EVALUATING DIFFERENT MODELS USED FOR PREDICTING THE INDOOR MICROCLIMATIC PARAMETERS OF A GREENHOUSE

Abstract

A robust adaptive model to predict and to control the greenhouse microclimatic condition is pivotal for better crop production and growth. The current research assesses the use of multiple linear regression (MLR), autoregressive integrated moving average (ARIMA), and multi-layered perceptron (MLR) for predicting indoor microclimate greenhouse, located in South Korea. The data were collected from the local weather station and regional weather station data named Ml (local weather station data combined with the regional weather station data), M2 (regional weather station data), and M3 (local weather station data), which were used as the input variables for the prediction. Four dependent variables were predicted (two temperature variables and two humidity variables) by each of the models using M1, M2, and M3 data sets. Performances of the models were evaluated with the coefficient of determination (R-2), the root mean square error (RMSE), the mean square error (MSE), and the mean absolute error (MAE). The simulation results showed that the prediction by the MLP model was highly correlated to the measured data with less RMSE, MSE, and MAE. Besides, seasonal based analysis results reinforce that the MLP performs a better simulation in different environmental conditions. Moreover, the Ml data were propitious for better performance than other data sets, which specifically improves the accuracy of the simulation results for humidity predictions. The present study developed a simple and powerful MLP model to predict the microclimate of the greenhouse, which may integrate into greenhouse controller devices through cloud technology in the future.