Study on neural network modeling of greenhouse environment based on partial least squares algorithm

The greenhouse environment model is an important basis to make control strategies and optimize control method. If correlation among multi-variables of greenhouse environmental model exists, the accuracy of the model decreases. In this paper, utilize partial least squares (PLS) to extract principal components of data, adopt radial basis function neural network (RBFNN) to construct control model of greenhouse environment in northern region of china. And this model is compared with the Orthogonal Least Square (OLS) algorithm in performance. The results indicate that RBF network model of the greenhouse environment based on PLS has smaller network structure, and is superior to OLS algorithm in the approximation ability and generalization ability. The model has laid a good foundation for designing control scheme and structure of greenhouse environment.