Artificial Intelligence for the POME-Centred Water–Energy–Waste Nexus in Palm Oil Production: Towards a Resource-Nexus Digital Twin

This review addresses a key research gap: palm oil milling’s water, energy, Palm Oil Mill Effluent (POME), and solid-residue flows are typically managed independently, while AI applications remain fragmented and rarely support integrated resource-allocation decisions. To address this gap, the review synthesises 109 sources retrieved under a PRISMA-reported protocol and classifies every reported capability as demonstrated in palm oil (A), demonstrated in analogous systems only (B), or proposed as an integration by this review (C), each class being given an operational definition and a validation scale. The term artificial intelligence is used here in the restricted sense of data-driven predictive modelling, model-based optimisation and closed-loop control, and these three functions are kept distinct throughout. The review states the nexus as a formal coupled decision problem with declared variables, objectives, constraints and time scales; specifies the dynamic material flow analysis and in-loop life cycle assessment that existing reviews identify as missing, including reconciliation, functional unit and allocation rules; and specifies how calibrated predictive uncertainty enters resource allocation. These are assembled into an AI-enabled resource-nexus digital twin operated under human oversight. The review finds that current palm-oil AI applications are concentrated in plantation sensing, single-unit treatment modelling, and response-surface optimisation, with validated industrial applications mainly involving anaerobic digestion, membrane separation, and selected biomass-conversion processes. It further identifies limited integration of Material Flow Analysis (MFA) and Life Cycle Assessment (LCA) into real-time operational decisions and inadequate propagation of feedstock and POME variability into resource allocation. Practically, the proposed framework provides a pathway for mills to coordinate water, energy, wastewater and residue management. The framework is a conceptual architecture accompanied by a stated evaluation protocol covering mass-balance closure, predictive accuracy and interval coverage, allocation performance against an incumbent policy, life-cycle consistency, robustness to effluent variability and computational feasibility.