Complex water conveyance systems often involve mixed flow conditions comprising pressurized pipe networks and free-surface open channels. Every single model has its limitations in long-distance and complex water transfer projects; coupling different models can leverage their respective advantages. For this reason, this study proposes an integrated modeling framework that couples the Storm Water Management Model (SWMM) and Environmental Protection Agency Network Evaluation Tool (EPANET) models to simulate water quality in pressurized–unpressurized coupled systems. Taking the typical Yin Chao Ji Liao water diversion project as a case study, total nitrogen (TN) and total dissolved solids (TDS) were selected as representative pollutants. A total of 32 simulation scenarios were systematically evaluated across four concentration levels at four distinct monitoring points. During this synthesis, key emergency response indicators, specifically T0 (optimal gate-opening time) and T2 (drainage operation duration), were quantified. The results indicate that pollutant dispersion in the unpressurised tunnel section is primarily governed by pollutant loading, whereas in the pressurised pipeline section, it is controlled by flow velocity and pressure differentials. High-concentration pollution in the tunnel section allows for a relatively longer emergency response time (T0 = 453 min). In contrast, due to rapid transport in the pipeline section, emergency operations must be completed swiftly (≤30 min). This coupled approach effectively addresses the challenge of water quality simulation in cross-regime conveyance systems. Future research will focus on integrating real-time sensor monitoring data with this coupled model to further optimize automated early warning protocols for long-distance diversion projects.

Recurrent neural networks (RNNs) are widely used in reservoir operation because they can represent temporal dependencies, but their performance depends strongly on hyperparameter selection. This study compares eight evolutionary strategies—genetic algorithm (GA), particle swarm optimization (PSO), simulated annealing (SA), differential evolution (DE), ant colony optimization (ACO), artificial electric field algorithm (AEFA), slime mould algorithm (SMA), and white shark optimization (WSO)—for optimizing the hidden-node number and initial learning rate of long short-term memory (LSTM) and gated recurrent unit (GRU) models. Under common experimental settings, ACO required the shortest computation time in all reported dataset–architecture blocks, whereas the most accurate strategy depended on the dataset, architecture, and metric. For the total dataset, ACO–LSTM and AEFA–GRU produced the lowest reported MAEs (609.876 and 582.942 m³/s, respectively); WSO–LSTM was best for flood-season MAE (948.326 m³/s), and GA–GRU was best for non-flood-season MAE (516.088 m³/s). An independent chronological holdout experiment further verifies the search boundaries and validation-fitness procedure and shows that the fitness response also varies with hydrological period and RNN architecture. The results support objective- and dataset-specific optimizer selection, subject to the limitations of a single-reservoir case study and a restricted two-variable search.