Dynamic, adaptive and modular Digital Twin framework for resource-efficient Controlled Environment Agriculture.

Area of Science:

• Agricultural Engineering
• Computer Science
• Systems Biology

Background:

• Controlled Environment Agriculture (CEA) is shifting from fixed set-point control to dynamic strategies for efficiency and quality.
• Dynamic control requires real-time monitoring, data integration, and predictive modeling, often facilitated by Digital Twins (DTs).

Purpose of the Study:

• To introduce a novel, open-source Digital Twin (DT) framework for supporting dynamic control strategies in CEA.
• To address challenges in scalability, generalizability, and interoperability of DTs in CEA.

Main Methods:

• Utilized the IoT platform ThingsBoard for scalable, vendor-agnostic data acquisition and management.
• Developed a physics-based modeling backend using Modelica and Functional Mock-up Units (FMUs) with a parameter estimation pipeline for calibration.
• Implemented a Multirate Moving Horizon Estimation (MMHE) state estimator for continuous estimation of unmeasured variables like plant biomass.

Main Results:

• Verified the DT framework's architectural feasibility and virtual modeling capabilities through a lettuce growth simulation in a vertical hydroponic farm.
• Achieved low prediction error (RMSE of 0.221g, NRMSE of 6.62%) for the calibrated growth model on an independent test set.
• Demonstrated the MMHE state estimator's effectiveness in maintaining continuous biomass estimates despite data sparsity and model mismatch.

Conclusions:

• The developed DT framework provides a robust and extensible foundation for physical DT implementations in CEA.
• The framework supports dynamic, data-driven, and energy-aware operations, paving the way for advanced agricultural management.