Neuro modelling and control of an automotive air conditioning system

Abstract

Air Conditioning system is very important in our world nowadays. Air conditioning system provide the temperature comfort not only in building but also inside car. The objective of this study is to identify the dynamic model for an Automotive Air Conditioning (AAC) system by using Recursive Least Square (Traditional Method) and Artificial Neural Network (Intelligent Method). Beside that, both model are also to design a self tuning PID controller for variable speed of AAC compressor system by using Matlab-SIMULINK software. In the project, system identification techniques (Traditional Method) namely Recursive Least Square (RLS) and Artificial Neural Network (Intelligent Method) was use do estimate dynamic model of AAC system. The input and output of the data used to estimate the dynamic model of AAC system were obtained from the experimental. The system identification techniques were used with (ARX) model structure and the Intelligent techniques were used with (NARX) model structure. The validation of the RLS and ANN models were based on the Mean Square Error (MSE). Comparative performance of the RLS and ANN model in this project were discussed. From the discussion, it found that RLS model with order number of 2 give a better performance with minimum value of MSE. For the ANN model, number of neuron in hidden layer of 10 and delayed input/output data of 6 have been optimized as the best model with the minimum value of MSE. After that, the best model of RLS and ANN are used with PID controllers. The PID controllers are tuned using Auto tuned and heuristic tuning method. From this research, it found that, RLS model using PID controller tuned using Heuristic tuning method had shown the best performance, but on other hand, for the ANN model it was shown that Auto tuned PID have a better performance in comparison to Heuristic tuned PID.