Load Matching, Grid Interaction, and Advanced Control - Modeling, Design, and Optimization of Net-Zero Energy Buildings - page 317

notablygeneralityandphysicalinsight-differentkindsofmodelsareuseful
for different kinds of applications. Different approaches commonly used in
advanced controls are presented in Section 6.4.
6.3.2.3 Modeling of Noncontrollable Inputs
In conventional feedback-based building control systems, modeling the
inputs (internal gains, solar gains, outdoor temperature, etc.) is not critical.
Heating or cooling is provided “as required” by comparing the current
temperature with the desired value, as in the case of a PID loop. Forecast
information is rarely used in conventional control systems.
In contrast, in predictive control, it is important to have a reasonably good
idea of the relevant input profiles to anticipate the building needs. The
quality of the forecasts deserves as much attention as the quality of the
model. The effort spent in the development of a model might be wasted if
the forecasts are neglected. Furthermore, while low-resolution data might
be sufficient for design purposes, in the development of a control strategy,
higher-resolution data are required.
Thetwomainsetsofinputsthatinfluencetheperformanceofabuildingare:
(a) weather variables (exterior temperature, solar radiation, humidity) and
(b) occupancy and related energy use and heat gains.
Forecasts released by weather service organizations are becoming
increasingly accurate and detailed, and tend to contain information with
higher spatial and time resolution (Poulin, 2006). They are often available
online, free of charge. This has not always been the case: the earliest studies
on predictive control in buildings relied on “home-made” predictors based
on qualitative forecasts or historical records (Chen and Athienitis, 1996),
or the likelihood of weather conditions for the following day based on the
current conditions (Nygård-Ferguson and Scartezzini, 1989). In the case
of large commercial buildings, which are less dependent on weather
conditions, optimal control studies were based on typical hourly variations
from average measurements for a given day of the year (Braun, 1990).
The information available in weather forecasts may not be set in a format
that could facilitate its immediate incorporation in building simulation. For
example, solar radiation on a horizontal surface is usually available, but
some kind of solar radiation modeling must be used to calculate solar gains
into the building.