A dual-scale multi-reservoir operation method for in-channel and off-channel water demand in the Lijiang River
-
Abstract
Dry-season water replenishment of the Lijiang River is not merely a matter of increasing reservoir releases: the river simultaneously sustains the ecological baseline of Guilin, its landscape and tourism-navigation corridor, and a set of off-channel municipal and irrigation intakes. Existing studies have addressed multi-reservoir optimization, ecological operation and replenishment-flow routing separately, so that medium- to long-term identification of water-supply capacity and short-range formulation of implementable operation plans remain disconnected, while in-channel and off-channel demands are usually evaluated in isolation. This study therefore defines dry-season replenishment as a dual-scale multi-reservoir operation problem addressing in-channel and off-channel water demand, and develops a coupled model that links a daily medium- to long-term water-supply-capacity analysis with an hourly short-range replenishment simulation and operation-plan generation module. The medium- to long-term model performs a daily water balance for each reservoir, separates the baseline release governed by ecological-flow and minimum power-generation rules from active replenishment, and identifies deficits through a sequential water-supply scheme in which the Guilin section is secured first and the residual deficit at the Yangshuo section is then computed after deducting the routed contribution of the Guilin replenishment, thereby avoiding double counting between the two sections. The short-range model first identifies the adjustable storage of each reservoir dynamically, so that reservoirs whose effective storage cannot even cover the baseline release over the operating horizon are excluded from the decision space and participate only as simulated boundaries; it then couples Muskingum flow routing along a table-driven river network, regulation by the Shuangtan hydropower station, and threshold-based off-channel withdrawal, so that only the flow actually attainable at a control section after propagation, efficiency loss, withdrawal deduction and engineering regulation is used to judge compliance. In Experiment 1, a benefit-driven trigger compares the baseline-release scenario with the maximum-permissible-release scenario and starts optimization only when replenishment yields a real gain in demand-satisfaction performance; the optimization then minimizes the total planned release subject to no loss of demand-satisfaction performance relative to the maximum-release reference for individual evaluation targets and an adjacent release-to-storage ratio penalty, and is solved by an improved differential evolution algorithm. Four numerical experiments were carried out for the Qingshitan, Xiaorongjiang, Chuanjiang and Fuzikou reservoirs, with the Si'anjiang reservoir held at dead storage. Experiments 2–4 instead compare predefined allocation paths using two aggregate performance indicators. Under the baseline condition, the recommended plan showed 100.0% average satisfaction for in-channel and off-channel demand at the reported precision with a total planned release of 82.8 m3/s, a 51.9% reduction relative to the maximum permissible release of 171.9 m3/s, whereas baseline release alone (27.0 m3/s) satisfied only 15.7% of the in-channel demand at Guilin. Under compound pressure, with local reach inflows reduced by 20%–40% and demand increased by 20%–30%, the fixed S3 plan degraded progressively to 58.0%, 47.0% and 36.8%; the reported adaptive candidate plans showed 100.0% average satisfaction at the reported precision in all these cases, with total releases of 122.9, 129.7 and 165.7 m3/s, while the water-saving rates of the reported candidate plans decreased from 50.9% under the baseline to 28.5%, 24.5% and 3.6% across the tested pressure scenarios. When the replenishment-efficiency coefficient fell to 0.75, even the maximum permissible release satisfied only 99.7% of the in-channel demand, and the adaptive solution degenerated into the maximum-release solution itself, the two coinciding in both decision variables and demand-satisfaction results; 100.0% satisfaction was restored at 0.80, indicating different demand-satisfaction performance between the two tested efficiency levels. In the predefined allocation-path comparison, the predefined linear-interpolation allocation plan (129.7 m3/s, 24.5% reduction) was not the most water-saving option, yet its release concentration of 0.474 and maximum reservoir drawdown of 3.27 m represented a trade-off relative to the tested priority allocation schemes, whose concentration reached 0.672 or whose drawdown reached 4.27 m, equal to that of the maximum-release scheme. These results indicate that the dual-scale framework can conceptually connect daily risk assessment with short-range plan generation; that the maximum-permissible-release scenario serves as an engineering high-release reference rather than a routine operating rule, and short-range plans should be updated as hydrological and demand conditions change; that replenishment-efficiency changes materially influence demand satisfaction in the tested scenarios; and that the predefined allocation paths reveal conditional trade-offs among total planned release, release-share dispersion and reservoir drawdown. The method provides model support for the four-stage water-resources allocation service of the digital-twin Lijiang River, and a reference for dry-season ecological replenishment of comparable urban and tourism-oriented rivers.
-
-