Li et al. (2026) Event-driven rotational irrigation scheduling in pressurised tree-structured networks under stochastic head fluctuations and node fault–recovery events
Identification
- Journal: Agricultural Water Management
- Year: 2026
- Date: 2026-11-14
- Authors: Qianxi Li, Li Cao, Yang You, Chenchen Lou, Wenè Wang, Fan Yang
- DOI: 10.1016/j.agwat.2026.110946
Research Groups
- Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas, Ministry of Education, Northwest A&F University, Yangling, 712100, China
- POWERCHINA Northwest Engineering Corporation Limited, Xi'an, Shaanxi, China
Short Summary
This study develops an event-driven framework for rotational irrigation scheduling in pressurized networks, accounting for stochastic head fluctuations and node fault-recovery events. It demonstrates improved irrigation achievement and hydraulic feasibility under disturbances, alongside significant computational efficiency gains through hierarchical caching.
Objective
- To develop and assess an event-driven rotational irrigation scheduling framework that accounts for cumulative water delivery and changing equipment availability under stochastic source-head fluctuations and node fault–recovery events.
- To integrate unfinished-task reconstruction and hydraulically coupled pressure-driven evaluation into multi-objective reoptimization within the remaining operating horizon.
- To quantify the integrated framework’s effects on irrigation delivery and hydraulic feasibility through paired numerical comparisons with static scheduling.
- To compare optimization algorithms, decision information policies, deterministic caching, and tree-based surrogate screening in terms of solution quality, computational cost, and result equivalence.
Study Configuration
- Spatial Scale: A gravity-fed, branched irrigation pipe network serving 57.3 hectares in Xinjiang, China, comprising 71 nodes, 70 pipe segments, and 120 irrigation units. Main pipes have a diameter of 250 mm, and lateral pipes have a diameter of 160 mm.
- Temporal Scale: A 5-day (120-hour) operating horizon, with each irrigation group assigned an operating duration of 4 hours. Source head fluctuations were sampled hourly from uniform distributions (20–30 m for operating heads, 8–20 m for low-pressure cases).
Methodology and Data
- Models used:
- Hydraulic evaluation: Custom-developed hydraulically closed pressure-driven analysis (closed-PDA) model based on path-incidence matrix, Darcy-Weisbach equation, Colebrook-White equation, and a linear pressure-driven discharge relationship.
- Reference hydraulic models: EPANET 2.2 and WNTR 1.5.0.
- Multi-objective optimization algorithms: Nondominated Sorting Genetic Algorithm II (NSGA-II), Strength Pareto Evolutionary Algorithm 2 (SPEA2), Reference Vector Guided Evolutionary Algorithm (RVEA), and Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D).
- Surrogate models: CatBoost, LightGBM, and XGBoost for screening candidate schedules.
- Data sources:
- Case study network parameters: Obtained from field surveys and design records of an existing irrigation system in Xinjiang, China.
- Stochastic conditions: 100 random source-head conditions for hydraulic model verification (U(20, 30) m and U(8, 20) m). 100 independent source-head scenarios (U(20, 30) m) for normal operation. 100 paired scenarios for single-fault/single-recovery and 500 paired scenarios for multiple-fault/multiple-recovery experiments.
- Surrogate model training data: 30,000 samples from the actual network and three synthetic tree-structured networks, with an additional 12,000 external test samples from different synthetic topologies.
Main Results
- Hydraulic Model Verification: The closed-PDA evaluator showed high agreement with EPANET, with nodal pressure-head mean absolute error not exceeding 0.00435 m and mean absolute percentage error (MAPE) for aggregated nodal outflow, pipe flow, and pipe velocity not exceeding 0.0064% across 100 random conditions.
- Improved Irrigation Performance under Disturbances: The event-driven strategy, with additional reoptimization, increased irrigation achievement by 2.09 percentage points (single-fault) and 3.73 percentage points (multiple-fault) compared to static scheduling. It also increased affected-unit target attainment rates by 25.99 percentage points (single-fault) and 15.94 percentage points (multiple-fault), and reduced mean deficits by 3.91 m³ per unit (single-fault) and 6.99 m³ per unit (multiple-fault).
- Enhanced Hydraulic Feasibility: Under normal operation, NSGA-II increased the hydraulically feasible group ratio to 74.7% from 17.7% for random rotation and 0% for the original schedule. Under fault-recovery conditions, dynamic reoptimization increased the mean proportion of hydraulically feasible groups by 1.55–2.96 percentage points.
- Computational Efficiency through Caching: Combined pipe-segment and irrigation-group caching (L1+L2) achieved an 8.56-fold speedup in mean cold-cache runtime (from 24.06 s to 2.81 s). Full three-level caching achieved a 390-fold speedup for exact hot replay. A complete event-triggered rescheduling cycle with prestarted parallel workers took a median of 4.258 s.
- Surrogate Model Limitations: Surrogate-assisted optimization reduced exact hydraulic evaluations by 32.7% but increased total optimization wall-clock time by 47.2% to 81.9% and failed to meet the predefined equivalence criterion for most operational metrics.
Contributions
- Developed an event-driven framework that integrates cumulative water delivery accounting, equipment availability updates, and hydraulically coupled pressure-driven evaluation for dynamic rotational irrigation scheduling.
- Quantified the operational benefits of event-driven reoptimization in improving irrigation achievement, reducing water deficits, and enhancing hydraulic feasibility under stochastic source-head fluctuations and node fault–recovery events.
- Demonstrated that hierarchical deterministic caching significantly accelerates hydraulic evaluations without compromising accuracy, providing a computational basis for real-time event response in pressurized irrigation networks.
- Provided a comparative analysis of multi-objective optimization algorithms, decision information policies, and the trade-offs between deterministic caching and surrogate-assisted optimization for computational efficiency and solution quality.
Funding
- Key Research and Development Program of Shaanxi (Program No. 2026NC-ZDGG-33)
Citation
@article{Li2026Eventdriven,
author = {Li, Qianxi and Cao, Li and You, Yang and Lou, Chenchen and Wang, Wenè and Yang, Fan},
title = {Event-driven rotational irrigation scheduling in pressurised tree-structured networks under stochastic head fluctuations and node fault–recovery events},
journal = {Agricultural Water Management},
year = {2026},
doi = {10.1016/j.agwat.2026.110946},
url = {https://doi.org/10.1016/j.agwat.2026.110946}
}
Original Source: https://doi.org/10.1016/j.agwat.2026.110946