A simple yet effective general regression neural network (GRNN) paradigm is suggested for heating, ventilating, and air conditioning (HVAC) control applications. Unlike the popular backpropagation paradigm, the proposed GRNN is simple to implement, requires only one parameter, and works well with sparse and random data. A simple local HVAC control example for a heating coil is chosen to test the GRNN effectiveness. The GRNN is used to capture the static characteristics for both valve/dampers and coils. Both simulated and experimental characteristics are used as identification as well as test data for the GRNN. The GRNN captures the characteristics remarkably well and, due to its simplicity, it exhibits promise for implementation in real controllers. A combined feedforward and feedback control algorithm is explored that can utilise the GRNN method to identify static characteristics and can then subsequently be used in a feedforward controller to generate control signals based on the identified characteristics.
KEYWORDS: year 1996, building services, air conditioning, heating, ventilation, controls, expert systems, heating coils, testing, performance, properties, valves, dampers, coils, algorithms, calculating
Citation: Symposium Papers, Atlanta, GA, 1996
Product Details
- Published:
- 1996
- File Size:
- 1 file , 1.2 MB
- Product Code(s):
- D-17205
