Multi-Parametric Tuning of Dynamic Air Conditioning Models Using Experimental Data

Parameter tuning of air conditioning system models is a critical step in constructing accurate dynamic models for use in control design and optimization. However, traditional manual or simulation-based methods of parameter tuning are tedious, time-consuming, or simply infeasible due to the large number of parameters. In this paper, an approach to tune multiple parameters of HVAC&R systems is proposed and shown to be effective in practice. An accurate and computationally efficient model is derived, and a wavelet decomposition approach is adopted in formulating the cost function that seeks to minimize the error between the predicted and measured data. Wavelets are advantageous in handling the multi-domain signals and capturing large and small dynamic features. In order to reduce the optimization time, a hybrid parameter tuning method is proposed. This method first uses a stochastic global tuning method (genetic algorithm) to find a set of estimated parameters, and these parameters serve as the initial values of a gradient search method. The results show that the proposed method can effectively tune many simultaneous parameters using large data sets.