머신러닝을 활용한 IoT 작물재배시스템 내부온도 예측에 관한 연구

Abstract

This study focuses on efficient production, away from research that focused on automating cultivation facilities and increasing convenience. In this study, we discuss a model that predicts the internal temperature of the IoT crop cultivation system using machine learning, which derives the internal temperature prediction model of the crop cultivation system and ensures crop growth under optimal environmental conditions. In this study, weather data were collected through external temperature data, internal temperature data, controller temperature data measured in crop farming systems and Agricultural Weather Information Service API before deriving the optimal prediction model. Afterwards, the collected data was refined to the appropriate form for the analysis. And preprocess of the data was carried out with process of parsing the information needed for the analysis and missing value. One of the learning methods of machine learning, Supervisited Learning, is used to predict optimal internal temperatures by utilizing pre-processing data, and multi-layer perceptron algorithm is utilized as a neural network model to efficiently derive predictive models. In addition, two predictive models were classified according to the input variables. The effect of cultivation materials on the system, which is one of factors that affect internal temperatures data, is Formulated. And by comparing the results of predictive models that applied this formula with those of predictive models that did not apply the expression, we derive a model that predicts patterns similar to the temperature within the real system.