Predicting energy consumption using Artificial Neural Network (ANN): A case study in a cold store in the North Island of New Zealand

Abstract

This study was developed to predict energy usage, based on the amount of fruit stored and environmental factors, using an artificial neural network model (ANN) in a New Zealand cold store. In this study, after investigating several transfer functions and structures, an ANN model was developed to reduce margins of error in energy auditing projects in a cold store. The final ANN model developed was based on weekly numbers avocado and kiwifruit bins and outside temperatures. The comparison between different models demonstrated the amount of stored fruits have more sensitivity than the outside temperatures in cold stores. Comparing actual and predicted energy usage in the studied cold store showed that the ANN model could be fitted to energy usage data and accounted for around 95% of the variance.