Post on 02-Apr-2023
SIMPLIFIED APPROACH FOR THERMO FLUID-DYNAMIC ANALYSIS
OF LOOPED DISTRICT HEATING NETWORK
Elisa Guelpa1, Vittorio Verda2*
1 Energy Department, Politecnico di Torino,
C.so Duca degli Abruzzi 24, 10129 Torino, Italy,
e-mail: elisa.guelpa@polito.it
2 Energy Department, Politecnico di Torino,
C.so Duca degli Abruzzi 24, 10129 Torino, Italy,
e-mail: vittorio.verda@polito.it
Keywords: district heating, modelling, compact model, black box model, network modelling, smart thermal grids
Abstract
DH network modelling constitute a crucial way for simulating the system responses to changes in operating
conditions. These modifications can be planned for increasing the DH efficiency (such as heat storage
installation or optimization of operations) or they can be unexpected, such as the occurrence of malfunctions
or pipeline breakups.
In this framework, an important issue regards the possibility of simulating of the entire DH networks. DH
networks can be fully modelled through a numerical approach, which considers fluid flow and transient
heat transfer. The model is based on conservation equations and it has been validated by comparison with
experimental data of the Turin district heating network. In some cases, i.e. when a) super-real time
simulations b) very large network c) optimization through heuristic methods are considered, faster
computational approaches can be useful for simulating thermal fluid-dynamic within the network. The goal
of this work consists in proposing a compact model for the simulation of DH network for obtaining a
significantly the computational costs without compromising its reliability.
1. Introduction
The competitiveness of DH derives from the use of various sources and high efficiency plants for house
heating and domestic hot water production in densely populated areas [1.2]. DH allows a higher exploitation
of renewable sources [3-5], efficient technologies [6, 7] and waste heat [8, 10]. Furthermore, the
simultaneous production of thermal energy and electricity significantly decreases the fuel consumption for
building heating with a consequent reduction of carbon emissions. As a consequence, district heating system
is one of the most convenient technology for house heating in urban areas [11, 12] and it allows a reduction
of primary energy consumption to about 50% with respect to the use of condensing boilers [13].
Optimal control, design and management [14, 15] are crucial points for reducing primary energy
consumption and pollutant production and increasing economic benefit from selling heat and electricity. In
this framework, modelling of DH system is a crucial point in order to analyze effects of changes in a) the
network topology b) system management c) operations d) power plants and storage capacity. Typical
applications deals with thermal storage [16-18] or virtual storage [19,20] installation. is For this reason
modeling has been widely applied in both DH design [21, 22] and management [23, 24] phases.
The most common modelling approaches are physical and black box approaches. Physical models rely on
the definition and solution of the physical problem on a computational domain, which is the entire DH
network pipeline. The non-linear problem that characterize the DHN fluid-dynamic is solved by numerical
methods that are often computationally expensive. Physical models require the topological description of
the network and allows evaluating mass flows and temperature distributions in the network.
In black box approaches, the nonlinearity is overcome using standard transfer functions or neural networks.
In order to build and validate these kinds of models, a certain amount of experimental data, or simulation
results, are necessary [25, 26].
In the literature various different types of physical models have been developed, such as loop equation
method [27], aggregated models [28] (applied to 20 km long network, with 1079 nodes and 10 MW of
maximum heat production) and node based model [29]. Both aggregated and node-based models do not
solve fluid-dynamic and thermal transient problems in all the network nodes [30]. In [30] the thermal fluid-
dynamic problem has been solved for a medium size network, neglecting heat losses.
When extended DH networks are considered, the computational cost becomes a crucial characteristic of a
simulation tool. This is due to the large number of nodes and branches included in the network. Indeed the
solution of the thermal fluid-dynamic problem may require very high computational resources. An
alternative approach has to be used in these cases.
As far as no modelling of extended network is presented in the literature, in the present paper the main
strength of the physical approach are combined with the compactness of the black box model. This is done
for achieving fast solution of fluid dynamic problem in extended DH networks. The circulation of the water
within the network loops is computed through a black box approach while mass flows and temperature are
evaluated by solving the mass and energy conservation equation for all the nodes of the network. It allows
keeping computational expenses as low as possible, with a negligible error.
2. Methodology
2.1 Physical model description (full model)
Physical description of DH networks usually rely on a one dimensional model based on the description of
the network through a graph based approach [31]. Namely, each pipe of the network is considered as branch
that starts from a node, the inlet node, and ends in another node, the outlet node. Incidence matrix A is used
to describe the connection between nodes and branches of the network. A general element Aij is equal to 1
or -1 if the branch j enters or exits the node i and 0 otherwise. A includes as many rows as the number of
nodes and as many columns as the number of branches.
The thermal fluid-dynamic model here used includes the mass and energy conservation equations applied
to all the nodes and the momentum conservation equation to all the branches. The mass balance equation
written using matrix form is:
0 extGGA , (1)
G is the vector including the mass flow rates in branches and Gext the vector includes the mass flow rates
exiting the nodes towards the extern.
The steady-state momentum conservation equation is considered for all the branches of the network,
considering an incompressible fluid, neglecting the velocity change within a single pipe and including the
gravitational term in the static pressure. This can be written in a matrix form as described in (6)
pumps
T pYPAYG , (2)
where the diagonal matrix Y represents the fluid dynamic conductance of branches:
1
2
1
2
1
k
kLD
f
S
GRY
(3)
The dependence of Y on pressure makes the system of equation non-linear, as already mentioned in the
introduction section. The solution of the mass and momentum equations is performed using a SIMPLE
(semi implicit method for pressure linked equation) algorithm [32], which is a common guess and correction
method use for solving Navier Strokes equations.
The energy conservation equation is written neglecting conduction in the fluid along the network and
volumetric heat release within the control volume. The energy equation is written in transient form because
the thermal capacity cannot be neglected and the temperature perturbations travel the network with the fluid
speed. The matrix equation, including an equation for each network nodes, assumes the following form:
𝑴�̇� + 𝑲𝑻 = 𝒈 (4)
Where M is the mass matrix that includes the coefficient of the unsteady term, K is the stiffness matrix,
including the terms multiply by T and g is the known term.
The model equations are detailed in the simple and matrix form in Table Further details on the method are
available in [33].
Mass conservation
equation
Momentum conservation equation Energy conservation equation
Extended form ∑𝐺𝑖𝑛 −∑𝐺𝑜𝑢𝑡 = G𝑒𝑥𝑡 (pin − pout) =1
2
f
DLG2
ρS2+1
2∑βkk
G2
ρS2− t
𝜕(𝜌𝑐∆𝑇)𝑖𝜕𝑡
∆𝑉𝑖 +∑𝑐𝐺𝑗𝑗
𝑇𝑗 = 𝑈𝑇𝑂𝑇(𝑇𝑖 − 𝑇𝑒𝑛𝑣)
Matrix form 𝐀 ∙ 𝐆 + 𝐆𝐞𝐱𝐭 = 0 𝐆 = 𝐘 ∙ 𝐀T ∙ 𝐏 + 𝐘 ∙ 𝐭 𝑴�̇� + 𝑲𝑻 = 𝒈
Solution Method SIMPLE SOLVER FOR LINEAR EQUATION
SYSTEM
Computational
cost (for the
examined
network)
1,7 s 10-?
Table 1 Full model details
2.2 Compact model description
As regards the thermal fluid-dynamic model discussed in section 2.1, the momentum equation is solved
mainly to catch the water distribution within the loop. In fact the knowledge of pressure is not mandatory
when the aim is the evaluation of the thermal request at the plants. This constitute a significant effort for
the computational point of view. The main question that authors want to answer in this paper is if an
alternative way exist to evaluate the water distribution in the loops in order to keep low computational costs.
The main idea of the work is to use a black box approach to evaluate the water distribution within the loop,
and to solve the continuity equation on an equivalent tree-shaped network.
In Fig. 1a a looped network is reported; it includes 3 loops. It is possible obtaining an equivalent tree shaped
network by eliminating a branch in each loop (as indicated in Fig. 1b). In lieu of each deleted branch, the
mass flow rate flowing in the branch has to be added as exiting the inlet node and entering the outlet node
of the examined branch (red arrows in Fig.1b).
The main issue is the evaluation of the mass flow circulating in the deleted branch. In this work, the mass
flow is considered as linearly dependent from the mass flow produced at each power plants. This procedure
allows significantly reduce the computational cost of thermal fluid dynamic model.
This approach does not provide the pressure distribution along the network because it overcomes the
solution of the momentum equation that is computationally expensive. Indeed the compact model solves
the mass and energy conservation equation in the tree-shaped network and provides temperature and
pressure in all the branches of the network.
Fig. 1 a) looped network b) equivalent tree-shaped network
3. System description
The network considered in this work is the Turin network, which is located in the north of Italy. It is the
largest network in Italy and one of the largest in Europe. This network has been used to validate the
proposed approach. The main system characteristics are:
5500 buildings supplied
total volume of 56 million m3
annual thermal request of 2000 GWh
maximum thermal power request is over 1.3 GW.
Total length about 2500 km
It includes more than 60000 nodes.
Fig. 2 Schematic o Turin transportation line. In red the main loops
The network outline is shown in Fig. 2. Thermal power is produced and stored in six thermal plants, that
are located in different areas of the network. There are 3 cogeneration groups, 5 groups of boiler systems
and 3 groups of storage tanks. The main characteristics of the power plants is reported in Table 2.
Plant Type Nominal Power [MW]
Torino Nord
Cogeneration 220
Boilers 340
Storage 150
Moncalieri 2 Cogeneration plants 520
Boilers 141
Politecnico Boilers 255
Storage 60
Martinetto Storage 60
BIT Boilers 255
Mirafiori Nord Boilers 35
Table 2 Characteristics of the thermal plants
The DH network is considered as composed of two parts: the transportation network and many distribution
networks. The first consists in pipes with large diameter, usually larger than 200 mm. The second one
includes the smaller pipes and it is used to distribute water to the buildings located in the same area. The
supply transportation pipeline is 70 km long and it includes 9 loops. Multiple loops are used to limit possible
effects of failures as well as to allow better flow distribution and reduce pumping effort. In the Turin
network there are 182 distribution networks. The nodes that connect the transportation line to each
distribution networks are called barycentres.
4. Results
In this section, mass flow rates obtained with both the full and the compact model have been reported and
compared. In Fig. 3 the value of mass flow rates in the deleted branches are depicted. Both mass flows
evaluated by solving the momentum equation and mass flow obtained through the simplified approach are
reported. Three different thermal loads have been considered in the analysis: 30%, 60% and 90% of the
maximum thermal load that can be provided to the network. The sign of the mass flow rates reported in
Fig. 3 is due to the direction of the flow respect to the reference one. It is possible to notice that in all the
considered cases the compact model correctly detects the direction of the flow. Furthermore, the error
performed in the prediction of the mass flow rate value is low in all the cases considered.
Fig. 3 Mass flow rates in the deleted branches, obtained through momentum equation and simplified
approach
In order to better visualize the performance of the compact model, the errors performed in mass flow rates
calculation are shown in Fig. 4. The figure reports the percentage relative error in the mass flow rate
evaluation for three load values. Only the cases when the mass flow rate is not negligible are considered in
the analysis. Error are always lower than 5 %. These results are particularly encouraging as regards the use
of simplified method to compute the mass flows in looped network.
Fig. 4 Error in mass flow rates evaluation
Fig. 5 shows the mass flow rate circulating within the transportation network in various cases. Results are
reported for 3 different loads and for both the full and the compact approaches. It can be seen that the mass
flows entering the network in the power plants nodes increases, while increasing the thermal load. This is
especially clear in the main branches near the Moncalieri power plants (south) and near the Torino Nord
power plant (north). Indeed, the lines representing these nodes in Fig. 4, become more and more yellow as
the load increases. Fig. 4 allows also noticing that the mass flow rates predicted with the two presented
approaches are very similar in all the nodes of the network. In particular, any differences cannot be detected
in Fig. 4. Observing the various loops it is possible to notice that no differences in the amount of mass flow
rate computed occur.
Fig. 5 Mass flow rates distribution [kg/s] in all the branches of the transportation network
In Fig. 6 the absolute error obtained in the mass flow evaluation for all the branches of the network is
depicted. Maximum deviation is 36 kg/s in case with 30% of the thermal load, 31 kg/s in case with 60% of
the thermal load and about 20% in case of 90% thermal load. These values are very low if compared with
the mass flow rates circulating the network (in Fig. 5). Indeed, 36 kg/s, which is the maximum value,
represents about 1 % of the maximum mass flow rate circulating in a branch of the network. Furthermore,
it is possible to notice that the maximum error occurs when 30% partial load is considered and it decreases
while the load increases. This point is very important because the network simulations are mainly used to
reproduce the system behavior when a high request occurs.
Mean error in all the branches of the network are respectively 5.4 kg/s, 4.1 kg/s and 2.6 kg/s. These values
represents the 0.1-0.3% of the maximum mass flow rate circulating in a branch of the network.
Fig. 6 Absolute error in the evaluation of mass flow [kg/s] in the transportation network
Fig. 7 shows the thermal request of all the 182 distribution networks (in ascending order), evaluated with
both the full model and the compact model. Fig. shows that the thermal request are very similar. The average
relative error is about 6 %. However this result strongly depends on the thermal power required by each
building. In fact, when the thermal request is low, although the absolute error is high, the relative error can
be high. In order to overcome this problem the total load required to the network is considered.
Fig. 8 Absolute error in the evaluation of mass flow [kg/s] in the transportation network
In order to evaluate the error performed in the evaluation of the total load required, the sum of the thermal
request required to the various users is done. Results are reported in Fig. 8 for three different thermal loads.
The maximum error performed (load 60 %) is about 1%. In case of 30 % and 90 % of load request, the error
are significantly lower (0.83 % and 0. 32%). This is a very satisfying result especially if the the difference
in the computational cost that the two models require are compared.
Fig. 8 Absolute error in the evaluation of mass flow [kg/s] in the transportation network
As regards the computational cost, the full model solution requires about 1.5 s for each iteration on a single
3.3 GHz CPU. An iteration of the compact model requires about 0.0015 s. Ratio between the computational
time is about 10-3.
5. Conclusions
The current work presents a compact model for describing the thermal fluid dynamic behavior of DH
networks. The model relies on the main strength of both physical and black box approaches. It allows
obtaining mass flow rates evolution in all the nodes of the network with reduced computational resources.
The model is based on the idea of evaluating the water distribution within the loop through a black box
approach. This allows solving a looped network as an equivalent tree-shaped network. The continuity and
energy equations are solved on the tree-shaped equivalent network. This avoid high computational costs
required to the solution of the non-linear coupled equations (continuity and momentum).
The proposed approach has been tested to the Turin district heating network, one of the largest network in
Europe, in order to analyze the start-up transient. The model provides mass flow rates in all the branches
and then it can be used, coupled with a thermal model, the evolution of the temperatures in all the nodes.
The computational time to solve the fluid-dynamic model applied to the entire network is about 1.5 s. The
water distribution within the loops can be obtained by using the compact model in 0.0015 s. The compact
model detects the mass flow direction in all the network loops. The error performed by the compact model
respect to the full model is always lower than 5%.
In conclusion, the use of compact model properly defined, can be considered a useful toll to simulate
thermal fluid-dynamic behavior of large district heating networks with low computational cost.
Acknowledgments
This work has been conducted within the European Project H2020-LCE-2016-2017 PLANET (Planning and
operational tools for optimizing energy flows and synergies between energy networks)
References
[1] Lund H., Moller B., Mathiesen B.V., Dyrelund A, The role of district heating in future renewable energy
systems. Energy 2010; 35: 1381–1390.
[2] Verda, V., Guelpa, E., Kona, A., & Russo, S. L. (2012). Reduction of primary energy needs in urban
areas trough optimal planning of district heating and heat pump installations. Energy, 48(1), 40-46.
[3] Fahlén, E., & Ahlgren, E. O. (2009). Assessment of integration of different biomass gasification
alternatives in a district-heating system. Energy, 34(12), 2184-2195.
[4] Lindenberger, D., Bruckner, T., Groscurth, H. M., & Kümmel, R. (2000). Optimization of solar district
heating systems: seasonal storage, heat pumps, and cogeneration. Energy, 25(7), 591-608.
[5] Yildirim, N., Toksoy, M., & Gokcen, G. (2010). Piping network design of geothermal district heating
systems: Case study for a university campus. Energy, 35(8), 3256-3262.
[6] Casisi, M., Pinamonti, P., & Reini, M. (2009). Optimal lay-out and operation of combined heat & power
(CHP) distributed generation systems. Energy, 34(12), 2175-2183.
[7] Ziębik, A., & Gładysz, P. (2012). Optimal coefficient of the share of cogeneration in district heating
systems. Energy, 45(1), 220-227.
[8] Fang, H., Xia, J., Zhu, K., Su, Y., & Jiang, Y. (2013). Industrial waste heat utilization for low
temperature district heating. Energy policy, 62, 236-246.
[9] Holmgren, K. (2006). Role of a district-heating network as a user of waste-heat supply from various
sources–the case of Göteborg. Applied energy, 83(12), 1351-1367.
[10] Coss S, Guelpa E, Rebillard C, Verda V, Le-Corre O. Industrial waste heat integration for providing
energy service to district heating networks. Proceedings of ECOS 2016. Portoroz, Slovenia, June 19-23
June 2016.
[11] Li, Y., Fu, L., Zhang, S., & Zhao, X. (2011). A new type of district heating system based on distributed
absorption heat pumps. Energy, 36(7), 4570-4576.
[12] Sciacovelli, A., Guelpa, E., & Verda, V. (2014). Multi-scale modeling of the environmental impact and
energy performance of open-loop groundwater heat pumps in urban areas. Applied Thermal
Engineering, 71(2), 780-789.
[13] V. Verda, G. Baccino, A. Sciacovelli, S. Lo Russo (2012). Impact of District Heating and Groundwater Heat Pump
Systems on the Primary Energy Needs in Urban Areas. Applied Thermal Engineering.
10.1016/j.applthermaleng.2012.01.047
[14] Benonysson, A., Bøhm, B., & Ravn, H. F. (1995). Operational optimization in a district heating system.
Energy conversion and management, 36(5), 297-314.
[15] Brundu, F. G., Patti, E., Osello, A., Del Giudice, M., Rapetti, N., Krylovskiy, A., ... & Acquaviva, A.
(2017). IoT Software Infrastructure for Energy Management and Simulation in Smart Cities. IEEE
Transactions on Industrial Informatics, 13(2), 832-840.
[16] Wang, H., Yin, W., Abdollahi, E., Lahdelma, R., & Jiao, W. (2015). Modelling and optimization of
CHP based district heating system with renewable energy production and energy storage. Applied Energy,
159, 401-421.
[17] da Cunha, J. P., & Eames, P. (2016). Thermal energy storage for low and medium temperature
applications using phase change materials–a review. Applied energy, 177, 227-238.
[18] Sciacovelli, A., Guelpa, E., & Verda, V. (2014). Second law optimization of a PCM based latent heat
thermal energy storage system with tree shaped fins. International Journal of Thermodynamics, 17(3), 145-
154.
[19] Guelpa, E., Barbero, G., Sciacovelli, A., & Verda, V. (2017). Peak-shaving in district heating systems
through optimal management of the thermal request of buildings. Energy, 137, 706-714.
[20] Verda, V., Guelpa, E., Sciacovelli A., F. G., Acquaviva, A., & Patti. (2016). Thermal peak load shaving
through users request variations. International Journal of Thermodynamics, 19(3), 168-176.
[21] Dalla Rosa A, Li H, Svendsen S. Method for optimal design of pipes for low-energy district heating,
with focus on heat losses. Energy 2011; 36(5): 2407-2418.
[22] Ancona MA, Melino F, PerettoA. An Optimization Procedure for District Heating Networks. Energy
Procedia 2014; 61: 278-281.
[23] Li L, Zaheeruddin M. A control strategy for energy optimal operation of a direct district heating
system. International Journal of energy research 2004; 28(7): 597-612.
[24] Sciacovelli, A., Guelpa, E., & Verda, V. (2013, November). Pumping cost minimization in an existing
district heating network. In ASME 2013 International Mechanical Engineering Congress and Exposition
(pp. V06AT07A066-V06AT07A066). American Society of Mechanical Engineers.
[25] Keçebaş A, Yabanova İ. Thermal monitoring and optimization of geothermal district heating systems
using artificial neural network: A case study. Energy and Buildings 2012;50: 339-346.
[26] Guelpa E, Toro, C, Sciacovelli A, Melli R, Sciubba E, Verda, V. Optimal operation of large district
heating networks through fast fluid-dynamic simulation. Energy 2016;102: 586-595.
[27] StevanovicVD, Prica S, Maslovaric B, Zivkovic B, Nikodijevic S. Efficient numerical method for
district heating system hydraulics. Energy Conversion and Management 2007; 48: 1536–1543.
[28] Larsen HV, Palsson H, Bøhm B, Ravn HF. Aggregated dynamic simulation model of district heating
networks. Energy Conversion and Management 2002; 43: 995–1019.
[29] Benonysson A, Bohm B,Ravn HF. Operational optimization in a district heating system. Energy
Conversion and Management 1995; 36: 297-314.
[30] StevanovicVD, Zivkovic B, Prica S, Maslovaric B, Karamarkovic V, Trkulja V. Prediction of thermal
transients in district heating systems. Energy Conversion and Management 2009; 50: 2167–2173.
[31] HararyF.,GraphTheory.Narosa Publishing House, New Delhi;1995.
[32] PatankarS.V. Numerical Heat Transfer and Fluid Flow;1980.
[33] Guelpa E., Sciacovelli A., & Verda V. (2017). Thermo-fluid dynamic model of large district heating
networks for the analysis of primary energy savings. Energy.