1.1 Feed-Forward artificial neural networks

Artificial Neural Networks (ANN) are computing systems containing many simple nonlinear computing units or nodes, called neurons, interconnected by links. A feed-forward ANN has a large number of neurons arranged in layers. Generally, all neurons in a layer are connected to all neurons in adjacent layers through uni-direction links, which are represented by synaptic weights. In the three-layer perceptron, schematically depicted in Figure 3, the neurons are grouped in sequentially connected layers: the input, the output and the hidden layers. Each neuron in the hidden and output layer is activated by a non linear activation function that relies on the weighted sum of its inputs and the neuron parameter, called bias, b.

The output of a neuron in the output layer is:

where the h hidden units (processing elements) perform the weighting summation of the inputs xt and the nonlinear transformation by the sigmoid (log-sigmoid or tan-sigmoid) transfer function Y, (.)

a) Sun and daylight

Starting from a Reference south oriented facade equipped with solar control windows, two examples of substitution with glass collector were simulated to evaluate the incidence on the building’s luminosity figure 4, Variation 1 and Variation 2.

Variation 1 : The comparison with the Reference case permits to notice that the integration of the glass collector in lower side of the facade reduces notably the over-heating in the summer and middle seasons. It contributes to the sun and daylight distribution. In middle seasons and in winter its effects improve the light on the far end of the buildings. In all seasons its influence on daylight is negligible. The glass collector in the lower side of the facade does not introduce more electric lighting.

Variation 2 : The comparison with the Reference case permits to notice that the vertical integration and in the lower side of facade increase the limitation of over-heating in summer and middle seasons. In middle seasons and in winter its effects tend to be the same at the far end of the building than in the Reference case.

In all seasons its influence on daylight is minimal. The integration in vertical position and in the lower side of the facade introduces small electric lighting loads in summer.

The influence of its vertical position is not neutral and takes advantage of raytracing simulation.

The results given hereafter are extrapolated from Fraunhofer-ISE measurements, to take into account the evolution of the geometry of the glass collector between the current and the tested version [2].

Optical transmission:

The evolution of the optical transmission coefficient (t) function of incidence angle is given by figure 6.

The comparison of the curves relating to the transmission coefficient of an insulated glass with solar control, clearly shows the effect of the reflectors on the direct transmission of solar flow.

This effect is of course under-evaluated, if we take into account the thermal re-emission, due to the exchange between inside glass position 4 and inner building by convection and infra-red radiation.

Solar factor

The net solar flow in the room is the sum of three components :

• a first component is the optical transmission, figure 1,

• a second component is the thermal re-emission,

• a third component is the exchange between the solar collector and inner building. This last component represents the losses of the solar collector, in position 4. The relation which expresses the net flow is:

Qnet = x• IT • S + hi • a2 • S-(0v -0,)+ Umt • a, • S-(0c -0,) (5)

The experimental values established by Fraunhofer ISE [2] were obtained in accordance with calorimetric measurements in complex glazing [4], by considering the same temperatures for the interior, for the exterior and for the absorber. Thus the g solar factor does not take into account the 3rd component. The evolution of g according to the angle of incidence is given in the figure 6.

If we integrate the losses of the solar collector to the interior, we obtain a new estimation of an equivalent g-value. A very unfavorable assumption of its evolution curve is given in figure 6, with an absorber temperature of 80°C.

In this case, we notice that the net solar flow, Qnet accounts for approximately 20% of incident solar flow. This value is to be compared with a 0.33 g-value of a double glazing with solar control.

M. Soria, M. Costa, A. Oliva (Centre Tecnoldgic de Transferencia de Calor"),
D. Heineke, E. Antonicelli (University of Gottingen),

X. Farres, M. Vallds (Biosca & Botey),

J. Bordas, I. Hassanova (Jardineria Bordas),

M. Roa, J. L. Anguita (Arquitectura Produccions),

S. Juhling (Juhling),

M. Filippi, G. Marabini (Nature),

E. Martins (Percurso)

The central idea of the GREENFACADE project is to use decidous climbing plants to shade the facades of urban buildings. During winter, the vegetation loses its leaves and its effect is almost null, while during summer its effects tend to supress or reduce the overheating. However, the practical aplication of this simple idea is not straightfor­ward. With partial funding from European Comission, a multidisciplinar consortium has been gathered with the purpose of solving different problems associated with the use of vegetal elements as shadowing devices. In this presentation, an overview of the project after its first year will be given.

Introduction

Overheating during summer is a serious problem in Southern European climates, spe­cially in office buildings, mainly due to their bad passive behaviour. The use of air-conditioning relieves the problem, but at the expense of a great consumption of electricity.

Different solutions have been proposed to improve the passive behaviour of buildings with glazed facades. One of the better solutions is a combination of ventilated facades [5], [4], [6] with selective glazing. Such facades, if carefully designed, may have a thermal behaviour comparable to conventional (i. e., built with bricks) facades.

In our work, a well known traditional architecture solution is reconsidered. The central idea is to use decidous climbing plants to shade the facades. During winter, the vegetation loses its leaves and its effect is almost null, while during summer its effects are: (i) To reduce the solar radiation arriving to the facade. (ii) To decrease the ambient temperature due to evaporation and transpiration. (iii) To reduce the wind velocity at the facade surface. (iv) To reduce the heat lost by the facade due to thermal radiation. In general, solar radiation re­duction is the dominant phenomena, so the overheating during summer can be significantly reduced, if a part of the facade is covered by vegetation.

The goal of this presentation is to give a short summary of the GREENFACADE research project. The GREENFACADE consortium of SME companies and research institutions, with partial finantial support by the European Comission, has gathered a multidisciplinar team (engineers, biologists, architects, mathematicians and gardeners) with the purpose of solv­ing the problems associated to the use of vegetal elements as shadowing devices.

A good review of previous studies can be found in [3], where the effect of the vegetation is measured with an experimental set up and a conventional architecture simulation code (APACHE) is used to simulate different buildings. To do so, the vegetation is modelled as a set of passive materials. Other works, however, consider the situation mainly from an architectural point of view [7]. In our approach, architectural aspects are important, but other

*Centre Tecnologic de Transferencia de Calor (CTTC), Lab. de Termotecnia i Energetica,

Universitat Politecnica de Catalunya (UPC), labtie@labtie. mmt. upc. es, www. cttc. upc. edu

elements are also considered central, such as the optimization of the designs, considered from a heat transfer and solar energy engineering. This will allow us to quantify the amount of energy to be saved with this solution, and to compare the green facades with other bio­climatic architecture technologies.

The GREENFACADE project is oriented along two different lines:

• Find solutions to the architectural problems associated to the vegetation such as: acess for maintenance, aspect of the plants from the building indoor, lighting, outdoor design, identification of species of interest, etc.

• Develop a detailed (time dependant) numerical model of the vegetation, treating each of the previous phenomena separately, and using laboratory measurements of vegeta­tion properties. This model, integrated in the code AGLA [1], [2], will allow to determi­nate the amount of energy to be saved for different building types, facade orientation and vegetation density The model will be validated using experimental data from dif­ferent prototypes.

As mentioned before, there are three kinds of coatings that will be discussed here: partially transparent absorbing coatings, opaque reflective coatings and non-static, optically switchable coatings that change their optical properties due to a change in a control parameter or ambient condition.

Absorbing coatings as described above can inhibit unwanted refractive effects of micro prisms. Reflective coatings offer the same option as absorbing coatings and they are additionally capable to improve or completely change the light shading abilities. Switchable coatings like gaschromic switchable WO34 films can be applied to the structures to offer new actively controllable features to the sunshading-system. Switchable mirrors that change from a transparent to a reflective state are capable to completely change the behavior of sun-shading devices.5 For certain structures, this can be achieved without disturbing or prohibiting the direct view through the glazing. As mentioned above, one easy option for reducing the refractive effect of prism arrays is a semitransparent coating on the prism-face responsible for the refracted image. This will reduce the refracted image without disturbing the direct view, as can be seen in figure 9. The graph is similar to figure 3 with some slight differences. In figure 9, the incident light hits the surface at an angle of 35° with respect to the sample normal. Therefore, the peak representing the direct image is located at -35° and the peak representing the refracted image at about -60°. In the bold curve, the prism surface facing the sun is coated with a Ni layer with transmittance of about 25%. The refracted image is reduced to approximately one fourth as expected, while the direct image is nearly unchanged. Of course, the hemispherical transmission is also reduced and especially for small angles, this could be an unwanted effect. The advantage of absorbing coatings in comparison to reflective coating is that the light guiding properties of the system stay unchanged and therefore can be easily used with existing designs.

If light absorption is unwanted, reflecting coatings can be used. However, the properties and effects of reflective coatings have to be considered already in the design process. In

these first tests, the coatings have been applied to existing structures to study the effects. Optimising the structure for use with face selective mirrors will be the next step. Below, first results of simulations are presented. A good example for the effects of a face selective mirror is shown in figure 10. In this case, the structure shown in figure 4 is coated with reflective Al coating. In contrary to figure 4, the coating was applied to the lower prism face oriented nearly like the normal of the sample. In figure 10, the measured hemispherical transmission for changing incidence angles is shown for the structure with and without a coating. For small angles, the reduction of transmission is small, because the mirror is nearly parallel to the incident light. For large angles, where for unchanged structure the transmittance of the light is increasing again, the coated prisms still reflect most of the incident light. Thus, a coating is not only capable of reducing distracting visual effects, but it becomes a new free parameter for designing microprism arrays for light shading applications.

A. K. Athienitis*, A. Tzempelikos* and Y. Poissant**

*Concordia University, BCEE Dept, Montreal Canada
**CETC Varennes — PV & Hybrid Systems Program, Natural Resources Canada

Introduction

This paper describes an outdoor experimental investigation of double fagades with integrated photovoltaic panels. Results predicted with a simple analytical one-dimensional model are compared with the experimental results and the impact of convective heat transfer coefficient is studied.

Photovoltaic modules may be integrated in buildings to form the exterior envelope layer while generating electricity. Their cost-effectiveness is thus improved in comparison with stand-alone systems that need a separate support structure, particularly when they replace expensive envelope exterior layers such as high quality precast panels, granite or marble. However, their electrical conversion efficiency, which is presently about 15%, is rather low in comparison with thermal systems; typically about 70-80% of the solar radiation incident on the PV panels is lost by convection and infrared radiation to the outdoor environment. By placing the PV modules in the interior of an airflow window (attached to the inner or outer glazing), and passing fresh air through the glazing cavity, we achieve the twin objectives of capturing much of the absorbed solar energy that would otherwise be lost as heat, while cooling the PV panels and thereby raising their electrical conversion efficiency. Thus, the window or double fagade functions as a cogeneration device that generates both electrical and thermal energy. Usually, some daylight should be transmitted through the system, that is, the PV modules should not cover the whole available area, but rather 50-70%. Thus, a significant percentage of incident solar radiation is transmitted as daylight, potentially reducing by a corresponding amount the electricity consumption for lighting.

A PV/thermal workshop [1] organized by the International Energy Agency concluded that there is an urgent need for research on effective integration of photovoltaic and solar thermal systems so that physical systems can be developed that perform both functions in a reliable and optimal manner. While several BIPV projects have been built around the world, such as the Mataro Library [2] in Spain and some progress on modelling has been reported [3, 4], there is no systematic procedure for their overall optimization and prediction of reliability. Part of the reason is the complexity, particularly of PV-airflow windows, which require consideration of thermal, electrical and daylighting performance. Motorized blinds may be employed in an airflow window with PV to control daylight transmission into a room in conjunction with dimming of electric lights and heating/cooling needs.

While many researchers have considered various system configurations and developed thermofluid models to investigate performance, there is a need for systematic optimization of such systems to raise their overall performance and cost effectiveness.

The mechanisms amplifying the effects considered have been discussed. They include virtual mechanisms of electric charge generation on the screen surfaces and their accumulation. The “quasicapacitance effect” [30] can be related to these mechanisms, in which the charge generation is carried out according to the G. G. Thomson’s mechanism (1896) as well as according to the film thermal electromotive force mechanism [38] arising in thermally strongly differentiated superinsulation layers or according to the semiconductor surface charging mechanism as a result of interaction between the screens and the gas medium [39]. The electret basis of the screens (the PETF-DA-12 film is used in electric capacitors) at low temperatures allows to keep the accumulated charges during a very long time.

The general picture of the given oscillatory process of the parameters being investigated is fully regulated by the ratio between acceptor and donor gases in HIC as well as by the magnitude of relative gas inflow from the environment, by the magnitude of developed dimension-quantised film surface and by the surface’s Fermi level magnitude. It goes without saying that it is also assumed that the atmospheric medium with daily humidity and temperature variations is considered as the ambient environment.

2. EIHIS, EIHCIS, MADHM EFFECTS

The effects detected in the course of full-scale experiments [6,7] are as follows:

1. Effect of effusion induced hydrogen instability of the superinsulation (EIHIS),

2. Effect of effusion induced heat conduction instability of the superinsulation in
cryogenic and vacuum facilities (EIHCIS), 3. Effect of multiplication of the amount of desorbed hydrogen molecules (MADHM) in respect to the magnitude of the moist air inflowing molecules will allow to deeply understand the essence of phenomena occurring in foliated heat-insulating systems.

The underground heat store is an 1100m3 body of water that stores the heat generated by the PVT and solar thermal panels for use in the buildings during the colder months. The top of the store is insulated with a floating lid of 500mm expanded polystyrene. It is hinged around the perimeter to allow for the expansion and contraction of the water and the design also incorporates a suspension system to support the roof should the water level reduce. The sloping sides are uninsulated. As long as the ground around the store is kept dry, it will act as an insulator and additional thermal mass, increasing the capacity of the
store. The high specific heat capacity of water (4.2kJ/kg°C) makes it a good choice for storing heat.

During the summer there will be little or no demand for heat in the building, so the heat generated by the PVT array will stored in the heat store. In the autumn some of the solar heat generated will be used directly in the buildings and the excess will be added to the heat store. The temperature of the water in the store will gradually rise over the summer and early autumn. During the winter the solar heat generated will be less than the buildings heat load, and heat will be extracted from the heat store to heat the incoming air to the building. The temperature of the water in the store will drop as the heat is extracted. Some heat will also be lost to the surroundings. This is estimated to be about 50% of the total heat put into the store over the summer. The relatively low-grade heat from the store can be used to preheat the incoming air to the building, as the outside air will be at a lower temperature than the water.

Architectural Expression

No attempt was made to replicate the arts and crafts style of the original buildings in the new building works. The editions and replacements are expressed in a clean modern, albeit sympathetic, idiom reflecting the contemporary concerns of Renewable Energy Systems and the leading edge energy technologies deployed over the site and concealed within the buildings.

View from South West

PVT Array

Heat Store prior to lid being installed

In order to test the system performance in the current configuration on a seasonal time scale and to investigate the influence of the main system parameters, a dynamic simulation model was developed within the TRNSYS 14.2 environment. TRNSYS is a widely used system simulation programme [10] provided with libraries of subroutines called types which simulate specific components or processes. Beside standard ones, user written subroutines are available.

The model of the test building and of the system was obtained assembling the following types:

— the non standard Duct Ground Storage Model (DST) [11]

— the non standard Type 160 Floor heating and hypocaust [12]

— the standard Type 56 Multizone Building Component

and the types for the auxiliary system equipments (counter current flow heat exchanger, pumps). Data collected in the above described measurement campaign were used to calibrate the model. The main three types were calibrated individually and then linked together. Simulated and measured data agree within experimental errors, as it is shown in Figure 3 for two quantities, namely the air temperature in test room 1 and the thermal power extracted from the test room 1.