Modelling of desiccant wheels using radial basis neural networks

Abstract The desiccant wheel based air conditioning systems are superior over conventional vapour compression systems in terms of the efficiency and power consumption. Researchers are consistently focusing on the various aspects of the desiccant wheel such as the architecture of the wheel, effect of operating parameters on the performance of the wheel and the overall configuration of air-conditioning systems. For the simulations of the air-conditioning systems it is essential to have an accurately performing model of the desiccant wheel. Thus, for solving this purpose, the radial basis neural network is applied to forecast the performance of a desiccant wheel data. The input parameters are the inlet process air humidity, temperature, regeneration temperature, number of rotations per hour of the desiccant wheel and the ratio of the volume flow rates of the process air and the regeneration air. The output parameters are the temperature and humidity of the process air outlet stream. The model is able to predict the original data within ± 5% accuracy and the r-square value is obtained to be 0.9963. A parametric study presenting the variations of the process air outlet temperature and humidity with respect to the various input parameters is also included in the article.