.png){kind=icon}
.png){kind=icon}
Return
Harnessing machine learning for water energy food nexus sustainability: developing a surrogate to multi-objective optimization
.png){kind=icon}
PRE
AI
F
M
A
M
DOI：10.1016/j.compchemeng.2025.109526.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Integrated management of water, energy, and food resources is critical for achieving sustainability under rising environmental and demographic pressures, yet existing approaches either lack computational efficiency for real-time decision support or fail to capture the full complexity of sectoral interdependencies. This study presents a Water Energy Food Nexus Machine Learning based surrogate Model (WEFN-MLM). The key innovation lies in training a Random Forest algorithm on comprehensive multi-objective optimization outputs from thousands of diverse scenarios, enabling the model to learn complex nonlinear interdependencies and resource trade-offs without requiring explicit mathematical formulation of system relationships. A high-resolution Multi-Objective Optimization WEFN model (MOO-WEFN) is used to generate the training dataset, incorporating constraints for resource availability, caloric requirements, and environmental thresholds. The trained model demonstrates high predictive accuracy, with most output variables achieving R² values above 0.90 and cosine similarity scores near 1.0. Normalized absolute error analysis reveals strong performance consistency across system-level metrics, with select deviations in sector-specific outputs, particularly those highly sensitive to scenario dynamics or underrepresented in the training space. Compared to traditional optimization, the surrogate model achieves up to a 300,000-fold reduction in computation time. The surrogate model is validated using a randomly generated test set of scenarios that enables direct comparison between surrogate predictions and optimization results. The results highlight the model’s effectiveness for high-resolution nexus analysis and scenario exploration, while also acknowledging trade-offs between speed and precision. Findings underscore the importance of diverse training scenarios, careful application boundaries, and integration with policy processes to support resilient resource planning.
Journal
C
IF:
3.9
Papers:
8.1K
Citations:
1.7W
Organization
Cited Papers