Minimization of actuator repositioning using neural networks with application in nonlinear HVAC1 systems

In process control the main goal is not only the output tracking but also the design of a suitable control signal. It is due to the fact that it is dealt with the actuator's performance directly. This means that the undesired oscillations should be removed to alleviate the malfunction of actuators. More than this, saving energy is an important point in energy-consuming systems such as Heating, Ventilating and Air Conditioning (HVAC) systems. In this paper, a nonlinear saturating element is designed with the help of neural networks. This element chops the undesired oscillations of the control signal. The performance of the system is guaranteed by the use of a number of PID controllers as central controllers and implementation of a performance measurement index to reduce the error between the outputs and desired input signals. A multi-layer perceptron network is used and trained with Error Back Propagation algorithm. The method is applied to a nonlinear model of an HVAC system, which demonstrates the good performance of the proposed design method.