Accepted Manuscript

Micro hydro power generation from water supply system in high rise buildings
using pump as turbines

Jiyun Du, Hongxing Yang, Zhicheng Shen, Jian Chen

PII: S0360-5442(17)30381-X

DOI: 10.1016/j.energy.2017.03.023

Reference: EGY 10490

To appear in: Energy

Received Date: 25 August 2016

Revised Date: 30 January 2017

Accepted Date: 02 March 2017

Please cite this article as: Jiyun Du, Hongxing Yang, Zhicheng Shen, Jian Chen, Micro hydro power
generation from water supply system in high rise buildings using pump as turbines, Energy (2017),
doi: 10.1016/j.energy.2017.03.023

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights

1. The performance of pump as turbine (PAT) applied in water supply pipes is studied.

2. The PAT has a good performance in head reduction and hydropower generation.

3. A PAT installation and control strategy is proposed and tested.

4. The PAT installation and control strategy could avoid excessive head reduction.
 ACCEPTED MANUSCRIPT

1 Micro hydro power generation from water supply system in high rise buildings
2 using pump as turbines
3
4 Du Jiyuna, Yang Hongxinga, Shen Zhichenga, Chen Jianb
5 a Renewable Energy Research Group, Department of Building Services Engineering, The
6 Hong Kong Polytechnic University, Hong Kong, China
7 b School of Energy and Power Engineering, University of Shanghai for Science and
8 Technology, Shanghai, China
9
10 Abstract
11 Water supply is of vital importance for urban development, so the overall pressure in urban
12 fresh water pipeline is usually very high to ensure consistent water supply throughout the urban
13 area. Currently, pressure reducing valves (PRVs) are widely used to reduce water pressure in
14 the water supply system (WSS) of high rise buildings to avoid water leakage and other danger
15 in Hong Kong. The application of pump as turbine (PAT) appears as an ideal way for water
16 head reduction as well as hydropower generation in water pipelines. In the present paper, the
17 feasibility and performance of PAT used in WSS for electricity generation was investigated. In
18 the research process, a PAT was first selected by empirical equations based on the given
19 working conditions in one typical high rise building. After selection, CFD simulation and
20 laboratory tests were carried out to study the performance of the selected PAT. Both the
21 simulated and measured results show that the selected PAT is feasible for power generation
22 and water head reduction. The comparison between numerical and experimental results
23 indicate that the simulation outcomes could be used for performance prediction of PAT
24 although the deviation exists. A PAT installation and control strategy was designed and tested
25 in the hydraulic test rig, and the results show that the proposed PAT installation and control
26 strategy performed well for avoiding excessive water head reduction at higher flow rates.
27
28 Keywords: Pump as turbine; water supply system; high rise building; micro hydro power
29 generation; water head reduction
30
 ACCEPTED MANUSCRIPT

31 1 Introduction
32 High rise buildings have obtained considerable development in the past decades due to global
33 urbanization and limited building land in cities. In Hong Kong, which is a densely populated
34 hilly city, most government funded public rental buildings are usually 30-40 storeys high and
35 the average height of residential buildings is about 26 storeys [1,2]. Compared to traditional
36 low rise buildings, high rise buildings possess many advantages, e.g. high living quality, less
37 land occupation and better use of daylight and thermal mass [3,4], however, they also put
38 forward higher requirement for energy and water supply [5]. Usually, in order to keep normal
39 water supply through the whole urban area, water pressure in urban mains is relatively high,
40 but the pressure is not high enough to send water to topmost users of high rise buildings, so
41 fresh water from urban mains has to be pumped up to the gravity-fed water tanks located on
42 the rooftops of high rise buildings and then distributed through down feed pipes to users [6].
43 Before pumping, the water pressure must be reduced to a proper value (about 10-45m water)
44 to minimize water leakage and damage of water supply appliances. The commonly used
45 technique for pressure reduction in water supply system (WSS) is the application of pressure
46 reduction valve (PRV). However, the PRV reduces overall water pressure in pipeline by
47 decreasing throttling area to increase water head losses, which means the excess energy is
48 wasted and the whole energy efficiency in WSS would be adversely effected [7].
49
50 Instead of dissipating excess energy, energy recovery from WSS using micro hydro turbines
51 or pumps as turbines (PATs) has attracted increasing attention as an effective way to control
52 pressure level in fresh water supply pipelines [8]. Generally, micro hydro turbines are
53 categorized into impulse turbine (i.e. Pelton, Turgo and Crossflow turbines) and reaction
54 turbines (i.e. Francis, Propeller and Kaplan turbines) [9,10]. Pelton and Turgo turbines can be
55 used at high or medium heads, but they are commonly installed in open environments and the
56 runner is not submerged in water [11], so these turbines are not suitable to be used in the WSS.
57 Based on our previous research, the crossflow turbine has a good performance when used in
58 the WSS to generate electricity, but its performance on water head reduction is very poor [12].
59 Similar to crossflow turbines, the propeller turbine is proved a promising device for electricity
60 generation in the WSS, but the head reduction through it is far less than the demand of water
 ACCEPTED MANUSCRIPT

61 supply in high rise buildings [13]. When applied in micro sites, the efficiency of Francis
62 turbines is usually between 30% and 60%, however, the cost in design and manufacture of
63 Francis turbines is relatively high [8,13].
64
65 Compared to traditional turbines, the PATs, which means centrifugal pumps operate inverse as
66 turbines for electricity generation, have important advantages, such as available in the market,
67 low investment cost, easy to install and maintain, etc [9]. Furthermore, according to the study
68 of Ramosetal. [14], the PAT could reach a relatively high efficiency (up to 85%). A case study
69 in Murcia, Spain indicated that the maximum installation capacity in the main pipe of water
70 supply with flow rate 300m3/h and water head 30m could reach 100kW by using PATs [15].
71 In Germany, eight parallel centrifugal pumps with total installation capacity 300kW were used
72 in WSS for pressure control and electricity generation in pipelines, the generated electricity
73 was fed into local grid [16]. Fontana et al investigated the water distribution system in one
74 district of Naples to assess the energy potential of hydropower generation using the PATs [17].
75 In this research, a numerical model was firstly developed using genetic algorithms to determine
76 the optimal locations of PRVs in the water distribution system for better water pressure
77 management. After that, all or some of the PRVs were replaced by the PATs for power
78 generation, results showed that although do not have a fine regulation of water head, the PATs
79 did achieve a good balance between hydropower recovery and water head reduction.
80
81 In the research of Fecarotta et al, a multistage pump for WSS application was studied by
82 numerical and experimental methods [18]. The results showed that the performance of a PAT
83 was very different from that of a PRV. For a PRV, the water head reduction could be adjusted
84 by changing its opening degree. As to a PAT, due to the fixed geometrical characteristics and
85 inner flow space, its performance on head reduction was easily effected by the variation of flow
86 rates. In order to solve this problem, A. Carravetta et al designed a PAT installation scheme, in
87 which a bypass and two regulating valves were adopted to cooperate with the PAT to adjust
88 the water head and flow rate through the PAT [19], so the working condition of PAT could
89 stay around its best efficiency point and the water pressure reduction was also ensured.
90 Relevant cost-benefit analysis of this PAT installation scheme was tested on a water pipe
 ACCEPTED MANUSCRIPT

91 network, results showed that the proposed installation scheme was economically acceptable
92 and had little impact on the reliability of pipe networks [20].
93
94 Current researches about the PATs applied in WSS mainly focused on the main pipelines,
95 which have relatively stable and sufficient flow rate, of district water supply networks.
96 However, researches about the PATs used in the WSS of one individual high rise building for
97 hydropower generation and head reduction have been less studied. In the present paper, a
98 systematic study on the PATs used in the WSS of high rise building combining PAT selection
99 method and installation strategy was conducted. First, the flow characteristics in the WSS of
100 one typical high rise building in Hong Kong were analyzed to determine the PAT design
101 conditions. Thereafter, a PAT was selected by theoretical and numerical methods based on the
102 determined design conditions. Then the performance of selected PAT was tested in hydraulic
103 test rig. Finally, a PAT installation and control strategy was designed and tested to ensure the
104 balance between water head reduction and normal water supply.
105
106 2 Materials and methods
107 The PAT selection process in this paper is shown in Fig 1, which can be divided into three
108 parts. Firstly, a PAT was selected by applying empirical equations based on its expected
109 working conditions. After that, the 3-D model of the selected PAT was built using CAD
110 software and CFD simulations were carried out to determine whether the PAT is workable
111 under its design conditions. If the simulated performance of the selected PAT model could not
112 meet the selection aims, the geometry of 3-D model would be modified. Once the simulated
113 performance of the PAT model was close to the selection aim, the test stage was launched. If
114 the experimental performance of the selected PAT model could not meet the selection aims,
115 the geometry modification and simulation would be repeated until the PAT performance
116 satisfies the requirements.
117
118 INSERT Fig.1
119
120 2.1 Determination of design conditions
 ACCEPTED MANUSCRIPT

121 Flow rate and head are two key design parameters for the selection of PATs. Compared to
122 rivers, sewage water or other sources for micro hydropower, fresh water in the WSS of one
123 individual high rise buildings is more fluctuant due to the variation of user demand over time
124 [9, 21]. This feature possesses a challenge to determine a design flow rate for the PAT selection.
125 Fig. 2 presents the flow rate characteristics in the WSS of one typical high rise building in
126 Hong Kong. It can be seen that the minimum flows occur in the nighttime while water usage
127 mainly appear in the daytime (from 10:00 a.m. to 10:00 p.m.), but the flow rates vary greatly
128 from hour to hour. In order to determine the design flow rate of the PAT, the low flow rates at
129 night are abandoned as the power potential is relatively low. Although the flow rates during
130 daytime also fluctuate in a wide range from 5m3/h to 18m3/h, most of these flow rates are
131 10m3/h or above, for the balance between power potential and the daily working time of the
132 selected PAT, 10m3/h is determined as the design flow rate for the PAT selection. In practice,
133 the water head has to be reduced by 30-35m water head at the entrance of the WSS of high rise
134 buildings, so in this paper, the 30m water head is determined as the design head for the PAT
135 selection. Considering the above analysis, high water head and low flow rate are the main
136 characteristics of the water flow in WSS of high rise buildings
137
138 INSERT Fig.2
139
140 Like other hydraulic turbines, the PAT converts hydro power into mechanical shaft power, then
141 further transfers the mechanical power into electricity with a generator. Hence the theoretically
142 available power can be calculated based on Eq. (1). However, for the calculation of mechanical
143 shaft power and actual power output, mechanical loss and conversion efficiency of the PAT
144 and generator have to be considered as energy loss is inevitable in the process of energy
145 conversion, as shown in Eq. (2) and Eq. (3).

Pth   gHQ 3600 (1)

Pme  nT 9.55 (2)

Pac  t me g  gHQ 3600 (3)

 ACCEPTED MANUSCRIPT

Pme
t  (4)
Pth

146 Where Pth is the theoretically available power (W),  is the density of water (kg/m3), g is the

147 acceleration of gravity (m/s2), H is the available water head (m), Q is the flow rate in pipeline

148 (m3/h), Pme is the mechanical shaft power (W), n is the rotation speed (rpm), T is the shaft

149 torque (N·m), Pac is the actual power output (W), t is the efficiency of the PAT, me is the

150 overall mechanical efficiency, and  g is the conversion efficiency of generator. The efficiency

151 of the PAT is regarded as the ratio of the mechanical shaft power to the theoretical input power
152 by water, as shown in Eq. (4).
153
154 2.2 Criteria for the PAT preliminary selection
155 As the range of flow rates over which a PAT can operate is much narrower than that in a
156 conventional turbine, the centrifugal pump to be used as turbine should be carefully selected
157 before installation to obtain the best possible efficiency [17,22]. However, the pump
158 manufacturers usually could not provide the performance curves of their pumps in turbine
159 mode [23]. Many researchers have put their study emphasis on establishing a correlation
160 between the performance characteristics in pump mode and that in turbine mode. Stepanoff
161 [24], Childs [25], Hancock [26], Sharma [27], Luneburg and Nelson [28], Schmiedl [29] and
162 Alatorre-Frenk [30] developed the correlation based on efficiency in pump mode; while
163 Kittredge [31], Diederich [32], Grover [33], Lewinski-Keslitz [34] and Buse [35] derived the
164 relations on basis of the specific speed of the pump. Sun-Sheng Yang [23] developed a
165 theoretical method for PAT performance prediction and verified its accuracy in numerical and
166 experimental ways. The comparison between the derived method and two other methods
167 proposed by Stepanoff [24] and Sharma [27] respectively showed that the proposed method
168 was more accurate in PAT performance prediction than other two methods. In this paper, the
169 method proposed by Sun-Sheng Yang (as shown in Eq. (5) and Eq. (6)) was adopted for PAT
170 preliminary selection.
 ACCEPTED MANUSCRIPT

Qt 1.2
 0.55 (5)
Qp  p

H t 1.2
 (6)
H p  1.1
p

171 Where Qt is the flow rate in turbine mode (m3/h), Q p is the flow rate in pump mode (m3/h),

172  p is the efficiency in pump mode, H t is the water head in turbine mode (m), H p is the water

173 head in pump mode (m).

174
175 2.3 Computational fluid dynamics simulation
176 No matter which method was chosen for PAT selection, once a pump is selected to operate as
177 a turbine, its performance in turbine mode should be verified [16]. In recent decades,
178 computational fluid dynamics (CFD) has been a promising numerical method for performance
179 prediction and inner flow analysis in the domain of rotating machinery design and optimization
180 [36,37]. In this paper, the CFD simulation was used to examine the performance and inner flow
181 characteristics of the selected PAT.
182
183 2.3.1 Grid generation
184 After theoretical selection, 3D model of the selected PAT, which is divided into four parts--
185 inlet reducer, impeller, volute and outlet reducer, was built in SolidWorks, a solid modeling
186 computer-aided design software. The whole computational domain was separated into two
187 main domains: stationary domain and rotary domain. Inlet reducer, volute and outlet reducer
188 belong to the stationary domain while impeller is the rotary domain. For simulation, the sliding-
189 mesh interface method was adopted to allow for the mesh motion between static and rotating
190 parts [38,39].
191
192 INSERT Fig. 3
193
194 After that, the model was imported into ICEM-CFD to generate unstructured tetrahedral grids.
195 In the meshing process, the “tetra mesh” was employed to generate grids in the domains far
 ACCEPTED MANUSCRIPT

196 from boundaries, while the “prism mesh” was used for grid generation in the domains near
197 boundaries, i.e. the volute wall and blades. Such strategy could achieve a good balance between
198 calculation time and accuracy. To minimize the numerical uncertainty in the solution, a grid
199 independence test was conducted. Fig.3 indicates the sensitivity of mesh on head reduction, it
200 can be seen in Fig.3 that variation of head reduction is relatively small when the number of
201 grid is more than 1.54 million. The final mesh of the model is shown in Fig.4, and the mesh
202 number of inlet reducer, impeller, volute, outlet reducer and total number are 0.39, 0.54, 0.31,
203 0.3 and 1.54 million respectively.
204
205 INSERT Fig.4
206
207 2.3.2 Numerical model
208 The transport equations that describe conversion of mass and momentum for incompressible
209 flows are continuity and Navier-Stokes equations, which can be written as

ui
0 (7)
xi

ui 1 p  2ui 
ui  v  (uiu j )  Fi (8)
x j  xi xi x j x j

210 Where u is the velocity, p is the pressure, v is the kinematic viscosity,  is the density and

211 uiu j is the turbulent stress.

212
213 To solve these equations, three main methods are used: the direct numerical simulation (DNS)
214 method, the large eddy simulations (LES) method and Reynolds averaged Navier–Stokes
215 (RANS) method. The DNS method is a powerful tool to solve the continuity and Navier-Stokes
216 equations without introducing any model, but it requires much more computing resources and
217 is only suitable for flows with relatively low Reynolds number and for simple flow passages
218 until now [40,41,42]. In the LES method, the large eddies are solved while the small eddies are
219 removed by a spatial filtering procedure, it has a good performance for the calculation of high
220 Reynolds number flows, but special near-wall treatment has to be introduced for proper
 ACCEPTED MANUSCRIPT

221 resolution in the near-wall region, which possesses high computational cost in terms of
222 computing power and time [40,43]. The RANS approach is the most common and mature
223 simulation method in the field of turbulent calculation [44, 45]. This method provides adequate
224 accuracy for most practical application and needs less computing resources and time [46]. The
225 good performance of the RANS method for PAT simulation has been proved by many studies
226 [16,18,23,36,47], so in this paper, the RANS method is adopted.
227
228 In the RANS method, a turbulence model is needed to close the equations [40]. Unfortunately,
229 there is no turbulence model that could be used in all the problems and the turbulence model
230 should be chosen based on the characteristics of flow, available computing resources and
231 demand of calculation accuracy [12]. In references, standard k   [23,48,49,50]and SST
232 k  w [51,52,53,54] models are commonly used for the performance prediction of PATs or
233 Francis turbines, but studies about the comparison between these two models are still seldom.
234 In this paper, both of the standard k   and SST k  w models are adopted for simulation to
235 determine the more accurate model for performance examination of the PATs.
236
237 2.3.3 Boundary conditions
238 In the WSS of high rise buildings, the flow rate in the pipes and the water pressure at the outlet
239 of the PAT have to be ensured for normal water supply, so in the simulation, velocity was set
240 as the inlet boundary condition and static pressure was set as the outlet boundary conditions.
241
242 INSERT Fig. 5
243
244 As is shown in Fig .5 is the schematic diagram of flow path inside the PAT. It can be seen that
245 leakage exists in the space between impeller hub/shroud and casing, which means the actual
246 flow rate through the impeller is less than that under design conditions [55]. However, in the
247 study, leakage could not be simulated because the numerical model did not include the flow
248 field of leakage. To obtain a more accurate simulation results, the leakage is estimated as a
249 function of the geometry of the flow field and the head [56]:
 ACCEPTED MANUSCRIPT

QL  ( Af  Ab ) H m0.5 (9)

2cl H cl 0.5 2
 2g ( )( ) D0
Ds H m
A (10)
D D
2( 0 ) 2 (1  ( h ) 2 )
Ds D0

Dsf  1.29 D0 (11)

Dsb  1.32 D0 (12)

250 Where QL is the leakage flow (m3/h), H m is the design head (m), Af and Ab are the areas of

251 fore and back leakage passage respectively (m2),  is the leakage flow coefficient which is

252 assumed 0.4, cl is the size of the seal clearance which is set to0.00015, H cl H m is the

253 coefficient of head loss through the seal, which is assumed 0.75, Ds represents the annular gap

254 diameter of the seal (m), D0 is the eye diameter of impeller (m), Dh is the hub diameter (m).

255
256 INSERT Fig.6
257

258 The head reduction caused by leakage, H L , is estimated by the empirical formula developed

259 by Stepanoff [57]:

3 u12  u22
HL  (13)
4 2g

260 Where u1 and u2 are the inlet and outlet tangential velocities of the impeller (as is shown in

261 Fig.6) and they can be calculated as a function of the rotation speed and diameter:
 D1 N
u1  (14)
60
 D2 N (15)
u2 
60
262 Where D1 and D2 are the diameters of inlet and outlet of the impeller respectively (m), N is

263 the rotation speed of the PAT (rpm).

264 To obtain a more accurate simulation result, the leakage is reduced from the inlet discharge of
 ACCEPTED MANUSCRIPT

265 simulation and the head loss caused by leakage is also added to the simulated head reduction.
266 For all the wall boundaries, the roughness was set to 100μm.
267
268 2.3.4 Solution parameters
269 All the simulations were conducted in Fluent using a second-order-accurate finite-volume
270 discretization scheme and the maximum residual is set to 10-5. The torque output of impeller
271 was monitored for power output calculation, while the pressure at inlet and outlet were
272 collected to evaluate the water head reduction caused by the PAT. By changing the inlet flow
273 rate, the performance of PAT under different working conditions were obtained.
274
275 2.4 Experimental setup
276 After theoretical selection and numerical verification, a prototype of PAT was purchased and
277 tested to understand its real performance. A complete hydraulic test rig for PAT tests was built
278 in the laboratory of Renewable Energy Research Group at the Hong Kong Polytechnic
279 University. Fig.7 shows the schematic diagram of this test rig, which mainly consists of a water
280 tank, a power pump, control valves, meters and sensors. The entire system was driven by a
281 frequency converting controlled centrifugal pump with rated flow rate 81m3/h and water head
282 82m. By changing the frequency of power pump and the opening degree of ball valve located
283 at the end of the pipeline, the required flow rate and water head could be achieved in the test
284 rig. Besides, two pressure sensors were used to examine the pressure drop between the
285 upstream and downstream of the water turbine and one electromagnetic flowmeter was adopted
286 for the flow rate measurement in the pipeline, all the sensors and flowmeter are located far
287 from the PAT for accurate measurement. A 24V three-phase permanent magnet alternating
288 generator with low starting torque was chosen to translate shaft power of the PAT into electric
289 energy, after rectification in the controller, the electricity was stored in two lead acid batteries
290 and the power and voltage output could be obtained from the controller. The diagrams of
291 established test rig and the controller and monitoring system are shown in Fig.8 and Fig. 9
292 respectively. The pipeline diameter in the test rig is 100mm, which is the same value of pipeline
293 diameter in the WSS of aimed high rise building.
294
 ACCEPTED MANUSCRIPT

295 INSERT Fig.7

296 INSERT Fig.8
297 INSERT Fig.9
298
299 3. Results and discussion
300 3.1 Geometric parameters of the PAT
301 After theoretical calculation, a single stage vertical shaft centrifugal pump was selected to be
302 used as a hydraulic turbine for investigation. The main geometric parameters of the selected
303 pump are shown in Table 1.
304
305 INSERT Table1
306
307 3.2 Results of simulations and lab tests
308 After a series of simulations and lab tests, the numerical and experimental performance curves
309 of the selected PAT were achieved. To compare the numerical and experimental results,
310 mechanical losses in mechanical seal, mechanical transmission components and generator,
311 which are provided by the pump and generator manufacturer, are reduced from the simulated
312 output power and efficiency. Fig.10 shows the numerical and experimental power output and
313 efficiency, while Fig.11 shows experimental and simulated water head reduction. Both the
314 numerical and experimental performance curves indicate the same trend of power output at
315 different flow rates, and the simulated and actual best efficiency points (BEPs) are both located
316 in the flow rate range from 9.5m3/h to 10.5 m3/h. According to the laboratory test results, the
317 maximum power output of the selected PAT under design conditions could reach about 110W.
318 Besides, as the investigated PAT is to be used in the WSS of high rise buildings for water
319 pressure management, its performance on water head reduction is of vital importance. Fig.11
320 shows that both the simulated and measured water head reduction at 10 m3/h are in the range
321 of the given working condition (30-35m water head). The performance curve of the PAT versus
322 rotation speed is shown in Fig.12, the rotation speed at BEP is about 1780 rpm.
323
324 INSERT Fig.10
 ACCEPTED MANUSCRIPT

325 INSERT Fig.11

326 INSERT Fig.12
327
328 It can be seen from Fig.10 and Fig.11 that the simulated efficiency and head reduction of SST
329 k   model are more close to experimental results than that of k   model. Table 2 shows
330 the difference between simulation results of different models and experimental results. As can
331 be seen in this table, the SST k   model is more accurate in performance prediction of the
332 PAT. The main reason is that the SST k   model combines the standard k   model and
333 standard k   model, it takes the effects of turbulence shear stress into consideration in the
334 definition of the turbulence viscosity and could capture the micro flow in the viscous layer
335 [51,53].
336
337 INSERT Table2
338
339 However, there are some deviations between CFD results and experimental results, the
340 simulated PAT efficiency of SST k   model is around 19% while the measured efficiency is
341 only about 11.5%. Table 3 compares the values of simulated and experimental head reduction
342 under different flow rates. It can be seen that experimental water head reduction is higher than
343 that of numerical results, and the error decreases with the increase of flow rate. Similar
344 deviations could also be found in some references about numerical studies on performance
345 prediction of the PAT [58,59]. There are two main reasons for this quantitative difference. On
346 the one hand, the simplicity of CFD model may cause miscalculation on the hydraulic loss, but
347 this miscalculation is difficult to be measured and ruled out. Besides, water in the space
348 between impeller hub/shroud and casing results in severe disc friction, which reduces the
349 conversion efficiency from hydro power to mechanical shaft power significantly. The effect of
350 disc friction loss is likely to be greater in low specific speed turbines [57], however, it could
351 not be simulated in this study because the numerical model did not include the flow field
352 between impeller hub/shroud and casing.
353
354 INSERT Table3
 ACCEPTED MANUSCRIPT

355
356 Regardless of the difference between numerical and experimental results, the trends and BEP
357 locations of the simulated results are consistent with that of the testing results from the
358 qualitative point of view. So the numerical results could be used for performance prediction of
359 the PAT.
360
361 3.3 Head reduction distribution
362 The water head reduction distribution within the four parts of the PAT (inlet reducer, volute,
363 impeller and outlet reducer) is presented in Fig.13. As indicated in Fig.13, the water head
364 reductions in the impeller and volute account for a great proportion of the total water head
365 reduction, and the values increase with the growth of flow rates. Besides, the inlet reducer also
366 makes up a small part of water head reduction.
367
368 INSERT Fig.13
369
370 However, it can also be observed from the experimental results that at high flow rates, the water
371 head reduction is far beyond the design head reduction. For instance, the water head reduction
372 at 11m3/h is 45m while that at 12 m3/h is 50m. Excessive water head reduction may have
373 inverse impact on the normal water supply, so in this paper, a PAT installation and control
374 strategy is designed and tested to solve this problem.
375
376 4 The PAT installation and control strategy
377 4.1 The PAT installation and control strategy introduction
378 As discussed in part 2.1, the flow rate in the WSS of high rise buildings is changing all the
379 time, and the maximum flow rate could be up to 18 m3/h while the minimum flow rate could
380 be only 5 m3/h. As suggested in the research of Fecarotta et al [14], the water head reduction
381 through a PAT is easily effected by the flow rate variation due to the fixed geometrical
382 characteristics and inner flow space. In order to achieve a good balance between power output
383 and water reduction without affecting the normal water supply, a PAT installation and control
384 strategy was designed.
 ACCEPTED MANUSCRIPT

385
386 INSERT Fig.14
387
388 Fig.14 is the PAT installation strategy with a bypass to deliver excess flow rate. An
389 electromagnetic control valve, which is controlled by the controller, was installed in parallel to
390 the PAT in the bypass. A sensor, which was connected to the controller, was installed on the
391 PAT to detect the rotation speed.
392
393 INSERT Fig.15
394
395 Rotation speed of PAT is the direct result of flow rate, so the rotation speed was selected as the
396 control variable in the PAT control strategy. The working principle of the PAT control strategy
397 is shown in the flow chart of Fig.15. As can be seen, the control strategy is a closed loop control.
398 To begin with, an acceptable rotation speed range, for example 1200-1450rpm, was used in the
399 controller, and the real time value of the PAT rotation speed was obtained then sent to the
400 controller by the sensor. If the rotation speed was in the right range, the electromagnetic control
401 valve would have no action. If the rotation speed was beyond the set range, the controller would
402 control the electromagnetic control valve to open with a bigger opening degree, so the flow
403 rate through the PAT would be reduced and the rotation speed would decrease. If the rotation
404 speed was below the range, the controller would control the electromagnetic control valve to
405 close with a smaller opening degree, so the flow rate through the bypass would be reduced and
406 the PAT rotation speed would increase. The controller compared the real rotation speed with
407 the typed range every 10 seconds and the adjustment process would repeat until the rotation
408 speed keeps stable in the given range.
409
410 4.2 The PAT installation and control strategy test
411
412 INSERT Fig.16
413
414 In order to study the real performance of the PAT installation strategy, the hydraulic test rig
 ACCEPTED MANUSCRIPT

415 was modified. Fig.16 shows the test rig for the PAT installation strategy, it can be seen that a
416 bypass with a smaller diameter is installed in parallel with the main pipeline. A motorized
417 valve, which is mainly composed of a motor, a shaft and a ball valve, is installed to control the
418 flow rate through the bypass by changing its opening degree. The working voltage of the motor
419 is 24V, so the chargeable lead batteries could be used to provide power for the motorized valve.
420 It takes 12s to fully open or close the motorized valve, by setting the operating time of one step,
421 the opened degree of each step could be adjusted. In the test study, the operating time of one
422 step was set to be 0.5s, which meant that 24 steps were needed to fully open or close the
423 motorized valve and the opened or closed degree of each step is 3.75°. Besides, the given
424 rotation speed range was from 1200rpm to 1450rpm.
425 4.3 Test results of the PAT installation strategy
426
427 INSERT Fig.17
428 INSERT Fig.18
429
430 In the experimental process, the flow rate in the pipeline of hydraulic test rig was controlled to
431 increase gradually and the power output and water reduction of the PAT was recorded. Fig.17
432 shows the output power and rotating speed changing trend with operating time. At the
433 beginning of test, with the increase of flow rate, both the rotation speed and power output
434 experienced an increase. When the rotation speed exceeded the given range, the motorized
435 valve began to open, and after a slight fluctuation, the power output stayed stably at about 80W
436 and the rotation speed remained in the suggested range. Fig. 18 shows the value of water head
437 reduction at different flow rates. According to the previous test results, at or above 11m3/h, the
438 water head reduction in pipeline was more than 35m. After installation of motorized valve and
439 bypass, the water head reduction stayed around 33m at or above 11m3/h, which suited the
440 proposed working condition very well. For security reasons, a pressure reduction valve was
441 suggested to be installed after the PAT and bypass, so the water head of fresh water distributed
442 to users could keep safe and stable.
443 5. Conclusions
444 In this paper, a systematic investigation about the PAT used for hydro power generation in the
 ACCEPTED MANUSCRIPT

445 WSS of high rising building was carried out. On the basis of data collected from the WSS of a
446 typical high rise building in Hong Kong, a PAT was selected by empirical equations. After
447 that, the performance of selected PAT was studied by numerical and experimental methods,
448 and both the simulated and experimental results show that the selected PAT is feasible in hydro
449 power generation as well as water head reduction. According to the test results, the maximum
450 power output of the PAT under 10 m3/h is 110W with 34m water head reduced. Although the
451 deviations between CFD results and experimental results exist, the simulation outcomes could
452 be used for performance prediction of the PAT. The water head reduction distribution analysis
453 indicates that impeller and volute make up most of the total water head reduction, in addition,
454 the role of inlet reducer in water head reduction could not be ignored.
455
456 Lastly, a PAT installation and control strategy aiming to avoid excessive water head reduction
457 at higher flow rates was designed and tested in the laboratory test rig. The test result shows that
458 the proposed PAT installation strategy has a really good performance. The power output keeps
459 stable around 100W while the water head reduction remains about 33m at 10m3/h or above.
460
461 Acknowledgments
462 The authors would like to thank Sino Green Limited in Hong Kong for funding this project and
463 close collaborations for its success.
464
 ACCEPTED MANUSCRIPT

Reference
[1]. Chen X, Yang H, Zhang W. A comprehensive sensitivity study of major passive design
parameters for the public rental housing development in Hong Kong. Energy, 2015, 93:
1804-1818.
[2]. Cheng C L, Yen C J, Wong L T, et al. An evaluation tool of infection risk analysis for
drainage systems in high-rise residential buildings. Building Services Engineering
Research and Technology, 2008, 29(3): 233-248.
[3]. Park J H, Chung M H, Park J C. Development of a small wind power system with an
integrated exhaust air duct in high-rise residential buildings. Energy and Buildings, 2016,
122: 202-210.
[4]. Lotfabadi P. High-rise buildings and environmental factors. Renewable and Sustainable
Energy Reviews, 2014, 38: 285-295.
[5]. Cheung C T, Mui K W, Wong L T. Energy efficiency of elevated water supply tanks for
high-rise buildings. Applied Energy, 2013, 103: 685-691.
[6]. Venkatesh G, Chan A, Brattebø H. Understanding the water-energy-carbon nexus in urban
water utilities: Comparison of four city case studies and the relevant influencing factors.
Energy, 2014, 75: 153-166.
[7]. Ramos H M, Kenov K N, Vieira F. Environmentally friendly hybrid solutions to improve
the energy and hydraulic efficiency in water supply systems. Energy for Sustainable
Development, 2011, 15(4): 436-442.
[8]. Coelho B, Andrade-Campos A. Efficiency achievement in water supply systems—A
review. Renewable and Sustainable Energy Reviews, 2014, 30: 59-84.
[9]. Kaunda C S, Kimambo C Z, Nielsen T K. A technical discussion on microhydropower
technology and its turbines. Renewable and Sustainable Energy Reviews, 2014, 35: 445-
459.
[10]. Elbatran, A. H., et al. "Operation, performance and economic analysis of low head
micro-hydropower turbines for rural and remote areas: a review." Renewable and
Sustainable Energy Reviews 43 (2015): 40-50.
[11]. Paish O. Small hydro power: technology and current status. Renewable & Sustainable
Energy Reviews, 2002, 6(6):537–556.
[12]. Chen J, Yang H X, Liu C P, et al. A novel vertical axis water turbine for power
generation from water pipelines. Energy, 2013, 54(2):184-193.
[13]. Caxaria G A, Sousa D D M E, Ramos H M. Small Scale Hydropower: Generator
Analysis and Optimization for Water Supply Systems// World Renewable Energy
Congress – Sweden, 8–13 May, 2011, Linköping, Sweden. 2011:1386-1393.
[14]. Ramos H M, Mello M, De P K. Clean power in water supply systems as a sustainable
solution: from planning to practical implementation.. Water Science & Technology Water
Supply, 2010, 79(1):24-26.
[15]. Pérez García J, Cortés Marco A, Nevado Santos S. Use of centrifugal pumps operating
as turbines for energy recovery in water distribution networks. Two case study//Advanced
Materials Research. Trans Tech Publications, 2010, 107: 87-92.
[16]. Jain S V, Patel R N. Investigations on pump running in turbine mode: a review of the
 ACCEPTED MANUSCRIPT

state-of-the-art. Renewable and Sustainable Energy Reviews, 2014, 30: 841-868.

[17]. Fontana N, Giugni M, Portolano D. Losses reduction and energy production in water-
distribution networks. Journal of Water Resources Planning and Management, 2011,
138(3): 237-244.
[18]. Fecarotta O, Carravetta A, Ramos H. CFD and comparisons for a pump as turbine:
mesh reliability and performance concerns. Int J Energy Environ, 2011, 2(1): 39-48.
[19]. Carravetta A, del Giudice G, Fecarotta O, et al. PAT design strategy for energy
recovery in water distribution networks by electrical regulation. Energies, 2013, 6(1): 411-
424.
[20]. Carravetta A, Fecarotta O, Sinagra M, et al. Cost-benefit analysis for hydropower
production in water distribution networks by a pump as turbine. Journal of Water
Resources Planning and Management, 2013, 140(6): 04014002.
[21]. McNabola A, Coughlan P, Williams A P. Energy recovery in the water industry: an
assessment of the potential of micro‐hydropower. Water and Environment Journal, 2014,
28(2): 294-304.
[22]. Teuteberg B H. Design of a pump-as-turbine microhydro system for an abalone farm.
Stellenbosch: University of Stellenbosch, 2010.
[23]. Yang S S, Derakhshan S, Kong F Y. Theoretical, numerical and experimental
prediction of pump as turbine performance. Renewable Energy, 2012, 48: 507-513.
[24]. Stepanoff A J. Centrifugal and axial flow pumps. 1948.
[25]. Childs S. Convert pumps to turbines and recover HP. Hydrocarbon Processing and
Petroleum Refiner, 1962, 41(10): 173-174.
[26]. Hanckock J. Centrifugal pump or water turbine. Pipe Line News, 1963: 25-27.
[27]. Sharma K. Small hydroelectric projects-use of centrifugal pumps as turbines.
Kirloskar Electric Co., Bangalore, India, 1985.
[28]. Lueneburg R, Nelson R M. Hydraulic power recovery turbines. Centrifugal Pumps–
Design and Application, second ed., Gulf Publishing Company, 1992: 246-282.
[29]. Schmiedl E. Serien-Kreiselpumpen im Turbinenbetrieb. Pumpentagung: Karlsruhe,
Germany, 1988.
[30]. Alatorre-Frenk C. Cost minimisation in micro-hydro systems using pumps-as-
turbines. University of Warwick, 1994.
[31]. Kittredge C P. Centrifugal pumps used as hydraulic turbines. Journal of Engineering
for Power, 1961, 83(1): 74-78.
[32]. Diederich H. Verwendung von kreiselpumpen als turbinen. KSB Techn Ber, 1966, 12.
[33]. Grover K. Conversion of pumps to turbines. GSA Inter corp., Katonah, New York,
1980.
[34]. Lewinsky-Kesslitz H P. Pumpen als Turbinen fur Kleinkraftwerke. Wasserwirtschaft,
1987, 77(10): 531-537.
[35]. Buse F. Using centrifugal pumps as hydraulic turbines. Chemical Engineer, 1981,
88(2): 113-117.
[36]. Keck H, Sick M. Thirty years of numerical flow simulation in hydraulic
turbomachines. Acta Mechanica, 2008, 201(1-4): 211-229.
 ACCEPTED MANUSCRIPT

[37]. Croba D, Kueny J L. Numerical calculation of 2D, unsteady flow in centrifugal

pumps: impeller and volute interaction. International Journal for Numerical Methods in
Fluids, 1996, 22(6): 467-481.
[38]. Zhang B, Liang C. A simple, efficient, and high-order accurate curved sliding-mesh
interface approach to spectral difference method on coupled rotating and stationary
domains. Journal of Computational Physics, 2015, 295(C):147-160.
[39]. Ferrer E, Willden R H J. A high order Discontinuous Galerkin - Fourier
incompressible 3D Navier-Stokes solver with rotating sliding meshes. Journal of
Computational Physics, 2012, 231(21):7037–7056.
[40]. Rodi W. DNS and LES of some engineering flows. Fluid Dynamics Research, 2006,
38(2):145-173.
[41]. Wang W Q, Zhang L X, Yan Y, et al. LARGE-EDDY SIMULATION OF
TURBULENT FLOW CONSIDERING INFLOW WAKES IN A FRANCIS TURBINE
BLADE PASSAGE. Journal of Hydrodynamics, Ser. B, 2007, 19(2):201-209.
[42]. Nieuwstadt F T M, Eggels J G M, Janssen R J A, et al. Direct and large-eddy
simulations of turbulence in fluids. Future Generation Computer Systems, 1994, 10(2–
3):189-205.
[43]. Chen X, Xia H. A hybrid LES-RANS study on square cylinder unsteady heat transfer.
International Journal of Heat & Mass Transfer, 2016.
[44]. Gourdain N. Prediction of the unsteady turbulent flow in an axial compressor stage.
Part 1: Comparison of unsteady RANS and LES with experiments. Computers & Fluids,
2015, 106:119-129.
[45]. Coroneo M, Montante G, Paglianti A, et al. CFD prediction of fluid flow and mixing
in stirred tanks: Numerical issues about the RANS simulations. Computers & Chemical
Engineering, 2011, 35(10):1959-1968.
[46]. Efthimiou G C, Andronopoulos S, Bartzis J G, et al. CFD-RANS prediction of
individual exposure from continuous release of hazardous airborne materials in complex
urban environments. 2016:1-23.
[47]. Morros C S, Oro J M F, Díaz K M A. Numerical modelling and flow analysis of a
centrifugal pump running as a turbine: Unsteady flow structures and its effects on the
global performance. International Journal for Numerical Methods in Fluids, 2011,
65(5):542–562.
[48]. González J, Oro J M F, Argüelles-Díaz K M, et al. Flow analysis for a double suction
centrifugal machine in the pump and turbine operation modes. International Journal for
Numerical Methods in Fluids, 2009, 61(2):220-236.
[49]. Saeed R A, Galybin A N, Popov V. Modelling of flow-induced stresses in a Francis
turbine runner. Advances in Engineering Software, 2010, 41(12):1245-1255.
[50]. Morros C S, Oro J M F, Díaz K M A. Numerical modelling and flow analysis of a
centrifugal pump running as a turbine: Unsteady flow structures and its effects on the
global performance. International Journal for Numerical Methods in Fluids, 2011,
65(5):542–562.
[51]. Su X, Huang S, Zhang X, et al. Numerical research on unsteady flow rate
 ACCEPTED MANUSCRIPT

characteristics of pump as turbine. Renewable Energy, 2016, 94:488-495.

[52]. Zhang H, Zhang L. Numerical simulation of cavitating turbulent flow in a high head
Francis turbine at part load operation with OpenFOAM. Procedia Engineering, 2012,
31(1):156–165.
[53]. Choi H J, Zullah M A, Roh H W, et al. CFD validation of performance improvement
of a 500kW Francis turbine. Renewable Energy, 2013, 54(6):111-123.
[54]. Li D, Wang H, Xiang G, et al. Unsteady simulation and analysis for hump
characteristics of a pump turbine model. Renewable Energy, 2015, 77:32-42.
[55]. Liu X, Luo Y, Karney B W, et al. A selected literature review of efficiency
improvements in hydraulic turbines. Renewable & Sustainable Energy Reviews, 2015,
51:18-28.
[56]. Neumann B. The interaction between geometry and performance of a centrifugal
pump[M]. Mechanical Engineering Publications, 1991.
[57]. Stepanoff A J. Centrifugal and axial flow pumps: Theory, design, and application//
Hydraulic Fill Structures. ASCE, 2011:32-51.
[58]. Derakhshan S, Nourbakhsh A. Theoretical, numerical and experimental investigation
of centrifugal pumps in reverse operation. Experimental Thermal and Fluid Science, 2008,
32(8): 1620-1627.
[59]. Rodrigues A, Singh P, Williams A, et al. Hydraulic analysis of a pump as a turbine
with CFD and experimental data//IMechE seminar computational fluid dynamics for fluid
machinery. 2003.
 ACCEPTED MANUSCRIPT

FIGURES

Begin

Input design flow rate

and water head

Preliminary selection of PAT by

empirical equations

Modify geometry and perform

CFD simulation

Best Efficiency NO
Acceptable?

YES
Test

NO Performance
acceptable?

YES
End

Fig.1 Flow chart of PAT selection process

Fig.2 The flow rate characteristics at different time

ACCEPTED MANUSCRIPT

Fig.3 Mesh independence test

Fig.4 Mesh of the simulated model

 ACCEPTED MANUSCRIPT

Dsf Dh

Fig.5 Flow path inside the PAT

w1

u1 c1

u2
w2

c2

Fig.6 Impeller inlet and outlet velocity triangle

 ACCEPTED MANUSCRIPT

Pressure Sensor II Pressure Sensor I

PAT
Electromagnetic Computer
Flow Meter Generator

Controller
Ball Valve

Pressure
Meter Battery
Water Tank

Pump

Fig.7 Schematic diagram of hydraulic test rig

Fig.8 Hydraulic test rig for PAT tests

 ACCEPTED MANUSCRIPT

Fig.9 The controller and monitoring system

Fig.10 Numerical and experimental power output and efficiency

 ACCEPTED MANUSCRIPT

Fig.11 Experimental and simulated water head reduction

Fig.12 Performance curve of the PAT versus rotation speed

 ACCEPTED MANUSCRIPT

Fig.13 Water head reduction distribution inside the studied PAT

Pipeline PAT Rotation Speed Sensor

Controller

Generator Motorized Control Valve

Fig.14 PAT installation strategy

 ACCEPTED MANUSCRIPT

Fig.15 Working principle of PAT installation strategy

Fig.16 Test rig for the PAT installation strategy

 ACCEPTED MANUSCRIPT

Fig.17 Experimental power output of the PAT installation strategy

Fig.18 Experimental water head reduction of the PAT installation strategy

 ACCEPTED MANUSCRIPT

TABLES

Table 1 Geometric parameters of PAT prototype

Parameters Values
Impeller Inlet Diameter (mm) 160
Impeller Outlet Diameter (mm) 40
Impeller Width (mm) 4
Impeller Blade Thickness (mm) 5
Volute Base Circle Diameter (mm) 170
Volute Inlet Diameter (mm) 25
Volute Outlet Diameter (mm) 25
Blades Number 5

Table 2 Deviation between simulation results of different models and experimental

results

Comparison of efficiencies Comparison of head reductions

Lab tests SST k   k  Lab tests SST k   k 

Value 11.5% 18.8% 19.9% Value 35.4m 33m 31.6m

Error 0 7.3% 8.4% Error 0 -6.7% -10.7%

Table 3 Numerical and experimental water head reduction

Flow rate (m3/h) 7 8 9 10 11

Experimental value (m) 14 19.4 25.4 34.3 43.3

Numerical value (m) 12 16.4 22.7 33 40.3

Error (%) 14.3% 15.5% 10.6% 3.8% 6.9%