Predicting Desired Fertigation for Rose Using Internet of Things Sensors and Time-Series Model

Greenhouse provides opportunities to have big yield effectively and efficiently. However, many resources are required, such as fertigation, a kind of solution of nutrient. Resources supply is essential to cultivate crops. Inadequate supply will hinder plant growth whereas the surplus results in waste. In this paper, we are especially interested in the fertigation supply. Further, excess fertigation leads to drainage which is difficult to purify and threatens the environment. To address this challenge, we aim to predict the desired amount of fertigation. To achieve this objective, we first establish a prototype to record the climate conditions inside a rose greenhouse using Internet of Things sensors. Simultaneously, the desired fertigation amount is obtained with the help of weight scale and historical data of fertigation supply and drainage. Second, a method is proposed to predict the desired fertigation by taking the sensors’ data as input, with a time-series model. Extensive experimental results suggest the potential of our objective and method. To be specific, our method achieves an average MAE 0.032 in the validation datasets.