International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
International Scholarly and Scientific Research & Innovation 2(7) 2008 873 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
ΔU = Un +1 − Un (10)
International Science Index, Mechanical and Mechatronics Engineering Vol:2, No:7, 2008 waset.org/Publication/10550
⎜⎝ 2c 2c ⎠⎟⎟
⎛ u 2
− c 2
c − u ⎞⎟
⎜⎜ ⎟⎟
− ⎜⎜ 2c 2 c ⎟⎟
A =⎜ 2⎟ (11)
⎜⎜ (u + c)(c − u )2 (c − u ) ⎟⎟
⎜⎜⎝ − ⎟
2c 2c ⎠⎟⎟
⎛ ρ(u + c) ⎞⎟ ⎛ ρ(u - c) ⎞⎟
⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟
+ ⎜⎜ 2 ⎟ − ⎜⎜ 2 ⎟⎟
F =⎜ ⎟, F =
⎜⎜ ρ(u + c)2 ⎟⎟⎟ ⎜⎜⎜ ρ(u - c)2 ⎟⎟⎟
Fig. 1 A control volume in a general gas pipeline ⎜⎜⎝ 2 ⎠⎟⎟⎟ ⎝⎜⎜ 2
⎟⎟⎠
⎟
where c is the speed of acoustic wave in the gas flow. When
The governing equations in matrix form are (9) is applied to each grid point, a block tridiagonal system of
∂U ∂F algebraic equations will be obtained. This equations system
+ =R (4) can be solved at each time step using Thomas algorithm,
∂t ∂x
which results in ΔU. Next, U at the advanced time level can
where be calculated using (10).
⎡ ρu ⎤ ⎡ 0 ⎤
⎢ ⎥
F = ⎢⎢ 2 ⎥, R = ⎢ ρu u ⎥ (5) IV. FLOW TRANSFER FUNCTIONS
ρu + P ⎥⎥ ⎢− f − ρg sin α ⎥
⎣⎢ ⎦ ⎢⎣ 2D ⎥⎦
To obtain the flow transfer functions, P0, T0, A0, and ρ0 are
Another form of the relations (1) and (2) versus the gas considered as the reference values and the nonlinear partial
pressure and the mass flow rate can be written as [13] differential equations (6) and (7) are linearized about them.
⎛ ⎞⎟ Moreover, these reference values are also considered to define
∂ ⎜⎜ P ⎟⎟ + 1 ∂m = 0
⎜⎜ (6)
ˆ ) RT ⎟⎟⎟ A ∂x
∂t ⎜⎜⎝(1 + kP the corresponding dimensionless variables expressed as
⎠⎟ x
ξ=
∂P 1 ∂m ∂ ⎛⎜1 + kP
ˆ ⎞ L
=− − RT ⎜⎜ 2 m 2 ⎟⎟⎟ tc
∂x A ∂t ∂x ⎝⎜ A P ⎠⎟ *
t =
(7) L
f m m Δ
ˆ ) RT − g h P
−
2DA2 P
(1 + kP L (1 + kP
ˆ ) RT P * = ρ* =
P
(12)
Po
where m shows the mass flow rate and kˆ is an experimental m * = mc
/ PoAo
parameter which is used to compute the compressibility factor,
u
i.e. u* = o
c
ˆ where uo is the average gas velocity in the pipe and is
Z = 1 + kP (8)
calculated as [13]
International Scholarly and Scientific Research & Innovation 2(7) 2008 874 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
∂ΔP * (s )
= − ⎡⎢ u * fL* + s ⎤⎥ Δm * (s ) +
∂ξ ⎣ ⎦
⎧ fL * ⎫ (18)
⎪
⎪ g Δh ⎪
⎨ u * u * − 2 + 2u *s ⎪ *
⎬ ΔP (s )
⎪
⎪ 2 c ⎪
⎪
⎩ ⎭
After imposing the boundary conditions, the above system
of ODE can be solved. For example, if the gas pressure at the
inlet and the mass flow rate at the pipe outlet are specified as
functions of time, the above system of ODE results in
⎧
⎪
⎪
⎪ΔPout * (s ) = e γ / 2 2b
⎪ ΔPin * (s ) +
⎪
⎪ 2b cosh (b ) − γ sinh (b ) Fig. 2 A simulink model when the pipeline inlet pressure and the
⎪
⎪
⎪ 2α sinh (b ) outlet gas flow rate are known
⎪
⎪ − ΔMout * (s )
⎪
⎪ 2b cosh (b ) − γ sinh (b )
⎪
⎪
⎪
⎨
⎪
⎪
(19) At the present work, a Simulink library for each type of the
⎪
⎪
⎪ 2β sinh (b ) boundary conditions is made in the MATLAB-Simulink
⎪
⎪ΔM in * (s ) = ΔPin * (s ) +
⎪
⎪ 2 b cosh (b ) − γ sinh ( b ) browser that is called as shown in Fig. 3. In this library each
⎪
⎪
⎪ 2b
⎪
⎪ + e−γ / 2 ΔMout * (s ) block has two inputs which are known from the boundary
⎪
⎪
⎩ 2 b cosh ( b ) − γ sinh (b ) conditions, and two outputs as the results of the transient
⎪
where α, β, b and γ are defined in appendix A. After Taylor- simulation. Then, the proposed approach is extended to
expansion of the hyperbolic terms in (19), the simplified simulate a gas network. A typical network which has been
transfer functions are studied by Ke and Ti [10] is considered and simulated with
⎧ the proposed approach. Fig. 4 shows a schematic of this
⎪
⎪ΔP * (s ) = FP ,P ΔPin * (s ) + FM ,P ΔMout * (s )
⎪ out out in out out network and its Simulink model is illustrated in Fig. 5. The
⎨ (20)
⎪Δ *( )
= Δ *( )
+ * accuracy of the obtained results and the computational
⎪
⎪
⎩
M in s FMin ,Pin Pin s FM in ,Mout ΔMout (s )
efficiency of the proposed simulation are discussed in the next
where section.
International Scholarly and Scientific Research & Innovation 2(7) 2008 875 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
International Science Index, Mechanical and Mechatronics Engineering Vol:2, No:7, 2008 waset.org/Publication/10550
VI. RESULTS AND DISCUSSIONS Fig. 7 illustrates the present results of FSM for pressure time
The results of the proposed transient simulation are changes at the pipe outlet, along with those of the others [8],
compared with those of the implicit FSM as an accurate [9], [15] and the experiments. There are some differences
nonlinear finite difference scheme. In order to verify the between the present nonlinear FSM results with those
accuracy of the present implicit FSM, a 72259.5 m long obtained by the others. However, when they are compared
pipeline of 0.2 m diameter was considered as a test case. The with the experiments, it seems that all of the numerical
test case which its experimental results are available, has been methods have the nearly similar differences with experiments.
studied by Taylor et al. [15], Zhou and Adewumi [8], and also The interesting point is the accuracy of the results of the
by Tentis et al. [9]. proposed transfer function model. As it is seen in Fig. 7, the
International Scholarly and Scientific Research & Innovation 2(7) 2008 876 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
2.1E+04
20
15
Flow Rate (S-m3/Hr)
2.0E+04
10
1.9E+04
0 20 40 60 80 100 120
Time (min)
1.7E+04
Implicit FSM
present Transfer Function modeling
7
1.6E+04
0 4 8 12 16 20 24 6.5
Time (Hr)
6
Fig. 6 A 24-hour irregular flow imposed at the pipe outlet
5.5
outlet pressure (bar)
5
30
AML (Tentis 2003) 4.5
28 Taylor et al. (1962)
TVD (Zhou Adewumi 1995) 4
26 Implicit FSM
Experimental data 3.5
24 present Transfer Function modeling
3
22
outlet pressure (bar)
2.5
20 2
0 20 40 60 80 100
18 time (min.)
12
Finally, a typical network as shown in Fig. 4 was considered
10 to confirm the results of the present gas network simulation.
8
The geometrical data of the network is introduced in Table I
and the gas demand at the nodes 2 and 3 are illustrated in Fig.
6
0 2 4 6 8 10 12 14 16 18 20 22 24 10. The pressure source in the network is node 1 which is
time (Hour) maintained at a constant pressure of 50 bar. The gas specific
Fig. 7 Comparison of pressure time history at the outlet gravity is approximately 0.6, the operational temperature is
278 K, and the friction factor is considered to be constant and
A harmonic demand as shown in Fig. 8 was imposed at the equal to 0.003. The present simulation results are compared
pipe outlet as another test case. From Fig. 9, it is observed that with those obtained by Ke and Ti [10] in Figs. 11 and 12. As
the present transfer function model can well follow the results is shown in the figures a good agreement is observed although
of the implicit FSM after a few minutes. The relatively large some differences exist at the sharp points. This behavior
errors at the initial times are expected because at these times implies that the transfer function model results in the sharp
the outlet pressure does not achieve its purely harmonic changes in the outlet pressure if the demand at the outlet is
behavior. sharp.
International Scholarly and Scientific Research & Innovation 2(7) 2008 877 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
60 TABLE I
Demand from node 2 SPEED COMPARISON BETWEEN USED METHODS
55
Demand from node 3
50 method CPU time (s)
45
transfer function
1.18
Flow Rate (S-m3/sec)
40
(current study)
35
30
implicit method 34.52
25
20 VII. CONCLUSION
15 The proposed simulation can be applied to analyze the
10 transient flow of natural gas in pipelines and networks with a
5
sufficient accuracy. Since the proposed simulation is used the
transfer functions of the transient gas flows, it is more
0
0 2 4 6 8 10 12 14 16 18 20 22 24 computationally efficient than the other finite difference
International Science Index, Mechanical and Mechatronics Engineering Vol:2, No:7, 2008 waset.org/Publication/10550
time (Hour) methods. On the other hand, it is an easy task to analyze the
Fig. 10 Demands versus time for nodes 2 and 3 of the simulated transient flows with any boundary condition types using the
network proposed MATLAB-Simulink library. Moreover, one can
assemble the transfer functions of all the network pipes to
TABLE I simulate the dynamic behavior of a gas network. The present
PIPE GEOMETRICAL DATA FOR THE RELATED NETWORK
study is shown that the proposed simulation extremely reduces
Gas Duct ID
From To Diameter Length the computational time comparing the other numerical
node node (m) (km) schemes. However, because the present simulation is based on
1 1 3 0.6 80
the flow transfer functions it only gives the endpoints results
2 1 2 0.6 90 and not those distributions along the pipelines.
3 2 3 0.6 100
48.8
50.4
48
50 present Transfer Function library
Ke and Ti (1999)
49.6 47.2
Pressure (bar)
49.2 46.4
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (Hour)
48.8
Fig. 12 Outlet pressure results for nodes 3
48.4
48 APPENDIX A
47.6
In this appendix, the algebraic expressions of the
0 2 4 6 8 10 12 14 16 18 20 22 24 parameters used in (19) and (21)-(24) are presented. α, β, γ
Time (Hour)
and b which are used in (19) are stated as [13]
Fig. 11 Outlet pressure results for nodes 2
International Scholarly and Scientific Research & Innovation 2(7) 2008 878 scholar.waset.org/1999.8/10550
World Academy of Science, Engineering and Technology
International Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engineering Vol:2, No:7, 2008
α (s ) = α1 + α2s 1 2 1
d2 = aˆ2 + γ2 + γ2aˆ1
(A-
β (s ) = β1s 8 2 14)
γ (s ) = γ1 + γ2s (A-1)
b (s ) = γ 2 + 4αβ / 2 ACKNOWLEDGMENT
where Khuzestan Gas Company and Shahid Chamran University
are acknowledged for providing technical, administrative, and
α1 = u * fL*, α2 = L / c , β1 = L / c , financial assistance.
fL* * * g Δh (A-2)
γ1 = u u − 2 , γ2 = 2u *L / c REFERENCES
2 c [1] E. B. Wylie, M. A. Stoner, and V. L. Streeter, “Network Transient
Calculation by Implicit methods”, Soc. Pet Eng. J., 1971, 356-362.
The other parameters which have been used in (21)-(24) are [2] C. A. Luongo, “An Efficient Program for Transient Flow Simulation in
[16] Natural Gas Pipelines”, 15th Annual Meeting Pipeline Simulation
Interest Group (PSIG), New Orleans, October 1986.
International Science Index, Mechanical and Mechatronics Engineering Vol:2, No:7, 2008 waset.org/Publication/10550
International Scholarly and Scientific Research & Innovation 2(7) 2008 879 scholar.waset.org/1999.8/10550