IBPSA1991

Several attempts have been made these past few years to obtain a simplified dynamic model of thermal behaviour of buildings from recorded data. System identification techniques are well known in other fields, such as aeronautics.

In conducting and teaching Building Simulation, we often find two main disadvantages of conventional models: inconsistency of simulation results obtained by different users of the same model, and long machine times required for annual simulations of relatively simple buildings. In searching for better simulation methods, we decided to depart from the conventional method and to introduce machine learning into mathematical modelling of buildings. This resulted with a new model, based on learning of building energy properties from monitored data.

The CLIM 2000 software proposes a host architecture for dynamic, modular building energy modelling. CLIM 2000 provides a library of basic models, which in particular allows simulation of hydraulic heating networks. Unlike those oriented towards simulation of the building envelope, these models pose specific resolution problems. Owing to the modelling adopted, the elements making up the hydraulic networks require knowledge of the topology of the heating circuit.

Changing working processes not only in manufacturing and assembly but also in office work require advanced buildings which allow a maximum of flexibility towards building structure and all its services (HVAC, telecommunications etc.). To gether with the growing importance of shared tenant services in intelligent buildings, theattempt to reach low-energy consuming buildings despite of their required flexibility in supplying a good infrastructure leads to a growing importance of "building performance".

Models used for forecasting of future energy demand are often econometrically based and do not attempt to look in detail at end uses of energy. It has become clear in recent years that 'this approach to energy forecasting can produce more realistic estimates of future demand in the building sector if it is used in conjunction with models which-cover important physical and social parameters and the contraints which these impose (Grubb 1990; Grubb 1991). Such complementary models provide insights which would otherwise be missed.

A set of statistical regression equations was developed to predict relative heating and cooling loads of external zones of commercial buildings. The equations were derived from the coil loads predicted by several thousand DOE-2 simulations.

The topic of this paper is the use of low temperature air (40 F or 5 C) for room cooling. Cold air systems can offer energy and space savings relative to higher temperature cooling systems. As the supply temperature and flowrate are reduced, considerations such as adequate flowrate, jet dumping or separation, condensation on duct walls, and decreased relative humidity become increasingly important. Cold air jet separation from the ceiling can be a problem resulting in unacceptable thermal discomfort in the occupied zone.

In this paper, we develop a discrete approach to describe the transport of condensible vapors through a microporous substance. We consider only isothermal water migration under uniform atmospheric air pressure, at temperature lower than 100C with negligible gravity. The pore-structure which is supposed to be representative of the material is built on a 2D random network of tubes. The basic phenomena (adsorption/desorption, diffusion, condensation) that occur during the water vapor transport in a single cylindrical pore at the steady state are taken into account.

With the accelerating use of building performance prediction models in a design context, the need for comprehensive program accreditation procedures is becoming more pressing. This paper recognises the importance of the validation component of such a procedure and makes a case for containing much of the present knowledge about validation within an interactive facility centred on test cells.