Optimal Energy Management in Indoor Farming Using Lighting Flexibility and Intelligent Model Predictive Control

TL;DR

This work addresses the high energy burden of indoor farming by leveraging plants' tolerance to brief light interruptions to design cost-aware, 24-hour lighting recipes. It couples model predictive control with transformer-based forecasts of solar radiation and electricity prices to generate optimal lighting schedules that meet physiological constraints such as daily light integral targets. The approach embeds plant-health boundaries into the optimization and incorporates other greenhouse loads to maximize peak-demand reductions. Case studies and simulations on lettuce in a large greenhouse demonstrate substantial energy, cost, and peak-load savings, highlighting the method's potential for sustainable, economically feasible indoor farming. The results suggest practical pathways for integrating intelligent lighting with grid services and other distributed energy resources.

Abstract

Indoor farming enables year-round food production but its reliance on artificial lighting significantly increases energy consumption, peak load charges, and energy costs for growers. Recent studies indicate that plants are able to tolerate interruptions in light, enabling the design of 24-hour lighting schedules (or "recipes") with strategic light modulation in alignment with day-ahead pricing. Thus, we propose an optimal lighting control strategy for indoor farming that modulates light intensity and photoperiod to reduce energy costs. The control strategy is implemented within a model predictive control framework and augmented with transformer-based neural networks to forecast 24-hour ahead solar radiation and electricity prices to improve energy cost reduction. The control strategy is informed by real-world experimentation on lettuce crops to discover minimum light exposure and appropriate dark-light intervals, which are mathematically formulated as constraints to maintain plant health. Simulations for a one-hectare greenhouse, based on real electricity market data from Ontario, demonstrate an annual cost reduction of $318,400 (20.9%), a peak load decrease of 1.6 MW (33.32%), and total energy savings of 1890 MWh (20.2%) against a baseline recipe. These findings highlight the potential of intelligent lighting control to improve the sustainability and economic feasibility of indoor farming.

Figures (9)

• Figure 1: Schematic representation of a greenhouse.
• Figure 2: Fluorescence response of three leaves under different conditions.
• Figure 3: High-level block diagram of the system modules and their interactions.
• Figure 4: Overall procedure for online prediction of electricity price and solar radiation.
• Figure 5: Trial recipes for plant growth experiments. Yellow cells are light intervals while grey cells are dark intervals.