Академический Документы
Профессиональный Документы
Культура Документы
net/publication/330845303
CITATION READS
1 4
5 authors, including:
Enbin Liu
Southwest Petroleum University
38 PUBLICATIONS 61 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Enbin Liu on 04 February 2019.
Enbin Liu1,2 , Liuxin Lv2, Qian Ma2, Jianchao Kuang1 and Lu Zhang2
Abstract
A pipeline operation optimization model with minimum energy consumption as the objective function was established
based on the dynamic programming method. The model was applied to a 3840 km gas pipeline whose designed pipeline
capacity was 170 3 108 t/a. There were 40 stations in the line, including 22 compressor stations and 32 compressors.
The solution time was controlled within 60 s to show that the algorithm was fast and effective. The number of starting-
up compressors in the optimized scheme is two more than that in the actual operation scheme, and the total pressure
drop of the pipeline decreased by 3.40 MPa, the average efficiency of the gas turbine units increased by 4.234%, the aver-
age efficiency of the electric drive units increased by 4.875%, and the power decreased by 18,720.38 kW, confirming the
validity and feasibility of the optimization model.
Keywords
Dynamic programming, energy consumption, gas pipeline, operation optimization
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering
1. Objective function: In pipeline optimization, the optimization models, but research on and energy
objective function is usually the maximum consumption models of gas pipelines with very large
throughput,7 the maximum filling of the pipe- compressors based on dynamic programming is insuffi-
line,8,9 or the lowest pipeline energy consump- cient. In our research, based on the characteristics of
tion.10,13 Of these, the latter is the most common long distances and multiple compressor stations, an
focus. In this study, the operation conditions of optimal operating plan was determined that accommo-
a gas pipeline were optimized using the lowest dated the actual conditions of a pipeline. By taking a
energy consumption as the objective function. gas pipeline as an example, we created a relevant
2. Constraint condition: There are usually three dynamic programming optimization model whose
kinds: inequality constraints, equality con- results have shown its feasibility.
straints, and compressor constraints. Inequality
constraints are used to limit the flow, pressure,
and temperature of a pipe within a specific
Modeling and method
range.11 Equality constraints are governing Modeling
equations of gas flow in the pipeline, including
Objective function. Our objective function reflected the
a mass conservation equation, a pressure equa-
objectives of pipeline operations. In the actual situa-
tion, and a temperature equation.10 Compressor
tion, the lowest energy consumption is taken as the
constraints were first established based on ideal
objective function, and because it consists mainly of the
compressor assumptions, which ignored the
energy consumption of compressor stations, the objec-
nonlinear relations between the pressure head,
tive function can be the minimum sum of that energy
power, efficiency, compression ratio, and flow
consumption
of a compressor. Wu et al.12 and Liu et al.13
built up a set of polynomials including surge m X
X n
curves and stagnation curves to describe feasible min F = fi kij , Qij , Pdi ð1Þ
regions that overcame the shortcomings of i=1 j=1
the ideal compressor model. Sanaye and
Mahmoudimehr14 promoted their compressor where F is the total energy consumption of all compres-
constraints by considering correction para- sor stations; fi is the energy consumption of the ith com-
meters relevant to the ambient temperatures, pressor station, i = 1, 2, 3,., m; kij is the switch state
component load running, and excess operations of the jth compressor in the ith compressor station,
of compressors, which were more likely to rep- where kij = 0 means no power on, and kij = 1 means
resent actual compressor operation. In addition power on; Qij is the flow rate of the jth compressor in
to the operation parameters of a compressor, the ith compressor station in m3/s (it is divided evenly
the operation state (on or off) is also an essen- when the compressor configuration in the station is the
tial parameter to be optimized, especially for a same); and Pdi is the outbound pressure of the ith com-
system with a set of compressors.15 pressor station in MPa.
3. Optimization variables: Optimization variables
are arguments of an optimization model, which Optimization variables. The outbound pressure and the
generally consist of pressure, temperature of running state of the compressors are directly related to
each node, flow rate of each unit (pipe and com- the objective function of the optimization model. The
pressor), and the running state and efficiency of energy consumption of a compressor depends on its
each compressor. The pressure, temperature, compression ratio, flow rate, and temperature, but
and efficiency of a compressor are continuous, when the inlet condition of the compressor station is
whereas the running state of a compressor is known, the energy consumption can be simplified as a
discrete.16–18 function of compression ratio and temperature. The
inlet and outlet temperatures of a compressor also
Over the past decades, experts have promoted many depend on its compression ratio, so the optimization
algorithms to solve the pipeline optimization problem. variables can be set to be the compression ratio that
Dynamic programming, generalized reduced gradient, optimizes the outbound pressure. As a result, we set the
and linear programming were the most common. But optimization variables of the pipeline optimization
dynamic programming became the most successful model to be the number of powered-on compressors
algorithm for solving this kind of problem because of and the outbound pressure of a typical compressor
its advantages of ensuring global optimization and easy station
handling of nonlinear situations.19–23
After years of effort by experts, many algorithms are X = Pdi , kij ð2Þ
now available to create natural gas pipeline
Liu et al. 3
where Pi is the node i pressure in MPa, Pi min is the min- where M is the overflow rate of the compressor in kg/s,
imum pressure allowed by node i in MPa, and Pi max is H is the polytropic head of the compressor, and h is the
the maximum pressure allowed by node i in MPa. efficiency of the compressor.
The head curve is calculated according to
Pipe stress constraints. When there are Np pipes in the
gas pipeline system, for the sake of pipeline safety, the H = h1 S 2 + h2 SQ + h3 Q2 ð8Þ
pressure in pipe k must be less than the maximum
allowable operating pressure, which is where h1, h2, and h3 are the fitting coefficients of the
head curve, S is the speed of the compressor, and Q is
the actual overflow rate of the compressor in m3/d.
Pk ł Pk max k = 1, 2, . . . , Np ð5Þ
The efficiency curve is calculated according to
where Pk is the pressure of natural gas in pipeline k in
MPa, and Pk max is the maximum pressure allowed in H
= e1 S 2 + e2 SQ ð9Þ
pipeline k in MPa. h
Compressor power constraints. The power of every the phase) was the outbound pressure (Pd, k1 ) (k 2)
compressor (station) was limited to the characteristics of station k – 1, and the stage effect Vk was the energy
of the compressor consumption of station k. Taking the total energy con-
sumption of the pipeline as the optimization objective,
Nj min ł Nj ł Nj max ðj = 1, 2, . . . , Nc Þ ð12Þ the dynamic programming model for the operation
optimization of each compressor station of the pipeline
where Nj is the power of the jth compressor (station) in
can be established:
W, Nj min is the minimum power of the jth compressor
Stage variables
(station) allowed in W, and Nj max is the maximum
power of the jth compressor (station) allowed in W. xk = Pd, k1 , k = 2;m + 1 ð14Þ
Dynamic programming. When the distribution branch f k ðx k + 1 Þ = min V1, k ½xk + 1 , P1, k ðxk + 1 Þ, k = 1;m
P1, k ðxk + 1 Þ
pipes along a gas pipeline are simplified to discharge
ð19Þ
nodes, the operation of the pipeline can be regarded as
a multistage process; hence, dynamic programming is Function recurrence equation
eligible for optimizing compressor stations along the
pipeline, as shown in Figure 2. fk ðxk + 1 Þ= min ffk1 ðxk Þ+Ck ðxk + 1 , dk Þg, k = 1;m
dk 2Dgk ðxk + 1 Þ
We established the dynamic programming model of
the optimal configuration of each compressor station ð20Þ
in the pipeline as equations (14)–(22) after setting the
Initial conditions
number of compressors as m and considering the gas
transmission process between the compressor station x1 = Pd, 0 ðSupply pressure of gas sourceÞ ð21Þ
k – 1 and station k as stage k of the problem. In that
problem, the state variable xk (for the starting state of and
Liu et al. 5
Figure 4. Length of pipe between each compressor station and number of compressors in each station.
Station number h1 h2 h3 e1 e2 s1 s2 s3 s4
Station Number of Pitted Outbound Pitted Outbound Driving type Unit Power
number starting-up pressure pressure temperature temperature efficiency (kW)
compressors (MPa) (MPa) (°C) (°C) (%)
Daily operation report (the practical operation were all 9.8 MPa, which was the design pressure of the
program). We obtained the process parameters of each pipeline.
station in the case from the practical operation report, Number of starting-up compressors, Power of com-
as shown in Table 2. pressor units, pitted and outbound pressure of each sta-
As shown in Table 2, in the actual operation scheme, tion, pressure drop along the pipeline, average
the total number of starting stations of the compressors efficiency of the driving device are compared in the
was 23—at least one for every station listed, except for daily operation report and the optimized operation
compressor station 12. All the other compressor sta- scheme, as shown in Figures 6–10.
tions were in the starting state, of which only stations 1 In the actual operation scheme, there were 23
and 16 opened two compressors. For a compressor sta- starting-up compressors. In the optimization operation
tion driven by a gas turbine, unit efficiency refers to the scheme, there were 25 starting-up compressors: one
combined efficiency of the gas turbine and the com- compressor was added to stations 2, 5, 7, 9, and 10,
pressor. For a compressor station driven by a motor, and one compressor was removed from stations 1, 4,
unit efficiency refers to the combined efficiency of the and 8.
electric drive and the compressor. The average effi- According to the daily operation report, the total
ciency of a gas turbine unit was 21.09%, the average power used was 245,495.4 kW, and the total power
efficiency of a motor-driven unit was 65.58%, the total used in the optimization scheme was 226,775.1 kW.
pressure drop was 44.67 MPa, and the total power was The optimization scheme reduced the power used by
245,495.4 kW. In actual operation, the energy con- 18,720.3 kW, or 7.6%. The power used by stations 1, 4,
sumption was high, the pressure drop was large, the 8, 13, 16, 22, 24, and 29 decreased substantially, and
unit efficiency was low, and there was still much room the optimization results were obvious.
for optimization. After optimization, the outbound pressures of sta-
tions 1, 3, 5, 6, 9, and 13 fell. In stations 4, 8, 11, 12, 13,
16, 17, 19, 20, 22, 24, 26, and 29, the pressures increased
Optimized operation program. After calculations, we (the pitted pressure was low in station 5 because of a
obtained the optimum operation program shown in pressure crossing), thereby reducing the energy con-
Table 3. In the optimum program, stations 4, 8, and 12 sumption of the compressors.
were powered off; for stations 3, 7, 10, 11, 15, 16, 17, In the daily operation report, the total pressure drop
19, and 20, a total of nine stations, outbound pressures of the pipeline was 46.46 MPa, whereas in the
8 Advances in Mechanical Engineering
Figure 9. Comparison of pressure drops between daily operation report and optimized operation scheme.
Note: section number ‘‘13 + 14’’ refers to pipelines 13 and 14 between the two compressor stations.
X
n
Ni ti
Wg = ge ð23Þ
i=1
hgi
X
n
Ni ti
We = ð24Þ
i=1
hei
Standard GB 2589-81 of 1.33 kgce/m3, v2 is the electric optimization problem for small gas pipelines in 1968.
coal conversion coefficient based on GB 2589-81 of As can be seen in the table, for many years, more and
0.1229 kgce/(kWh), and Tur is the turnover in more experts began to use dynamic programming to
107 N m3 km. optimize the energy consumption of gas pipelines.
Figures 11 and 12 directly reflect that every energy However, one limitation was that the pipelines studied
consumption index was decreased after optimization, were too small (few compressor stations and few com-
which confirms the feasibility and validity of the opti- pressors). Also, some scholars had different views on
mization model. Compared with the daily operation the actual operation of a compressor.
report, the optimization scheme reduced total gas con- Basing our work on our predecessors, in this study,
sumption by 527.45 (104 N m3), total electricity con- we used the dynamic programming method to optimize
sumption by 406.69 (104 kWh), total production energy the energy consumption of a large gas pipeline. The
consumption by 7515.25 tce, and total energy consump- description of the compressors was in line with engi-
tion by 8.65%. Therefore, the optimized operation plan neering practice. (The total length of the pipeline was
can save much energy. 3840 km. There were 40 stations in the line, including
Compared with the daily operation report, the optimi- 22 compressor stations and 32 compressors.) Such a
zation scheme reduced gas consumption by 11.39 (N m3/ large-scale, long-distance pipeline model can be solved
(107 N m3km), electricity consumption by 8.78 (kWh/ in 60 s (a small increase in size can add tens of times
(107 N m3km), and production consumption by more work and computation). The algorithm was fast
16.24 (kgce/(107 N m3km)). The length of the west–east and effective. Moreover, the compressor description
gas pipeline, the number of stations, the volume of trans- combined with the field data measured by the online
mission, and the number of compressors are large. The testing software of the Beijing Oil & Gas Control
proposed optimization program can reduce much energy Center was used to correct the fitting curve, making it
consumption and have great economic benefits. consistent with the actual operation of the compressor.
This had a more advanced and practical value com-
pared with previous treatments of the compressor,
Discussion which used only a simple curve equation and ignored
Table 4 shows that PJ Wong started to use dynamic the actual operation. This also distinguishes our study
programming to solve the gas pipeline operation from previous work.
12 Advances in Mechanical Engineering
1968 PJ Wong and 26 km gas pipeline, 10 compressor stations Compressor energy head equation
RE Larson6
1990 HS Lall and – The way to characterize the performance of a
PB Percell24 compressor is polytropic head and polytropic efficiency
2002 SZ Gao25 1000 km gas pipeline, 7 compressor The working condition was calculated according to
stations, 14 compressors the characteristic curve of centrifugal compressor
2011 D Danilovic26 71.4 km gas pipeline, no compressor station –
2015 X Zhang et al.19 1800 km gas pipeline, 8 compressor Defined by its maximum speed, minimum speed,
stations, 27 compressors, surge line, and stone line
2 sources, and 4 delivery points
2016 J Zhang27 150 km gas pipeline, 10 compressor stations Compressor constraints were mainly feasible domain
constraints and pressure, temperature, and flow of
the nonlinear function of the relation
2017 HA Behrooz28 420 km gas pipeline, 3 compressor stations, Surge line, choke line, maximum allowable speed, and
9 compressors minimum required speed were the four constraints
imposed by the compressor wheel map
2018 PE Du29 1631 km gas pipeline, 7 compressor stations, The characteristic curve of the original compressor
27 compressors factory was numeralized by the variable speed
quadratic fitting technology based on the least squares
principle
2018 E Liu et al. 3840 km gas pipeline, 22 compressor stations, The curve of the compressor was corrected and fitted
(this paper) 32 compressors, according to the online test software of the Beijing
5 injection stations, and 20 distribution stations Oil & Gas Control Center. The compressor’s operating
area is shown in Figure 1
Liu et al. 13
22. Fyk I, Oliynyk M, Kovalchuk Y, et al. Theoretical and 26. Danilovic D. Determination of optimal parameters of dis-
applied aspects of using a thermal pump effect in gas tributive gas pipeline by dynamic programming method.
pipeline systems. East Eur J Enterpr Technol 2018; 1: Petrol Sci Technol 2011; 29: 924–932.
39–48. 27. Zhang J. Study on the adaptability of gas pipeline opera-
23. Liu EB, Peng SB, Zhang H, et al. Blockages detection tion optimization algorithm. Comput Simul 2016; 33:
technology for oil pipeline. J Balkan Tribol Assoc 2016; 369–372.
22: 1045–1057. 28. Behrooz HA. Dynamic optimization of natural gas net-
24. Lall HS and Percell PB. A dynamic programming based works under customer demand uncertainties. Energy
gas pipeline optimizer. Anal Optim Syst 1990; 144: 2017; 134: 968–983.
123–132. 29. Du PE. Steady-state operation optimization of pipeline
25. Gao SZ. Study on optimal operation of compressor station network based on dynamic programming and improved
in gas transmission line. Master’s Thesis, Southwest Petro- genetic algorithm. Oil Gas Storage Transp 2018; 37:
leum University, Chengdu, China, 2002. 285–290.