Predicting Microclimate of a Closed Greenhouse Using Support Vector Machine Regression

Abstract

Abstract A semi-closed greenhouse has been studied in this research. The greenhouse is designed to keep the inside temperature and relative humidity within the optimum growing range throughout the year while maximizing the utilization of solar energy for the plants. The Venlo shaped greenhouse has 4 mm tempered glass as the covering material and three air handling units which control the microclimate. Using air handling units to manage the microclimate of a greenhouse is becoming increasingly popular in arid climates. The plant yield productivity and quality depend on the accurate monitoring and control of the greenhouse microclimate. This study aims to develop a dynamic model that predicts the temperature and relative humidity to improve climate monitoring control accuracy. The data-driven model determines the temperature and relative humidity by incorporating factors affecting the microclimate, such as solar radiation, ambient temperature, relative humidity, fan speed, etc. The available greenhouse data spans from April to June (3 months). Results illustrate that the model predicted accurate values for temperature and relative humidity with an R 2 value of 0.930 and 0.911 and an RMSE value of 0.826 and 1.740, respectively. By accurately predicting the temperature and relative humidity inside the greenhouse, the crop yield can be increased, while minimizing the energy consumption by the air handling units, making greenhouses in an arid climate a more economically feasible option.