Neural Models to Predict Temperature and Natural Ventilation in a High Tunnel
Abstract. As a response to the rising demand for local food, high tunnels (HTs) can help small producers become more profitable through crop protection and extension of the growing season. Proper ventilation that responds to changes in outside weather conditions can remove excess heat and humidity inside HTs and lead to better solar energy utilization while maintaining a favorable growth environment. Rather than depending on complex mathematical models, this study investigated an artificial neural network (ANN) for predicting the inside air temperature and ventilation rate of a HT. Energy balance calculations and measured values were compared to the ANN. Results showed that the average air temperature from an array of 15 thermistors inside the HT was predicted more accurately in terms of mean square error (MSE = 1.7°C2) and mean absolute error (MAE = 1.0°C) than a single inside temperature at the center of the HT (MSE = 17.7°C2, MAE = 3.3°C). Relative humidity and wind direction had the least significant impacts on the prediction of inside air temperature, and only four outside weather inputs were found to have significant impacts on the prediction of inside temperature: outside air temperature, door opening level, solar radiation, and wind speed. Moreover, the optimal ANN structure was determined as 29, 25, and 13 neurons in a single hidden layer and 30 neurons in two hidden layers for prediction of inside air temperature, ventilation rate based on measurement, door opening level, and ventilation rate based on modeling, respectively. Keywords: Air temperature control, Artificial neural network, High tunnel, Ventilation.