-time flood forecasting framework based on Data Based Mechanistic(DBM)model and Kalman filtering method.The framework is designed for the whole process of forecasting including model input,model itself and model output.Firstly,in terms of input,a method for optimizing the weight of rainfall gauges is proposed to solve the problem that the existing weighting methods are not suitable for the watershed area with sparse in situ networks.Secondly,in terms of the model itself and the model output,because of the uncertainty of the data and the process of runoff and confluence,the physics based,DBM model is introduced into flood forecasting,and Kalman filter is coupled to correct the model output in real time.The results show as follows:the accuracy of the forecasting result by the proposed weight optimization method is higher than that by the traditional methods such as Thiessen polygon method and simple averaging method in the data-sparse area.Kalman filtering can reduce the uncertainty of model input and model structure to a certain extent,and significantly improve the accuracy of the prediction,especially in the sub-basins with only rainfall information as the only input.(2)Considering the complicated problem of factor extraction and classification in previous flood classification and identification methods,the paper proposed a flood classification and identification flood forecasting method based on deterministic coefficient,combined with the real-time flood forecasting framework based on DBM and Kalman filter.Firstly,the flood real-time forecasting model based on DBM and Kalman filter is used to classify floods based on the deterministic coefficient which represents the process of flood occurrence and development.Then,the identification method is constructed by combining the forgetting mechanism with the deterministic coefficient.Finally,the method is carried out in Nierji Reservoir Basin.The results show that the method performs well in the sub-basins with obvious non-linear relationship between rainfall and flow,such as Shihuiyao sub-basin,Guli sub-basin,Kehou sub-basin and Jiagedaqi sub-basin.The degree of improvement by the method in the sub-basins with low non-linearity is worse than that of the former sub-basins.(3)The previous reservoir flood control operation evaluated the upstream and downstream flood control safety by the extreme values,which ignored the process indicators such as the process of destroying the upstream and downstream systems and the duration of the damage.However,these process indicators are also important for reservoir operation.The concept of resilience can effectively describe the process of system damage in the fields of ecology,society and engineering.For the first time,this paper introduces the concept of ’system resilience’ into reservoir flood control operation.Firstly,the necessity and feasibility of the application of’system resilience’ in reservoir flood control operation are analyzed.Then,the’system resilience’ of the downstream protection point by flood is defined and quantified.Finally,flood control optimization operational model is established with three objectives including the maximum water level of the upstream reservoir,the maximum discharge of the downstream flood protection point and the resilience of the downstream protection point,and it has been carried out by the NSGA-Ⅱ method.Results show as follows:1)The diversity of schemes is increased with the addition of the resilience objective in the optimization,which not only provides the decision makers with schemes that favors upstream security or downstream security,but also provides schemes that favors the resilience of the downstream protection points.2)The final recommendation scheme CB has been provided for decision makers.In the case of rainfall mode 1,the upstream risk of the final recommended scheme CB is consistent with the conventional scheme(the maximum upstream water level is 217.53m),and the downstream risk is lower than the conventional scheme(the average CB flow is 10200 m3/s,while the conventional scheme is 10340 m3/s),the resilience of the downstream protection point is 86.82%,which is superior to the conventional 86.18%.In the case of rainfall mode 2,the upstream risk and the downstream risk of the scheme CB are consistent with the conventional one(the maximum water level average is 218.15m,the average of downstream maximum flow is 9990m3/s),and t the resilience of the downstream protection point is 90.04%,which is higher than the conventional(89.31%).(4)Previous researches mainly focused on tending to increase the rate of flood resource utilization in reservoir forecast flood control operation by raising the limit water level.Few studies tend to increase the efficiency of flood forecast control operation by reducing the limit water level.This paper proposed a flood forecast control operational method considering the forecasting error to reduce the limit water level,which ensures that the incoming water can meet the recharge of the reservoir to the original limit water level.Considering the forecast error distribution and the different preferences of decision makers,the binary comparison method and fuzzy optimization model are used to evaluate the forecast scheduling scheme.The results show that the upstream risk,the downstream risk and the resilience of the downstream protection by the rule recommended in chapter 6 under different relative errors(-22%,1.3%and 19%)are better than the best rule recommended in Chapter 5 without considering the forecast except for the upstream risk under relative errors of-22%. |
