Controlling the Interior of Greenhouses using Elman Recurrent Neural Network

This work proposed Elman recurrent neural network prediction model for predicting greenhouses' interior temperature and relative humidity using four input parameters, viz., wind velocity, carbon dioxide concentration, outside temperature and humidity. Implementing a data acquisition technique we have gathered real time data for each of the parameters. The data was harvestedfrom the greenhouse found inside the premises of the Ethiopian Environmental and Forest Research Institute which is used to develop the model. Generally, 18000 data was collected and processed to develop the model using the Elman Recurrent Neural Network (ERNN). Analysis of several selection criteria has been done to decide the number of neurons in the hidden layer. After evaluating the performance by fixing the hidden neuron using convergence theorem we got a better mean square error of 0.0218 in turn showing98% accuracy of the proposed prediction model.