1、2010 ASHRAE 11ABSTRACTPlanned increases in train frequency and the use of airconditioned rolling stock on the London Underground Sub-Surface lines will increase the energy dissipated in the tunnelsand an assessment has been made of the potential impacts ofthis on tunnel temperatures. The Cooling the
2、 Tube Programmein London has been established to help manage thermal condi-tions on the network. An important early step is to identify thehot spots in the existing system, and predict the future envi-ronment in light of changes to train operations and expectedpassenger numbers. Aero-thermodynamic m
3、odelling of theindividual lines was considered to be essential in this process.This paper discusses the tunnel ventilation modelling ofthe southern portion of the sub-surface lines (District andCircle) during normal train operations. The line has severalnatural ventilation openings of various sizes
4、and orientations.The modelling is particularly challenging due to the bi-direc-tional nature of the tunnels. Important considerations whenbuilding datasets, modelling sensitivities, collecting and inter-preting temperature data, the effect of ground water migration,validation approaches and the perf
5、ormance of the model ispresented.INTRODUCTIONARUP has provided Cooling the Tube Programme (CTP)with tunnel ventilation modelling for the Sub-Surface Line(SSL) section of the London Underground. The modelincludes about 8.3 km (13.3 miles) of double-track Districtand Circle line tunnels and eleven sta
6、tions from west of SloaneSquare to east of Whitechapel Stations.London Undergrounds Circle line runs along the align-ment of the oldest underground railway in the world. Builtwith steam trains in mind, the tunnels have historically beenquite cool and well ventilated by the blow holes used to relieve
7、the steam. These blow holes now function as natural ventopenings to reduce tunnel temperatures.A Subway Environment Simulation (SES) model wasdeveloped for the existing trains and future upgraded rollingstock running on the southern portion of the SSL to estimatethermal conditions under normal and c
8、ongested operations.The model was validated by comparison to present dayconditions. The benchmarked model could then be applied tofuture train operations to estimate the air temperatures infuture, identify the warmer sections, and if needed, proposecooling measures to mitigate the high temperatures
9、encoun-tered.The simulations were aimed at predicting the peak airtemperatures at 5 P.M. peak operation on a relatively hot Julyafternoon in the selected portion of SSL.This paper presents selected key modelling issues relatingto the SSL and the specific techniques applied to tackle these.A validati
10、on or benchmarking process matching results totemperature monitoring data was used to tune these techniquesfor the current case. This gave more confidence in the predic-tions of the modelling for future train operations. The mainissues on the SSL are:Bi-directional tunnelsGround water movementFuture
11、 air-conditioned rolling stockThe paper discusses the benchmarking process andsome key sensitivities, but does not cover the modelling ofthe system upgrades except for the impact of air-condition-ing.Natural Ventilation in London Underground Sub-Surface LinesModelling for Normal OperationsJohn Alexa
12、nder Mohammad Tabarra, PhD, DIC, CEngMember ASHRAEJohn Alexander is an engineer and Mohammad Tabarra is an associate at ARUP, England.OR-10-002 2010, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions 2010, Vol. 116, Par
13、t 1. For personal use only. Additional reproduction, distribution, or transmission in either print or digital form is not permitted without ASHRAEs prior written permission. 12 ASHRAE TransactionsMODELLING BI-DIRECTIONAL TUNNELSThe SSL is a double-track or bi-directional tunnel, serv-ing both the ea
14、stbound and westbound trains. This results incomplex patterns of airflows and pressure peaks as trains passeach other, and poses the difficulty of modelling a three-dimensional airflow field using a one dimensional model.Several approaches were investigated for the SSL.Any one dimensional network mo
15、del can be challenged inproviding accurate temperature predictions with bi-direc-tional tunnels, where the passing trains can cause a cancella-tion (partial or complete) of train piston effects, and generatelow or null flow regions in the tunnels of the model. Therunning of trains “like clockwork” i
16、n an SES model can resultin trains passing each other at exactly the same point for theduration of a simulation, creating an enduring low flow regionin which heat can build up. The heat transfer in a one dimen-sional model is based on bulk airflow movement, and will beminimal in low flow sections. I
17、n reality, turbulent eddies andchurning of the air by the passage of trains could give rise tomore heat transfer to the walls than would necessarily bepredicted by SES. The low blockage ratio of the trains in theSSL also contributes to low bulk airflow velocities. This limi-tation can cause numerica
18、l instabilities and/or result in artifi-cially high temperature predictions. This is not usually anissue for unidirectional tunnels where trains travel in onedirection and higher blockage ratios result in higher bulkairflow velocities and heat transfer.This effect was noted early on and steps were t
19、aken toavoid this from happening with the SSL model. Differentheadways or time delays were used for the eastbound andwestbound routes to compensate for piston effect cancellationand avoid trains repeatedly crossing at the same point in thesimulation.This method has been used with both weekly average
20、(long-term) train operations and peak (short-term) train oper-ations to ensure reasonable temperatures and avoid “hot spots”in the model. It should be noted that in SES 4.1 it is usual torun “long-term” simulations in which wall temperatures arecalculated with peak train operations. The default heat
21、 sinkattenuation factor of 0.5 accounts for off-peak operationsallowing an overall daily average of 50% of peak-hour heattransfer into the walls. In SES SI an improved approach isavailable. Long-term simulations are run with average trainoperations and a heat sink attenuation factor set to 1. Wallte
22、mperatures from this simulation are then fixed and used in ashort-term model with peak train operations. This allows heattransfer to be more accurately based on actual average trainoperations but also provides flexible inputting for futurechanges in both peak and off-peak operations.Several compensa
23、tion methods were identified and arediscussed below.Staggering MethodIn this method the simulation for the long-term walltemperature calculation is repeated several times and averaged.In each repetition, the starting time of the routes in one of thedirections, say westbound, is delayed by a fraction
24、 of the head-way. This is repeated until a full set of headways are coverede.g., 0 delay, 1/4, 1/2, and 3/4 headway delay. The resultingwall temperatures are averaged to smooth out any hot/coldspots. The method is repeated for the short-term air tempera-ture simulations to give the result.Unequal He
25、adway MethodsThe headways of the eastbound and westbound routes areset differently. The system headway for the simulation compu-tation is the lowest common multiple of the two headways. Inthis scenario the eastbound and westbound trains cross atdifferent points continually, giving a better average s
26、ystemperformance. There are two slightly different approachespossible.1. Headways several secondsHere the EB and WB headways are slightly altered fromthe calculated value. For example on the SSL the averageheadway of 180s may be set to 178s WB and 182s EB. Thispreserves the average system headway an
27、d at first glanceappears to be a reasonable approach. However the resultingsystem period (the lowest common multiple of the headways)will be large. This is particularly the case for the SSL whereseveral types of rolling stock operateboth Circle line Cstock trains and District line D stock trains. It
28、 was found thatif large summary periods are used that the simulation mayshow erratic behaviour or nonconvergence i.e., the SES arraylimits may be breached. Following this, a second method wasdeveloped.2. Trains per hour (TPH) 1 TPHIn this approach the number of trains per hour between EBroutes and W
29、B roots are incremented while maintaining thetotal number of trains (and hence heat load) per hour in thesystem. The headways are calculated from the following:EB TPH = Average TPH + 1WB TPH = Average TPH 1For the current SSL the 28 TPH peak operations translateto 27 TPH eastbound and 29 TPH westbou
30、nd or vice-versa.Different EB and WB headways prevent the EB and WBtrains from crossing at the same position repeatedly to createaerodynamically stagnant hot points. The average number oftrains per hour in the whole system is preserved. This gives asystem headway of one hour for the computations.It
31、is important for the trains to be in the same position atthe start and end of a summary period if the heat sink calcu-lations are carried out, to ensure convergence of the calcula-tions. The above procedure would normally be suitable forapplication to both the long-term and short-term calculationsfo
32、r a bi-directional system. However, the SSL train sequencingof three D stock trains interspersed by a C stock train wouldnot repeat over the hour due to the different headways. So this 2010, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published i
33、n ASHRAE Transactions 2010, Vol. 116, Part 1. For personal use only. Additional reproduction, distribution, or transmission in either print or digital form is not permitted without ASHRAEs prior written permission. ASHRAE Transactions 13method is not to be applied to the long-term model where heatsi
34、nk calculations are carried out.Additionally a large summary period is required to ensuresame trains position at start and end which can significantlyincrease computational run time.Approach for SSLFor the reasons discussed above, a hybrid of the twoapproaches detailed was used for the SSL. It was d
35、ecided thatthe staggering method should be used for long-term simula-tions and different headways (Average TPH 1) for the short-term simulations. Some sensitivities carried out actuallyshowed minimal differences between the two methods.The simulation is repeated four times, each time with thesame he
36、adway. The starting time of the first westbound (WB)train is varied each time by a quarter a of train headway. Thisensures that the positions where the WB trains cross the east-bound (EB) trains are different between staggers. The walltemperatures for each of the simulations are extracted from there
37、start files.The wall and air temperature result for the long-termsimulation is the averaged of these four individual simula-tions. The averaged long-term wall temperature is used in theshort-term simulation.The short-term simulations model the peak operationsreflecting increased train frequency. Wal
38、l temperatures (Envi-ronmental Control Analysis) are not recalculated in short-termsimulations because it is assumed that the wall temperaturewould not vary greatly during this period. The average walltemperatures calculated from the long-term simulation areused for the calculation of air temperatur
39、es in the short-termsimulations.BENCHMARKINGThe first objective of the SSL study was to produce arepresentative model of the SSL during 2006, known as thebenchmark (BM) model. The BM model was then used topredict the temperatures of the future operations if no mitiga-tion strategy is implemented i.e
40、. cooling solutions. This futuremodel without cooling was named the “Do Nothing” (DN)model.The BM model must reflect the system conditions undercurrent operations as accurately as possible, if it is to be usedreliably to predict the effects of future operations. There isalways a risk that some assum
41、ptions or estimates would causethe BM models result to differ markedly from reality. A rigor-ous benchmarking process, in which the model outputs arematched to measured data, reduces this risk and gives greaterconfidence in the SES model of the SSL.Temperature Data CollectedFour sets of temperature
42、measurements were used toassist the verification of the BM:The main data used for benchmarking was 2006 plat-form temperatures in Victoria and Embankment stationas the train operations modelled were also based on2006 timetables. These were collected by Tiny Tag dataloggers (TT).Supplementary data: 2
43、008 TT platform temperatures inVictoria, Embankment, Blackfriars, Mansion House,Monument, and Tower Hill stations.Supplementary data: 2008 TT tunnel temperaturesbetween Westminster and Embankment station, andbetween Temple and Blackfriars station.Additionally, a set of spatial temperature variation
44、pro-files along the platforms of each station was collectedduring two hot July peak afternoons in 2008 whichreached and surpassed the design ambient temperatureof 26.8C (80.2F).Station Tiny Tag DataTemperature loggers (TT) placed within the stations bythe CTP recorded the air temperatures every 5 or
45、 15 minutes,depending on the logger.Two TT were placed on the SSL platforms of Victoria andEmbankment Stations from the design year, 2006. The datacollected from these TT were used to validate the benchmark-ing model, so that the simulation output of the model wouldmatch the temperatures at those tw
46、o points at design condition.This would act as a fixing point for the simulations. Externalambient data was available for Victoria Station and was usedfor normalizing platform temperatures at both Victoria andEmbankment to the design temperature. The correlationbetween the ambient temperature and th
47、e platform tempera-tures of the two TT are presented in Figure 1. Each platformTT data point is plotted against the ambient TT data point atthe same time. Assuming linear relationships between thetemperatures of platform and ambient air, a linear regressionline was fitted to the points.Figure 1 Plat
48、form versus ambient temperature for Victoriaand Embankment Stations (2006). 2010, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions 2010, Vol. 116, Part 1. For personal use only. Additional reproduction, distribution, o
49、r transmission in either print or digital form is not permitted without ASHRAEs prior written permission. 14 ASHRAE TransactionsThe analysis showed that at the external design ambientair temperature of 26.8C (80.2F), the platform air tempera-tures are very close to ambient. At lower ambient tempera-tures, platforms were warmer than outside. At highertemperatures, the data suggested that the District and Circleline platforms would be cooler than outside. This is likely dueto the heat sink of the tunnel walls and the infrequency ofoutside temperatures exceeding 27C (80.6F
