Figure 2.12: Effect of the "smoother" feature on the electric lighting control
2.2.3 Heating System
An efficient heating controller should have predictive and adaptive features.
Unfortunately, available controllers such as NEUROBAT [Krauss et al., 1998] consume
too much computational time. Indeed, optimization of a cost function (grouping
discomfort and energy consumption) is unsuited to our experiments with 15 office rooms
and 15 heating controllers to run. Thus, a simpler empirical heating controller has been
developed, that nevertheless has both predictive and adaptive features.
Since the present report is not focused on heating device control, the algorithm is not
describe here. The reader interested in this subject should refer to the PhD thesis of
Antoine Guillemin [...].
2.3 AdaptiveModels
The different controllers being defined, the adaptive models used by them are described
in the present section. All these models are adapting at a room level. Only the weather
data prediction model is achieved at the building level.
2.3.1 Weather Data Prediction Model
The vector of solar irradiance predicted over the six next hours on the horizontal plane is
needed by the control system. Such data could have been provided by public weather
forecast service but in this case the information supplied is often averaged over several
hours and is not directly usable for a six hours ahead prediction. Moreover, the necessary
solar radiation sensor is already available in our system because it is required for the
lighting and thermal controllers. Thus, a solar irradiance predictor is used within this
work. The approach used was developed and verified in the NEUROBAT project
[Krauss et al., 1998, Morel et al., 2001]. It was there shown that artificial neural
networks (see the book of Haykin [Haykin, 1999] for comprehensive explanations of
ANNs) are the most effective method for the prediction of the horizontal global solar
irradiance8. A new version of a similar feed-forward network has been re-developed.
Summary :
Figure 2.12: Effect of the "smoother" feature on the electric lighting control 2.2.3 Heating System An efficient heating controller should have predictive and adaptive features. 2.3.1 Weather Data Prediction Model The vector of solar irradiance predicted over the six next hours on the horizontal plane is needed by the control system.
Tags :
heating,controllers,prediction,solar,hours,system,data,control,weather,been,adaptie,used,controller