3671 -- Adaptive Control of HVAC Processes Using Predictive Neural Networks PDF

States that in practically all HVAC systems, the maintenance of process setpoints relies on feedback loop control. Proportional, integral and derivative (PID) algorithms are often used in these controllers. The optimisation of such loops has been the subject of many studies, most of which use techniques to first model the dynamic behaviour of the process and then to use the model to arrive at a series of control constants that would effect a first-order response to changes in setpoint. Demonstrates how artificial neural networks (ANNs) can be used for the same purpose, and compares the performance of two types of ANN controllers to a conventional PID controller.

KEYWORDS: Controls, algorithms, heating, ventilation, air conditioning, expert systems, comparing, experiment, predictive controls, adaptive controls, artificial intelligence.