Air Source Heat Pump Performance Evaluation Based on Experimental measurements and Neural Network

This paper will present a performance evaluation with neural network architecture which simulates the electrical power requirements of an air source heat pump used in winter commercial building heating. Methodology selected evaluates four types of known neural network architectures to select the proper model which fit better the heat pump. Using experimental measurements for training session of neural network, the weights established and with neural network architecture biases, is driven a yearly energy consumption simulation for five cities of Romania. Results indicate the appropriate total electrical energy consumption of heat pump and seasonal energy efficiency ratio obtained.