Purpose: The Manas River in Xinjiang is a typical inland river in arid areas. The Ken Swart Reservoir in the middle reaches has three functions: flood control, irrigation and power generation. However, the current irrigation and power generation dispatching mode of "on-demand discharge" of the reservoir is difficult to achieve the purpose of optimal operation, and may cause adverse effects on the health of the downstream ecosystem of the reservoir. Therefore, this thesis studies the optimal operation of "irrigation-power generation" in Ken Swart Reservoir, in order to improve the comprehensive utilization efficiency of water resources and alleviate the contradiction between supply and demand of water resources in the basin.
Methods: Firstly, the balance analysis of water demand of Ken Swart Reservoir was carried out by statistical method, and the theoretical frequency curve of annual runoff was completed by P-III curve distribution, and the typical hydrological year was determined. Secondly, used the decomposing-prediction-reconstruction method, the monthly runoff prediction model (ASEEMD-BES-ELM) was constructed by combining the Self-adaptation Decomposition Strategy (AS), Ensemble Empirical Mode Decomposition (EEMD), Bald Eagle Search (BES) algorithm, Grid Search method (GS) and Extreme Learning Machine (ELM). Then, combining basin data, hydrological data, reservoir characteristics data, water demand data, reservoir dispatching rules and other data, the construction of Ken Swart reservoir model based on MIKE BASIN was completed, and the model was calibrated and verified in combination with the measured discharge flow. Finally, the Non-dominated Sorting Genetic Algorithm Ⅱ (NSGA-Ⅱ) with elite strategy was used to construct and solve the optimal scheduling model of "irrigation-power generation" of Ken Swart Reservoir, and the non-inferior solution set was obtained. Then, the evaluation index system of multi-objective optimal scheduling scheme was constructed, and the index weight was determined by subjection-objective combination method, and the fuzzy optimization evaluation model was constructed. The dominance of non-inferior solution sets in different study years was calculated, and the optimal solution was input into the MIKE Basin-based Ken Swart Reservoir dispatching model for simulation dispatching.
Results:(1) The average annual inflow of Ken Swart Reservoir is 1.259×109 m³, the water demand is 1.452×109 m³, and the total annual inflow of water in wet, normal and dry years is 1.379×109 m³, 1.199×109 m³ and 1.074×109 m³, respectively.
(2) The Nash-Sutcliffe efficiency coefficient of ASEEMD-BES-ELM reservoir monthly runoff prediction model is 0.971, the mean absolute error is 5.173 m³·s-1, the root-mean-square error is 8.282 m³·s-1, and the mean absolute percentage error is 16.033%, indicating that the model has better prediction effect.
(3) In the rate period (2015~2018), the absolute errors of the measured discharge and the simulated discharge of MIKE BASIN reservoir are 2.4×107 m³, 2.0×107 m³, 9.0×106 m³ and 1.2×107 m³, respectively. The relative errors are 2.13%, 1.18%, 0.68% and 1.04%, respectively. During the validation period (2019~2020), the absolute errors of the measured discharge and MIKE BASIN simulated discharge are 2.8×107 m³ and 8.0×106 m³, respectively, and the relative errors are 2.21% and 0.69%, respectively, indicating that the simulation model had good applicability.
(4) In the study years, the optimized irrigation water shortage in wet, normal, dry and 2030 year is 8.9550×107 m³, 2.9025×108 m³, 3.7565×108 m³ and 2.8787×108 m³, respectively. The power generation is 2.8412×108 kW·h, 2.6979×108 kW·h, 2.4286×108 kW·h and 2.6962×108 kW·h, respectively, and the irrigation water shortage is reduced by 15.98%, 11.68%, 10.25% and 13.18%, respectively, compared with before optimization. Power generation increased by 1.53%, 3.13%, 3.75% and 3.62%, respectively.
Conclusion: The MIKE BAISN has good applicability in the simulation dispatching of Ken Swart Reservoir, and is feasible in the research of multi-objective reservoir optimal dispatching scheme based on NSGA-Ⅱ algorithm, which helps decision makers to better understand and manage water resources and formulate reasonable dispatching strategies. At the same time, the combination of ASEEMD-BES-ELM model can further improve the adaptability and flexibility of reservoir optimal dispatching scheme.
谷歌
火狐
