Self-learning neurocontroller for maintaining indoor relative humidity

An air-conditioning system is designed to meet maximum space cooling load. Thus the system's controller needs parameter adjustment periodically due to changes in the environment and operating conditions. For a constant-air-volume system at system part-load operation the indoor relative humidity may exceed the limit recommended for comfort and health. This paper describes the application of neural networks to develop an intelligent air handling. The purpose is twofold: 1) the controller self-learning capability will substitute conventional parameter adjustment; 2) in addition to controlling the indoor temperature, the controller will also limit the indoor relative humidity. With the designed cost function, the proposed controller is a promising tool to limit the rise in indoor relative humidity in this particular constant-air-volume system.