Digital twin of an urban-integrated hydroponic farm

Melanie Jans-Singh, Kathryn Leeming, Ruchi Choudhary, Mark GirolamiData-Centric 2020 · DOI

Abstract

This paper presents the development process of a digital twin of a unique hydroponic underground farm in London, Growing Underground (GU). Growing 12x more per unit area than traditional greenhouse farming in the UK, the farm also consumes 4x more energy per unit area. Key to the ongoing operational success of this farm and similar enterprises is finding ways to minimize the energy use while maximizing crop growth by maintaining optimal growing conditions. As such, it belongs to the class of Controlled Environment Agriculture, where indoor environments are carefully controlled to maximize crop growth by using artificial lighting and smart heating, ventilation, and air conditioning systems. We tracked changing environmental conditions and crop growth across 89 different variables, through a wireless sensor network and unstructured manual records, and combined all the data into a database. We show how the digital twin can provide enhanced outputs for a bespoke site like GU, by creating inferred data fields, and show the limitations of data collection in a commercial environment. For example, we find that lighting is the dominant environmental factor for temperature and thus crop growth in this farm, and that the effects of external temperature and ventilation are confounded. We combine information learned from historical data interpretation to create a bespoke temperature forecasting model (root mean squared error < 1.3°C), using a dynamic linear model with a data-centric lighting component. Finally, we present how the forecasting model can be integrated into the digital twin to provide feedback to the farmers for decision-making assistance.

Introduction: The Challenge of Bespoke Urban Farming

Controlled Environment Agriculture (CEA) is a cornerstone of future urban food security, yet it remains highly resource-intensive. "Growing Underground" (GU) is a prime example: situated 33 meters beneath London in repurposed WWII air raid shelters, it produces 12 times the yield per unit area of traditional UK greenhouses but consumes four times the energy.

Location and layout of the Growing Underground farm in London.

Unlike standard greenhouses, GU's environment is buffered from external weather but constrained by limited ventilation and heavy reliance on artificial lighting as the primary heat source. Standard agricultural models often fail in such bespoke, non-standard environments. This research addresses the need for a digital twin capable of synchronizing disparate data, decoupling confounded environmental variables, and providing actionable predictive feedback to optimize energy use and crop health.

Methodology: A Three-Stage Digital Twin Framework

The researchers developed a digital twin following a structured three-stage framework: Data Creation, Data Analysis, and Data Modeling.

The three-stage framework of the digital twin: Observations, Processing, and Output.
  1. Data Creation: A Wireless Sensor Network (WSN) monitored 89 variables across 25 sensors. This high-frequency data was integrated with "unstructured" manual records, such as ventilation dial settings and crop yield logs.
  2. The "Light Proxy": In the absence of dedicated light sensors, the team used energy meter data as a proxy. By analyzing power spikes (exceeding 30 kW), they inferred lighting status to model its significant thermal impact.
  3. Forecasting Models: The study compared a standard SARIMA (Static Seasonal) model with a customized Dynamic Linear Model (DLM). The DLM incorporated a "data-centric lighting component" to account for irregular operational shifts, such as extended lighting for tours or maintenance.

Key Insights and Environmental Dynamics

  • Lighting as "Weather": In the underground tunnels, LED lights are the dominant environmental driver. Activating the lights can raise internal temperatures by over 5°C, particularly when external temperatures are low.
  • Confounding Variables: Ventilation and external temperature effects are often confounded in practice, as operators typically increase ventilation on warmer days. The digital twin helps disentangle these drivers to understand true causal relationships.
  • Spatial Micro-Microclimates: Analysis revealed significant spatial variation. For instance, salad rocket thrived in the warmer "back" sections of the tunnel. The digital twin enables operators to map crops to their ideal "thermal niches."
  • Anomaly Detection: The model serves as a diagnostic tool. Significant deviations from forecasted temperatures often signal manual errors, such as a door left open or an incorrect lighting schedule.

Results and Operational Impact

The bespoke DLM achieved a Root Mean Squared Error (RMSE) of 0.72°C for 4-pm forecasts, significantly outperforming naive models. With a 96–98% coverage rate, the model's predictions were highly reliable. Operationally, the system flagged the 6% of instances where manual overrides deviated from the intended 18-hour photoperiod, providing a critical feedback loop for management.

Limitations and Future Directions

While the framework is robust, it currently relies on manual data entry for ventilation and crop logs, creating potential bottlenecks. Furthermore, the thermal coefficients are site-specific to the GU tunnels. Future work aims to move from decision support to automated control, where the digital twin directly adjusts HVAC and lighting systems in real-time to maintain optimal conditions autonomously.

Summary updated

Abstract

Abstract This paper presents the development process of a digital twin of a unique hydroponic underground farm in London, Growing Underground (GU). Growing 12x more per unit area than traditional greenhouse farming in the UK, the farm also consumes 4x more energy per unit area. Key to the ongoing operational success of this farm and similar enterprises is finding ways to minimize the energy use while maximizing crop growth by maintaining optimal growing conditions. As such, it belongs to the class of Controlled Environment Agriculture, where indoor environments are carefully controlled to maximize crop growth by using artificial lighting and smart heating, ventilation, and air conditioning systems. We tracked changing environmental conditions and crop growth across 89 different variables, through a wireless sensor network and unstructured manual records, and combined all the data into a database. We show how the digital twin can provide enhanced outputs for a bespoke site like GU, by creating inferred data fields, and show the limitations of data collection in a commercial environment. For example, we find that lighting is the dominant environmental factor for temperature and thus crop growth in this farm, and that the effects of external temperature and ventilation are confounded. We combine information learned from historical data interpretation to create a bespoke temperature forecasting model (root mean squared error &lt; 1.3°C), using a dynamic linear model with a data-centric lighting component. Finally, we present how the forecasting model can be integrated into the digital twin to provide feedback to the farmers for decision-making assistance.
BibTeX
@inproceedings{janssingh2020digital,
  author = {Melanie Jans-Singh and Kathryn Leeming and Ruchi Choudhary and Mark Girolami},
  title = {{Digital twin of an urban-integrated hydroponic farm}},
  booktitle = {Data-Centric},
  year = {2020},
  doi = {10.1017/dce.2020.21},
}
Figure