Abstract
In this paper, we propose a novel variable setpoint RNN controller for maintaining comfortable indoor
environment in double storey residential building by controlling the motorised thermostatic radiator valves
(TRVs) mounted on radiators. In order to monitor the indoor environmental condition of the building the
sensor interface collects information from different sensors and sends this information to random neural
network (RNN) controller. The RNN controller ensures comfortable environment for occupants by
regulating the air temperature of the building according to the setpoint suggested by PMV index based
variable setpoint estimator. The proposed RNN controller is compared with ANN controller and it is found
that accuracy of RNN controller is 26% more than ANN controller and conserves 2.75% more energy than
ANN controller at PMV index based temperature setpoints. The RNN controller has the capability to adjust
the room temperature to lower setpoints (not included in the training data) while ANN controller failed to
maintain accurate comfortable environment for the operating points not covered in the training data. The
results show that the percentage of accurate air temperature regulation for RNN controller is 95.69% while
for ANN it is 2.22%.
environment in double storey residential building by controlling the motorised thermostatic radiator valves
(TRVs) mounted on radiators. In order to monitor the indoor environmental condition of the building the
sensor interface collects information from different sensors and sends this information to random neural
network (RNN) controller. The RNN controller ensures comfortable environment for occupants by
regulating the air temperature of the building according to the setpoint suggested by PMV index based
variable setpoint estimator. The proposed RNN controller is compared with ANN controller and it is found
that accuracy of RNN controller is 26% more than ANN controller and conserves 2.75% more energy than
ANN controller at PMV index based temperature setpoints. The RNN controller has the capability to adjust
the room temperature to lower setpoints (not included in the training data) while ANN controller failed to
maintain accurate comfortable environment for the operating points not covered in the training data. The
results show that the percentage of accurate air temperature regulation for RNN controller is 95.69% while
for ANN it is 2.22%.
| Original language | English |
|---|---|
| Title of host publication | Proceeding of iiSBE Net Zero Built Environment 2014 17th Rinker International Conference, Gainesville, FL, 6&7 March |
| Publisher | University of Florida |
| Pages | 326-343 |
| Number of pages | 18 |
| Publication status | Published - 6 Mar 2014 |
Keywords
- random neural network, artificial neural network, Building Energy Management Systems,
- artificial neural network
- PMV index based control scheme
- building energy management systems