Data-driven economic dispatch towards operational management of distributed energy resources for grid-connected water–energy microgrids

This paper addresses the critical need for economic dispatch optimization in integrated water-microgrid systems by introducing innovative data-driven techniques to overcome computational challenges. Leveraging machine learning models, we achieved rapid resolution of objective functions and constraints in water–energy system dispatch optimization. It effectively tackles the computational challenges associated with dynamic economic dispatch problems in these nonlinear, interdependent, complex systems, showcasing remarkable improvements with computational speed enhancements of over 1 0 4 times compared to conventional numerical optimization methods. The accuracy of the machine-learning algorithms is demonstrated through the superior dispatch efficiency of the developed data-driven models for solving the economic dispatch problem, with approximately 99% accuracy for all dispatch predictions for the gradient boosting model. The modeling errors, however, are more pronounced with variations in water demand and grid connection profiles. Furthermore, the SHAP explainable artificial intelligence (XAI) technique is applied to interpret the model predictions and input–output relationships. Notably, the presented work enables the incorporation of offline data-driven systems for in-depth analysis of the impacts of dispatch decisions, thereby enhancing system robustness. This innovative approach holds significant promise for achieving cost-effective and sustainable power generation in line with global zero-carbon emission targets. • A data-driven economic dispatch approach is proposed for water–energy systems. • Data-driven machine learning methods optimized by the particle swarm optimization. • Data-driven economic dispatch errors optimized to NRMSE less than 0.01. • Computational gain of the proposed approach compared with the traditional approach.