Hourly cooling load prediction of a vehicle in the southern region of Turkey by Artificial Neural Network

In this study, Artificial Neural Networks (ANNs) method for prediction hourly cooling load of a vehicle was implemented. The cooling load of the vehicle was calculated along the cooling season (1 May–30 September) for Antalya, Konya, Mersin, Mugla and Sanliurfa provinces in Turkey. For ANN model, seven neurons determinated as input signals of latitude, longitude, altitude, day of the year, hour of the day, hourly mean ambient air temperature and hourly solar radiation were used for the input layer of the network. One neuron producing an output signal of the hourly cooling load was utilized in the output layer. All data were divided into two categories for training and testing of the ANN. The 80% of the data was reserved to training and the remaining was used for testing of the model. Neuron numbers in the hidden layer from 7 to 40 were tested step by step to find the best matching ANN structure. The obtained results for different numbers of neurons were compared in terms of root mean squared error (RMSE), coefficient of determination (R2) and mean absolute error (MAE). The best matching results for the training and testing were obtained as 8 neurons for the minimum testing RMSE value for the prediction of cooling load by the ANN model on the 23rd day of each month along the cooling season. For the model with 8 neurons RMSE, R2 and MAE (Training/Testing) were found to be 0.0128/0.0259, 0.9959/0.9818 and 78.81/174.71 W/m2, respectively. It is shown that the cooling load of a vehicle can be successfully predicted by means of the ANNs from geographical characteristics and meteorological data.