Project by Branislav Daráš: Advanced Model Predictive Control Design For a Smart Greenhouse

In the realm of modern agriculture, ensuring optimal growing conditions is crucial for maximizing yield and quality. This challenge is particularly pronounced in greenhouse environments where factors such as temperature, humidity, and CO2 levels must be meticulously controlled. Traditional control systems like Proportional-Integral-Derivative (PID) controllers have been the go-to for many years due to their simplicity. However, as systems become more complex, a more sophisticated approach is necessary. This is where Model Predictive Control (MPC) comes into play.

VESNA (an acronym for Versatile Simulator for Near-zero Emissions Agriculture) is an intelligent ecological greenhouse developed by young researchers and students at the Institute of Informatics, Automation, and Mathematics, FCHPT STU in Bratislava. The intelligent greenhouse integrates the latest smart technologies with an ecological approach to modern food production. The VESNA project combines these two directions into a unique ecosystem. This ecosystem includes the greenhouse itself, various sensors, actuators, and a communication interface that ensures data transmission from the device.

Key actuators in this project include a heater and humidifier located at the bottom of the smart greenhouse. Essential sensors measure temperature and humidity. Following the introduction of the temperature control loop in my bachelor's thesis, the next step was to integrate a humidity control loop into the system, using the humidifier as the control input and humidity as the control output. To achieve this, I applied a 25% step change to the humidifier to observe its step response.

I then normalized the data to determine the system gain.

To simplify, a first-order transfer function was used. The time constant was calculated as the time difference between when the step response reached 63.2% of the gain and the time the step occurred, resulting in a first-order transfer function:

With the control loop identified, a test was conducted. The results showed that the control stabilized quickly, with humidity reaching its setpoint, indicating successful control.

Having identified two control loops, it was time to introduce MIMO (multiple-input multiple-output) control to VESNA for the first time. The results demonstrated the interaction between temperature and humidity, with rising humidity lowering the temperature and vice versa.

However, humidity does not reach its setpoint,  suggesting room for improvement in the weight matrices.

On this Figure we can see that even though the temperature is within delta-neighbourhood, we send zero input to the heater and it doesnt response. Although the temperature was within acceptable limits, the heater did not respond appropriately due to residual warmth after being turned off.

To enhance control performance, several steps can be taken:

  1. Tuning the weight matrices of optimization problems objective function.
  2. Identify how heater affects the humidity and humidiser temperature, and see if MPC can resolve this problem by itself.
  3. Consider lighting and fans as other input for MIMO control, at least fans might be able to resolve the heaters residual warmth issue.

In summary, I successfully introduced a new control loop after identifying and testing the process. This was followed by implementing the first MIMO control for the VESNA greenhouse. We discussed the control quality and potential improvements for the future, and I am eager to continue working on this project.