A commercial PI-neural controller for the control of building services plant

Discusses some of the practical issues relating to the development of an implementation of a hybrid PI/neural network controller for commercial application in heating, ventilation and air-conditioning (HVAC) control systems. To ensure commercial application it is necessary that such a controller is no more complex to commission than a standard PI controller, therefore the commissioning of the neural network must be entirely automated. Methods for incrementally training the network are introduced which avoid the control problems resulting from using an initially untrained network. Results are presented from testing the controller on a full-scale test-rig