Improved MIMO system identification and control using genetic algorithms

Many industrial control systems are of the Multi-Input-Multi-Output (MIMO) type which require advanced control solutions based around multivariable system formulations. However, conventional multiloop Single-Input-Single-Output (SISO) control systems are still used in industry because the formulation of MIMO control schemes are not as straightforward. Hence, strong interaction between loops often significantly limit their effectiveness. Recent trends in control system theory have seen the emergence of a new breed of sophisticated multivariable, model based predictive, design techniques such as Dynamic Matrix Control (DMC), Internal Model Control (IMC), Generalized Predictive Control (GPC) and Proportional-Integral-Plus (PIP). With such methods comes a tacit requirement for modelling techniques which can provide efficient and accurate multivariable models, particularly since the quality of the control achievable with these approaches is directly related to the quality of the underlying model. In this paper, we demonstrate how Genetic Algorithms (GAs) can help the control engineer to solve some of the problems which arise when designing MIMO system, namely the model structure selection and input-output pairing problems. The effectiveness of the proposed strategies will be demonstrated by considering an industrial air-conditioning system example.