Content uploaded by Frédéric Haldi
Author content
All content in this area was uploaded by Frédéric Haldi on Nov 14, 2014
Content may be subject to copyright.
CITYSIM: COMPREHENSIVE MICRO-SIMULATION OF RESOURCE FLOWS
FOR SUSTAINABLE URBAN PLANNING
Darren Robinson1, Haldi, F.1, Kämpf, J.1, Leroux, P.1, Perez, D.1, Rasheed, A.1, Wilke, U1
1Solar Energy and Building Physics Laboratory (LESO-PB), Ecole Polytechnique Fédérale de
Lausanne (EPFL), CH-1015 Lausanne, Switzerland
ABSTRACT
In this paper we describe new software “CitySim”
that has been conceived to support the more
sustainable planning of urban settlements. This first
version focuses on simulating buildings’ energy
flows, but work is also under way to model energy
embodied in materials as well as the flows of water
and waste and inter-relationships between these
flows; likewise their dependence on the urban
climate. We discuss this as well as progress that has
been made to optimise urban resource flows using
evolutionary algorithms. But this is only part of the
picture. It is also important to take into consideration
the transportation of goods and people between
buildings. To this end we also discuss work that is
underway to couple CitySim with a micro-simulation
model of urban transportation: MATSim.
INTRODUCTION
It is estimated that over half of the global population
is now living in urban settlements (UN, 2004), in
which three quarters of global resources are
consumed (Girardet, 1999). Energy derived from
fossil fuels is key amongst these resources, so that
urban settlements are responsible for the majority of
greenhouse gas emissions. It is thus important that
existing urban settlements are adapted and that
proposed settlements are designed to minimise their
net resource consumption. Software for simulating
and optimising urban resource flows will play an
essential role in this process. In this paper we
describe progress that is being made in one such
initiative: the development of CitySim. We also
describe work that is underway to add further
functionality to CitySim and to couple CitySim with
a microscopic transport simulation model MATSim
to resolve for all key urban resource flows.
Achieving this would provide an invaluable platform
for the testing of planning interventions to improve
urban sustainability.
CITYSIM STRUCTURE
In common with its predecessor SUNtool (Robinson
et al, 2007) the use of CitySim’s Java-based GUI to
simulate and optimise building-related resource flows
proceeds according to four key steps:
- Definition of site location and associated climate
data.
- Choice and adjustment of default datasets for the
types and age categories of buildings to be studied.
- Definition of 3D form of buildings; definition of
energy supply and storage systems to be modelled;
refinement of building and systems attributes.
- Parsing of data in XML format from the GUI to the
C++ solver for simulation of hourly resource
flows; analysis of results streamed back to the
GUI.
The scale of analysis may vary from a
neighbourhood of just a few buildings, though a
district of several hundred to an entire city of tens of
thousand. But in each case the core modelling
capability and the data needs of these models is
similar. In the following we describe this modelling
capability and how this is being extended to facilitate
thoroughly comprehensive simulations of urban
resource flows and how these might be optimised.
CITYSIM: CORE MODELS
For the purposes of urban scale simulation, it is
important to achieve a good compromise between
modelling accuracy, computational overheads and
data availability. In all cases these have been the
criteria applied in the selection of an appropriate
modelling methodology.
Thermal Model
From the above criteria it was decided to develop a
model based on analogy with an electrical circuit;
more specifically based on a resistor-capacitor
network. In this case a conducting wall can be
represented by one or more temperature nodes
(Lefebvre, 1997). The heat flow between a wall and
the outside air can be represented by an electric
current through a resistor linking the two
corresponding nodes and the wall’s inertia can be
represented by a capacitance linked at that node.
In our model (Kämpf and Robinson, 2007), which is
a refinement of that due to Nielsen (2005), an
external air temperature node Text is connected with
an outside surface temperature node Tos via an
external film conductance Ke, which varies according
to wind speed and direction (Figure 1). Tos, which
also experiences heat fluxes due to shortwave and
Eleventh International IBPSA Conference
Glasgow, Scotland
July 27-30, 2009
- 1083 -
longwave exchange, is connected to a wall node Tw
of capacitance Cw via a conductance defined by the
external part of the wall. In fact this node resembles a
mirror plane, so that we have similar connections to
an internal air node Ta of capacitance Ci via an
internal surface node Tis. Tis may also experience
shortwave flux due to transmitted solar radiation and
a longwave flux due to radiant heat gains from
internal sources (people and appliances) and Ta may
experience convective gains due to absorbed
shortwave radiation, internal casual gains and heating
/ cooling systems. Finally, our internal air node may
be connected with our external air temperature node
via a variable resistance due to infiltration and
ventilation.
Figure1: Monozone form of the CitySim thermal
model.
For a whole building with many subspaces, the air
nodes of each zone are linked via the separating wall
conductance. Interzonal airflow can also be handled
through this conductance. To account for the
corresponding inertia the capacitance of the
separating wall is subdivided and allocated to the
neighbouring zone’s capacitance (Figure 2).
Figure 2: Interzonal connection between zones
represented by the two-node thermal model
In general, an n-node model may be represented by
the following differential equation:
() () () ()
tutTtAtTC
r
r
r
+⋅=
′
⋅ …(1)
Where
()
tT
r
represents the temperature vector at the
n-nodes
()
tT′
r
denotes its derivative with respect to
time and
()
tu
r
represents the source terms of each
node. C is the positive diagonal thermal capacity
matrix and A(t)is the symmetric heat transfer matrix.
In the particular case of our two-node multi-zone
model eq(1) can be expressed as follows:
()
() ()
()
()
()
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
′
′
+
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
′
′
⋅
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
′
′
⋅
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
tu
tu
tT
tT
GF
ED
tT
tT
C
C
w
a
w
a
w
ar
r
r
r
r
r
2
1
0
0 ...(2)
Where for the ith zone C1 and C2 are internal (air and
the wall separating zones i and j) and wall (external
and separating) capacitances. D, E, F and G
correspond to combined conductances for external
and separating walls.
Predictions from this simplified model compare well
with those of ESP-r (Clarke, 2001) for a range of
monozone and multizone scenarios.
Radiation models
In common with CitySim’s predecessor “SUNtool”,
the Simplified Radiosity Algorithm (SRA) of
Robinson and Stone (2004) is used to solve for the
shortwave irradiance incident on the surfaces
defining our urban scene. For some set of p sky
patches, each of which subtends a solid angle Φ (Sr)
and has radiance R (Wm-2Sr-1) then, given the mean
angle of incidence ξ (radians) between the patch and
our receiving plane of slope
β
together with the
proportion of the patch that can be seen σ (0 ≤ σ ≤ 1),
the direct sky irradiance (Wm-2):
(
)
∑=Φ= p
ii
dRI 1cos
ξσ
β
…(3)
For this the well known discretisation scheme due to
Tregenza (1993) is used to divide the sky vault into
145 patches of similar solid angle and the Perez all
weather model (Perez, 1993) is used to calculate the
radiance at the centroid of each of these patches. The
direct beam irradiance
β
b
Iis calculated from the
beam normal irradiance Ibn which is incident at an
angle ξ to our surface of which some fraction ψ is
visible from the sun, so that:
ξ
ψ
β
cos
bnb II
=
…(4)
Now the direct sky and beam irradiance contributes
to a given surface’s radiance R which in turn
influences the irradiance incident at other surfaces
visible to it, so increasing their radiance and vice
versa. To solve for this a similar equation to that used
for the sky contribution gives the reflected diffuse
irradiance. In this case two discretised vaults are
used, one for above and one for below the horizontal
plane, so that:
(
)
∑=Φ= p
ii
RI 21*cos
ξω
ρβ
…(5)
where
ω
is the proportion of the patch which is
obstructed by urban (reflecting) surfaces and R* is
the radiance of the surface which dominates the
obstruction to this patch (in other words, that which
contributes the most to
ω
). As noted earlier, R*
depends on reflected diffuse irradiance as well as on
the direct sky and beam irradiances. For this a set of
simultaneous equations relating the beam and diffuse
sky components to each surface’s irradiance, which
itself effects the reflected irradiance incident at other
surfaces, may be formulated as a matrix and solved
either iteratively or by matrix inversion (Robinson
and Stone, 2004).
The principle complication in the above algorithm
lies in determining the necessary view factors. For
- 1084 -
obstruction view factors, views encapsulating the
hemisphere are rendered from each surface centroid,
with every surface having a unique colour. Each
pixel is then translated into angular coordinates to
identify the corresponding patch as well as the angle
of incidence. For sky view factors then,
ξ
σ
cosΦ is
treated as a single quantity obtained by numerical
integration of Φ⋅ d
ξ
cos across each sky patch.
Likewise for
ξ
ω
cosΦ, for which the dominant
occluding surface is identified as that which provides
the greatest contribution. A similar process is
repeated for solar visibility fractions for each surface,
for which a constant size scene is rendered from the
sun position.
In addition to using the above view information to
calculate incident shortwave irradiance, this may also
be used to calculate longwave irradiance; given the
corresponding surface and sky temperatures (see
Robinson and Stone 2005). Eq (3-5) may also be
solved using luminance / illuminance as an input to
model the external luminous environment. Additional
renderings may be then computed to determine the
view information necessary to calculate the direct
and reflected contributions to internal illuminance (at
known points) as well as the incoming luminous flux
for internal reflection calculations (Robinson and
Stone, 2005; 2006).
Behavioural models
One of the key sources of uncertainty in building /
urban simulation relates to occupants’ behaviour,
which is inherently stochastic in nature. Key types of
behaviour and their main impacts on urban energy
flows are as follows:
Presence: metabolic heat gains and pollutants.
Windows: infiltration rates.
Blinds: illuminance and transmitted irradiance.
Lights: heat gains and electrical power demand.
Electrical appliances: heat gains and electrical power
demand.
Waste: production of combustible and recyclable
solids.
The central behavioural characteristic relates to
occupants’ presence, which of course determines
whether they are available to exercise any other form
of influence on resource flows. Based on the
hypotheses that all occupants act independently and
that their actions at time t+1 depend only upon the
immediate past (t), we model transitions in
occupants’ presence (present to absent (T10) or
present (T11) and vice versa) based on the Markov
condition that:
()
()
()
tTjXiXP
lXkXjXiXP
ijtt
Ntttt
:
,...,,
1
11
====
====
+
−−+ …(6)
For this we take as input a profile of the probability
of presence at each time t and a mobility parameter μ
– the ratio of the probability of changing state to not
changing state, so that the transition absent to present
is found from:
)1()(
1
1
)(
01 ++⋅
+
−
=tPtPtT
μ
μ
…(7)
And that of present to present from:
)( )1(
)1()(
1
1
)( )(1
)( )1(
)( )(1 0111
tP
tP
tPtP
tP tP
tP
tP
T
tP tP
T
+
+
⎥
⎦
⎤
⎢
⎣
⎡++⋅
+
−
⋅
−
=
+
+⋅
−
=
μ
μ
…(8)
The remaining transitions are simply: 1110 1TT −= and
0100 1TT
−
=
. In addition to this we also need to
consider long absences, for example due to illnesses
or vacations. For this we use a daily profile of the
probability of starting a long absence and a further
profile for the cumulative probability of the duration
of this absence. This takes precedence over the short
time step transitions in presence, described above.
For further details, we refer the reader to Page et al,
(2007a).
Although this occupancy model has been integrated
with CitySim, it is currently disabled. Rather, and as
an intermediate step, use is currently made of
deterministic rules / profiles describing occupants’
presence and behaviour. In the near future however,
the possibility will be provided to switch between
deterministic representations and stochastic models.
For this we will also use the models of: Haldi and
Robinson (2009a, b) for window openings and the
corresponding ventilation exchanges; Haldi and
Robinson (2009c) for interactions with blinds and the
impacts on solar radiation and daylight transmission;
Page et al (2007b) for electrical appliances use and
their associated power consumption and heat gain
production. In the first instance the simplified model
of Page (2007) will be used for solid waste
production.
For interactions with lights we will initially use the
same models as those integrated with Lightswitch-
2002 (Reinhart et al, 2004). In particular, the
probability of switching on lights at arrival as a
function of minimum internal workplane illuminance
(Ei,min) will be predicted by the expression due to
Hunt (1979). Switching on lights at intermediate
times is also predicted as a function of Ei,min, but by
the expression of Reinhart and Voss (2003) whereas
switching off at departure will be modelled as a
function of expected duration of absence (Pigg et al,
1996). Note that interactions with blinds will be
resolved before those with lights.
Plant and equipment models
This category of model includes both heating,
ventilating and cooling (HVAC) systems and energy
conversion systems (ECS).
The HVAC model is based on the psychrometry of
humid air, considering an ideal mixture of two
perfect gases: air and vapour. It computes the
psychrometric state (temperature and moisture
- 1085 -
content and hence the enthalpy h) of the air at each
ith stage in its supply (e.g. outside, heat recovered,
cooled and de-humidified, re-heated, supply). Given
the required mass flow rate m
& (which may be
defined by the energy to be delivered or the room
fresh air requirement) the total delivered sensible and
latent loads for all stages in the heating and/or
cooling of air q can then be calculated:
i
n
iihmq ∑Δ= & …(9)
The family of ECS models comprise a range of
technologies that provide/store heat and/or electricity
to buildings. These include a thermal storage tank
model for hot/cold fluids, boilers, heat pumps,
cogeneration systems, combined cogeneration and
heat pump systems, solar thermal collectors,
photovoltaic cells and finally wind turbines. These
ECS models are in general based on performance
curve regression equations whereas a simplified
thermal model simulates the sensible / latent heat
storage (so that phase change materials are also
accounted for).
If the ECS models have insufficient capacity to
satisfy the HVAC demands then the supply state is
adjusted and the predicted room thermal state is
corrected (using the thermal model).
Integration
The conceptual structure of CitySim is presented in
Figure 3 below. The GUI produces an XML file
which contains a geometric description of our urban
scene together with numerous attributes that relate to
each of the models influencing urban resource flows.
This XML file also contains pointers to a climate file
as well as to the appropriate entries in databases
relating to building constructions, occupatants’
behaviour and plant and equipment characteristics as
well as the fuels combusted. This data is parsed to the
CitySim solver, which creates instances of the
objects describing our scene. Each of the models are
then called in turn and are parsed the necessary data.
Shown in black are each of the functions that have
been developed within CitySim and shown in grey
are those that are either under development or are
planned to be developed in the near future. These are
described in the remaining sections of this paper.
Figure 3: Conceptual structure of CitySim
CITYSIM: EXAMPLE APPLICATION OF
CORE MODELS
We have applied CitySim’s core models to a group of
buildings in the district of Matthäus in Basel
(Switzerland). For this, we used a 3D model provided
by the city’s Cadastral Office, and completed the
physical description of the buildings by means of the
national census data for the year 2000 and results
from a recent visual field survey of the district.
Figure 4 shows a projected view of the group of
buildings simulated, in which we have represented
for one particular day and hour in the year the surface
averaged irradiance at its centroid by coloured points.
The colour map used represents the intensity in
irradiance, from red to blue, calculated by the SRA
model.
Figure 4: A projection of the group of buildings
simulated in Matthäus district, Basel, Switzerland.
Using the buildings’ construction date, renovation
status and with the help of renovation specialists
(EPIQR Rénovation), we attributed the physical
characteristics relating to the walls, roofs and
- 1086 -
windows. See Kämpf and Robinson (2009) for a
fuller description of this scene and the means for its
attribution.
By way of example, figure 5 shows the averaged
internal building temperature and the ideal heating
and cooling demands for a selected building in the
group, throughout the year on an hourly basis. The
temperature set-points were chosen to be 21°C for
heating and 26°C for cooling.
0
5
10
15
20
25
0 1000 2000 3000 4000 5000 6000 7000 8000 -20
-10
0
10
20
30
40
50
60
70
80
Temperature (celsius)
Power (kW)
Hour in the year
Zone averaged internal temperature (celsius)
Heating needs (kW)
Cooling needs (kW)
Figure 5: Internal volume-averaged temperature and
heating and cooling demands for one of a group of
buildings.
CITYSIM: FURTHER MODULES
In parallel with the above work in developing the
core solver for CitySim, we have also been active in
developing a range of complementary modelling
capabilities to enhance CitySim’s scope. These as
well as some planned models (on which work will
start shortly) are described below, to give a flavour of
what CitySim’s capabilities are expected to be in the
near future.
Further object attribution
The energy flows during the operation of an urban
development are just part of the story. It is also of
interest to consider the embodied energy content of
materials, to facilitate life cycle energy analysis.
Related to this is an issue of primordial importance to
the urban designer and developer – that of cost (both
capital and running). Realistically speaking, this
typically outranks environmental performance as a
fitness function to optimise. Work is thus underway
to add these additional attributes (cost and embodied
energy) to the properties of the objects that comprise
a CitySim urban scene. This is straightforward. Less
straightforward is the compilation of databases that
are rich enough to describe the attributes of the most
commonly used construction / servicing products for
potentially any users’ location, including the costs of
their acquisition and the local labour involved in their
utilisation (construction labour). Likewise, the time-
varying local energy purchase and feed-in tariffs.
Clearly much of the onus here will need to be placed
upon the user (although the ETHZ EcoInvent
database would be of use for embodied energy
content). However, one way of maximising the re-use
of this effort would be the use of an on-line open data
respository – providing open access to users’ data.
Further resource flows
The appliance model which is currently under
development for integration within CitySim has thus
far been calibrated exclusively to model electrical
appliances. It does not currently model water
consumption from appliances that consume
exclusively water or both water and electricity.
It is thus planned to develop a simplified model of
water flows, including:
- The consumption of water in buildings,
distinguishing the quality of water required to
support the activity concerned as well as the
quality of output waste water.
- Evapotranspiration from vegetated surfaces and the
associated consumption of water for irrigation.
- Harvesting and storage of rainwater and recycling
of consumed water.
- Supply of water from recycled, harvested and
mains sources.
Modelling of surface water drainage is currently
considered to be beyond the scope of CitySim.
As part of this water processing module the
anaerobic digestion of human waste will also be
modelled in a simplified way. Combustible biogas
and solid waste (predicted using the simplified solid
waste model mentioned under ‘core models’) will
then be available for input to the ECS models.
In the longer term, as we increase our scale of
analysis as well as the diversity of building types that
are modelled, it will be of special interest to model
possible synergetic exchanges of energy and matter
between buildings or the activities accommodated
within them. In this way it will be possible to test
hypotheses inspired by industrial ecology regarding
ways of minimising net urban resource use, through
improved circularity in their flows.
Consider the industrial park on the outskirts of the
town of Kalundborg in Denmark for example.
Amongst the numerous synergetic exchanges, the
waste heat produced by the power station meets the
space heating and hot water demands of the town’s
buildings and a fish-farm as well as the process heat
needs of a bioplant and an oil refinery. The bioplant
produces yeast for pig farmers as well as
fermentation sludge for local farmers. Gypsum from
the power station’s flue gas desulphurisation plant is
a key raw material of a plasterboard
manufacturer…and so on. The reduction in resource
imports and exports are considerable. These
principles are readily applicable to urban
developments, which tend to accommodate rich sets
of energy and matter flows.
- 1087 -
Since the processes involved may be represented in
an aggregate way (as is typical with mass flow
analysis), representing the potential exchanges need
should not be difficult; provided of course that the
necessary data is available.
Urban climate modelling
Compared with rural settings, in the urban context
more shortwave radiation is absorbed, less longwave
radiation is emitted and the mean wind speed is
lower, so that the mean air temperature is higher.
This urban heat island is exacerbated by
anthropogenic heat sources and the relative lack of
evapotranspiration due to vegetated surfaces. Due to
inertial differences, urban and rural temperature
profiles also tend to be out of phase, so that whilst
the mean urban temperature is higher, afternoon
temperatures may be lower, particularly in summer.
These urban-rural temperature differences can have
significant implications for predicted resource
demands and so needs to be account for in our
simulations.
Although the influencing mechanisms are reasonably
well understood, predicting the urban climate is
complicated by the scales involved: from buildings
within the urban canopy (the size of a few meters) to
large topographical features such as nearby water
bodies or mountains (the size of a few kilometres).
These scales cannot be satisfactorily resolved in a
computationally tractable way using a single model.
Our solution to this problem has been to couple
different models which each address different spatial
scales.
Firstly, freely available results from a global (macro)
model are input to a meso-model at a slightly larger
scale than that of our city. This meso-model is then
run as a pre-process to interpolate the macro-scale
results at progressively finer scales until the
boundary conditions surrounding our city are
resolved at a compatible scale. The meso-model may
then be run in the normal way. In the rural context
this may simply involve associating topography and
average land use data with each cell, the former
affecting temperature as pressure changes with height
the latter affecting temperature due to
evapo(transpir)ation from water bodies or vegetated
surfaces. In the urban context however, it is
important to account for the energy and momentum
exchanges between our built surfaces and the
adjacent air, which implies some representation of
3D geometry. For this we use a new urban canopy
model in which the velocity, temperature and scalar
profiles are parameterised as functions of built
densities, street orientation and the dimensions of
urban geometric typologies. These quantities are then
used to estimate the sources and sinks of the
momentum and energy equations.
Thus, a completely coupled macro, meso and urban
canopy model can be used to predict the temperature,
wind and pressure field in a city taking into account
not only the buildings from which it is composed but
also the scales which are bigger than the city itself.
This new modelling capability, which is described in
Rasheed et al (2009), may be run as a pre-process to
modify the climatic inputs to CitySim.
Evolutionary algorithms
For a new urban development, even with a relatively
limited number of variables (geometry, type of use,
occupancy and constructional characteristics, plant
and energy supply technologies), the number of
permutations is very large. The probability of
identifying an optimal configuration of these
variables by manual trial and error or simple
parametric studies is thus correspondingly small.
Moreover the response function computed by
CitySim may exhibit a non-linear, multi-modal and
discontinuous behaviour. Therefore heuristic
methods such as Evolutionary Algorithms are needed
to overcome possible local optima, keeping in mind
that we can never be sure of finding the global
optimum in a finite time frame.
Following from a review of available evolutionary
algorithms (EAs), we have developed a hybrid of two
algorithms (Covariance Matrix Adaptation –
Evolutionary Strategy [CMA-ES] and Hybrid
Differential Evolution [HDE]) which offers improved
robustness over its individual counterparts for a
larger range of optimisation problems (Kämpf and
Robinson, 2009a,b).
After having applied this new algorithm to a range of
solar radiation optimisation problems, we have
recently coupled this with CitySim and deployed it to
explore ways of optimising the energy performance
of part of a district of the City of Basel in
Switzerland, called Matthäus (see Kämpf and
Robinson, 2009c). Work will shortly start on the
integration of an interactive tool within CitySim with
which users will be able to select parameters to
optimise, together with the constrained range of
values they may take and the criteria with which
overall fitness is to be judged.
Uncertainty analysis
Finally, it is likely that the uncertainty in some of the
inputs to urban models, particularly in relation to the
conceptual design of new developments, may have a
significant impact on performance predictions and
the conclusions reached from these predictions.
To be able to accommodate these uncertainties in
urban scale predictions, we intend to borrow
techniques developed for the simulation of individual
buildings (e.g. De Witt (2001) and Macdonald
(2002)). A facility should then be provided within the
interface to define the uncertainties of the parameters
to which predictions are found to be particularly
sensitive, and to select these parameters for
simulation purposes.
- 1088 -
MULTI- AGENT TRANSPORT
SIMULATION (MATSIM)
Resources in urban settlements are consumed
principally by buildings and the activities
accommodated with them and by the transport of
goods and people to and between them. To account
for this transport related energy use, we use the
Multi-Agent Transport Simulation Toolkit MATSim-
T (1990).
MATSim-T simulates the sub-hourly transport of
individual people within a given urban scene.
For this a geometric description of the scene is
required, consisting primarily of the transport
network nodes and the links between them as well as
the locations of activities (work, home, education etc)
located at or adjacent to these nodes / links. Using
geo-coordinated census data a population of
households may then be created which is then
populated with individuals and associated with
attributes such as ID, driving licence ownership, car
availability, etc. Activity chains1 such as home-work-
leisure-work-home and the locations and preferred
timing of these activities are then associated with
these agents. With this initialisation complete the
scene may be simulated (Figure 6).
Figure 6: MATsim-T simulation of traffic results fo
Zürich, Switzerland (from: www.matsim.org).
First the travel time of alternative plausible routes
between the origin and destination of each journey is
calculated to determine the least cost routes. This is
carried out for all activity chains of all agents until all
journey plans have been deduced. A stochastic
queue-based traffic simulation then simulates each
agent’s journeys throughout the network. The actual
arrival time at each destination then depends on the
degree of network congestion.
Next a score is associated with the utility of the
achieved daily travel plan, considering the utility of
the time spent performing the required activities, the
time spent in travel and whether or not these
activities were started late. Using an evolutionary
algorithm in which each member of a given
generation corresponds to a new travel plan the next
travel plan in line is then simulated. The success of
these plans are then ranked. The worst performing
1 e.g. derived from Swiss microcensus data
ones are discarded and those that are retained are
used to create the next generation. This process
continues for a defined number of iterations until a
daily plan of near optimal utility has been identified
for each agent. This corresponds to the way in which
humans learn from experience what is the most
efficient time to leave one destination for another and
according to what route. For example we may choose
to leave home rather early in the morning to arrive at
work quickly and also leave earlier to avoid heavy
afternoon traffic.
With agents’ daily travel plans chosen, their journeys
may be simulated and the associated fuel
consumption calculated, using empirical performance
data.
Now to couple CitySim and MATSim all that is
required is that the two tools share a common XML
file holding (amongst other variables) building IDs
and the characteristics of the occupants. By running
MATSim as a pre-process to CitySim the means for
data exchange will be the arrival and departure time
of occupants – in effect replacing the current
stochastic occupancy model.
As noted earlier, MATSim’s initialisation process
associates a range of attributes to each individual
agent. These include age, gender and salary. To these
we may add other attributes such as environmental
preferences. In a CitySim pre-process we may then
divide apartment buildings into households of
appropriate size and allocate appliances to these
households. We may refine our stochastic models so
that behaviour is predicted as a function of say
preferred summertime temperature or indoor
illuminance…etc. In short, a coherent way for
integrating MATSim and CitySim is through the
multi-agent stochastic simulation of behaviour.
Work has recently started in developing this new
MAS behavioural modelling environment. It is hoped
that this modelling utility will be integrated into
future single building as well urban simulation
programs.
CONCLUSIONS AND OUTLOOK
Considerable progress has been made in recent years
to simulate the availability of renewable energy
within the urban context as well as the more general
demand for and consumption of resources (energy,
water and waste); culminating in the recent
development of CitySim, to which this paper relates.
But their remains a great deal of work to do if we are
to model urban resource flows, due to both buildings
and transport, in a truly comprehensive way, together
with means for optimising them. This is however a
very exciting challenge and one which, should we
succeed, would produce a model of acute value to
future planners of urban settlements, both new and
existing!
- 1089 -
ACKNOWLEDGEMENT
Financial support received for this work from the
European Commission as part of the CONCERTO II
Project HOLISTIC is gratefully acknowledged.
REFERENCES
Clarke, J.A. (2001), Energy Simulation in Building
Design, Butterworth-Heinemann: Oxford.
de Wit, M.S. (2001), Uncertainty in prediction of
thermal comfort in buildings, PhD Thesis, Delft
University of Technology, Delft.
Girardet, H. (1999), Creating sustainable cities,
Schumacher Briefing 2, Green Books.
Haldi, F. Robinson, D. (2009a), A comprehensive
stochastic model of window opening behaviour,
Building and Environment, in press
(doi:10.1016/j.buildenv.2009.03.025).
Haldi, F. Robinson, D. (2009b), A comprehensive
stochastic model of window usage: theory and
validation, ibid.
Haldi and Robinson (2009c) , A comprehensive
stochastic model of blind usage: theory and
validation, ibid.
Hunt, D.R.G. (1979), The use of artificial lighting in
relation to daylight levels and occupancy. Building
and Environment (14), 21–33.
Kämpf, J., Robinson, D. (2007), A simplified thermal
model to support analysis of urban resource flows,
Energy and Buildings 39(4), p445-453.
Kämpf, J., Robinson, D. (2009a), A hybrid CMA-ES
and DE optimisation algorithm with application to
solar energy potential, Applied Soft Computing 2(9)
p738-745.
Kämpf, J., Robinson, D., Optimisation of building
form for solar energy utilisation using constrained
evolutionary algorithms, Energy and Buildings
(submitted)
Kämpf, J., Robinson, D. (2009c) Optimisation of
urban energy demand using an evolutionary
algorithm, ibid.
MATSim-T, 1990, www.matsim.org/
Nielsen, T. R. (2005), Simple tool to evaluate energy
demand and indoor environment in the early stages
of building design. Solar Energy 78 (1), pp. 73-83.
Lefebvre, G., Bransier, J., Neveu, A. (1987),
Simulation du comportement thermique d’un local
par des methodes numeriques d’ordre reduit, Revue
Generale de Thermique 26(302) 106–114.
Macdonald, I.A. (2002), Quantifying the Effects of
Uncertainty in Building Simulation. Ph.D. Thesis,
ESRU, University of Strathclyde.
Nielsen, T.R. (2005), Simple tool to evaluate energy
demand and indoor environment in the early stages
of building design, Solar Energy 78 (1), p73–83.
Page, J. (2007), Simulating occupant presence and
behaviour in buildings, Unpublished PhD Thesis,
EPFL.
Page, J., Robinson, D., Morel, N., Scartezzini, J.-L.
(2007a), A generalised stochastic model for the
prediction of occupant presence, Energy and
Buildings, 40(2) p83-98.
Page, J., Robinson, D., Morel, N., Scartezzini, J.-l.
(2007b), Stochastic simulation of occupant presence
and behaviour in buildings, Proc. Tenth Int. IBPSA
Conf: Building Simulation 2007, Beijing, China.
Perez, R., Seals, R., Michalsky, J. (1993), All-
weather model for sky luminance distribution––
preliminary configuration and validation. Solar
Energy 50 (3), 235–243.
Pigg, S., Eilers, M., Reed, J. (1996), Behavioural
aspects of lighting and occupancy sensors in privates
offices: a case study of a university office building.
In: Proc. ACEEE Summer Study on Energy
Efficiency in Buildings 8, pp. 8.161–8.171.
Rasheed, A. Robinson, D. (2009), Multiscale
modelling of the urban climate, ibid.
Reinhart, C. F. (2004), Lightswitch-2002: a model
for manual and automated control of electric lighting
and blinds, Solar Energy, 77(1), p15–28.
Reinhart, C.F. and Voss, K. (2003), Monitoring
manual control of electric lighting and blinds.
Lighting Research & Technology 35(3), p243–260.
Robinson, D., Stone, A. (2004), Solar radiation
modelling in the urban context, Solar Energy, (77)3,
p295-309.
Robinson, D., Stone, A. (2005), A simplified
radiosity algorithm for general urban radiation
exchange, Building Services Engineering Research
and Technology, 26(4), p271-284.
Robinson, D., Stone, A. (2006), Internal illumination
prediction based on a simplified radiosity algorithm,
Solar Energy, 80(3), p260-267.
Robinson, D., Campbell, N., Gaiser, W., Kabel, K.,
Le-Mouele, A., Morel, N., Page, J., Stankovic, S.,
Stone, A. (2007), SUNtool – a new modelling
paradigm for simulating and optimising urban
sustainability, Solar Energy, 81(9), p1196-1211
Tregenza, P., Sharples, S. (1993), Daylighting
algorithms. ETSU S 1350-1993, UK.
United Nations (2004). Department of Economic and
Social Affairs / Population Division. World
Urbanization Prospects: The 2003 Revision. New
York: United Nations, p.3.
- 1090 -