Modeling greenhouse humidity by means of NNARMAX and principal component analysis

This paper is concerned with how far a combination of auto regressive moving average model with external input and neural network (NNARMAX) can ameliorate the modeling performance for the inside humidity of an unheated, naturally ventilated greenhouse in autumn under east-central China conditions. The environmental factors influencing the inside humidity, including outside air temperature and humidity, wind speed and direction, solar radiation, inside air temperature, are all collected as data samples. First, through grey relational analysis, the wind speed which has the least correlation degree with the output is omitted. Second, through principal component analysis (PCA) of the other input data samples, 3 principal components are extracted which are taken as the input of NNARMAX model. At last, a comparison of the modeling performances is made between the NNARMAX model and the other 4 models, and the NNARMAX model outperforms the other 4 models as indicated by the simulation results and the goodness of fit.