Evaluation of energy recovery and potential of hydrogen production in Iranian natural gas transmission network

Energy Policy 61 (2013) 65–77

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Evaluation of energy recovery and potential of hydrogen production in Iranian natural gas transmission network Sahar Safarian a,n, Yadollah Saboohi a, Movaffaq Kateb b a b

Sharif Energy Research Institute (SERI), Sharif University of Technology, Tehran, P.O. Box 14597-77611, Iran Department of Electrical and Computer Engineering, University of Tehran, Tehran, P. O. Box 14395-515, Iran,

H I G H L I G H T S







Formulation of a linear model that represent non-linear features of natural gas flow. Development of an optimization model for investment strategy of gas networks. Evaluation of energy recovery by turbo expander and ORC in natural gas networks. Evaluation of hydrogen production in natural gas supply networks.

art ic l e i nf o

a b s t r a c t

Article history: Received 16 October 2012 Accepted 1 May 2013 Available online 2 June 2013

In the natural gas transmission network, from supply points to demand nods there are various technological options that include processing, transportation, conversion and gas distribution. Comprehensive analysis of natural gas network requires evaluation of different chains of gas flow through various levels based on economical and environmental criteria subject to technical and operational constraints such as feasibility, operability and reliability of different alternatives. To aid decision-making process in the sector of natural gas, a generic optimization-based model has been developed for assessing long term energy issues related to planning and design of natural gas supply systems. The model is capable of identifying optimal investment strategies and build up of new capacities of an integrated gas supply system. Evaluation of the potential of energy conservation and hydrogen production in transmission network are also investigated by three energy recovery technologies: turbo expander, ORC and electrolyzer. The model has then been applied in studying the development of Iranian natural gas network. The results indicate the utilization of produced hydrogen by electrolyzer has considerable impact on minimizing the total cost. The total produced hydrogen of the case study is 1337 million kg, in the period 2011–30. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Natural gas network Hydrogen production Energy conservation

1. Introduction 1.1. Background Development and planning of natural gas transmission network involves capital intensive projects that have considerable impact on the energy economy in general and gas-rich countries (such as Iran) in particular. The investment costs and operation expenses of pipeline networks are extensive and marginal improvement in the system planning and utilization can substantially contribute to the economic efficiency of gas supply system (Azadeh et al., 2010; Dinca et al., 2007; Rios-Mercado et al., 2006).

n

Corresponding author. Tel.: +98 21 223 744 87. E-mail address: [email protected] (S. Safarian).

0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2013.05.002

Different aspects of the natural gas supply system have been analyzed in the past decades. The main elements of the network, that have been included in the previous studies, were gas pipelines and compressor stations and the objective has been minimization of the total costs i.e. investment and operational (Larson and Wong, 1968; Graham et al., 1971; Martch and McCall, 1972; Flanigan, 1972; Mah and Schacham, 1978; Cheesman, 1971). Based on dynamic programming, Larson and Wong (1968) developed a steady-state model which considers a pipeline segment connected to a compressor station. Their model estimates the optimal suction and discharge pressure from pipeline. The length and diameter of the pipeline was assumed to be constant due to the limitation of dynamic programming. Martch and McCall (1972) tried to modify the Larson and Wong model by considering a tree network. However, the transmission network was predestined because of the limitations of the optimization technique and assumption on constant pressure.

66

S. Safarian et al. / Energy Policy 61 (2013) 65–77

Cheesman (1971) suggested an optimizing code to enhance the capability of the previous models by taking into account the pipeline variables such as length and diameter of a pipeline. A mixed-integer nonlinear programming technique (MINLP) was introduced by Cobos-Zaleta and Rios-Mercado (2002) that minimizes fuel consumption in natural gas pipeline network. Decision variables in the MINLP were pressure at each node of network, the mass flow rate through the pipeline and the number of compressors to be installed within each station. The steadystate condition was assumed and the total supply flow was driven completely to meet the demand for natural gas with no loss along the network. The model was only capable of solving small problems for specific type of compressor units. Rios-Mercado et al. (2006), proposed a heuristic solution procedure where the fuel cost is minimized within a cyclic network topology. It means that a network containing at least one cycle with two or more compressor stations. The procedure included two stages. At the first stage, gas flow was assumed constant and optimal pressure was estimated. At the following stage, pressure was kept constant and an attempt was made to calculate the rate of flow that might improve the objective function. Their model was appropriate for both cyclic and noncyclic network but it was considerably cumbersome to solve a cyclic network. Kabirian and Hemmati (2007), developed an integer nonlinear programming in order to identify an optimal expansion plan of an existing natural gas network. The minimization of total costs of the network was considered as the objective function. The main outputs of the model have been reported to be type, location and installation schedule of pipelines and compression stations. A review of the existing analytical tools indicates that processing plants and transport units, such as pipeline segments and compressor stations, are considered in the above mentioned studies. Despite the improvement in the modeling of the gas transmission network, it may be concluded that technological options for enhancing energy efficiency at the transport and distribution levels have rarely been considered. But it is to be noted that energy loss in the compressor stations and pressure reduction valves at the distribution level is considerable and it is a determining factor of energy efficiency of natural gas network. Various technological options have been developed for recovery of energy losses in the natural network such as turbo expander and the organic rankine cycle system (ORCs) (Pozivil, 2004; Maddaloni and Rowe, 2007; Desai et al., 2009; Wang et al., 2011). A system of turbo expander is a substitute for the reduction valves in gas pressure reduction stations (GPRSs). And the ORC system retrieves thermal energy from high temperature exhaust gases of gas turbines employed in the gas pressure compression stations (GPCSs). The recovered energy is now available as electricity which can be sold to the grid or used in a local process. The next option is using the previous wasted energy to produce hydrogen gas, compressed and stored, ready for dispensing. Of course, hydrogen infrastructure developments will be worthwhile only if the hydrogen is produced in a sustainable way. This means that the production has to be based on efficient conversion technologies in the proper scale. The advanced hydrogen production technologies are natural gas steam reforming, reforming of fossil fuels, partial oxidation of heavy hydrocarbons and electrolysis (Miltner et al., 2010; Liao et al., 2010). In this study, electrolysis of water is considered as technology for the production of hydrogen, because it is only technology which is fed by electricity. The hydrogen production by electrolysis was widely investigated previously (Grigoriev et al., 2006; Marshall et al., 2007; Jomard et al., 2008; Wrana et al., 2010; Zeng and Zhang, 2010). Behavior of a turbo expander in GPRS has been modeled by Pozivil (2004) using the HYSIS environment. The model consisted

of steady-state calculations based on the isentropic efficiency of a turbo expander. The simulation was carried out at a variety of presumed inlet and outlet conditions and isentropic efficiencies. The model output included generation of electric power, preheating requirements and thermal efficiencies. Maddaloni and Rowe (2007) also investigated utilizing a turbo expander in pressure reduction station to produce electricity. The electricity can either be routed back into the electric distribution grid or used to produce small amounts of hydrogen. In addition to the turbo expander, heat recovery in turbocompressor stations provides a technological option for improving energy efficiency of gas transport network. Heat recovery through utilization of ORC is a viable option which could be considered. Desai et al. (2009), proposed a methodology for optimizing the operation of an ORC for 16 different organic fluids. They have reported 16.5% improvement in the thermal efficiency of an ORC system. Based on thermodynamic model, Wang et al. (2011) examined the performance of 9 different organic fluids at fixed net output power. The thermal efficiency of an ORC with various working fluids was examined based on application of different evaporating pressures and condensing temperatures. 1.2. Further development of model The modeling approaches in the referred studies can be classified according to the techniques used in these analytical tools, such as optimization, simulation, non-linear, linear, static and dynamic. Linear optimization is applied to identify the optimal design or operation of the system. Nonlinear programming problems are constrained optimization problems with nonlinear objective and/or constraint functions. A disadvantage of nonlinear programming over linear programming is that general purpose program is somewhat less effective because the nonlinear paradigm encompasses such a wide range of problems with a great number of potential pathologies and eccentricities. In the sense that certain active constraints become dependent, or are only weakly active. Curvature in the objective or constraint functions (a second-order effect not present in linear programming), and differences in this curvature between different directions, can hinder identification of global optimal point, especially when second derivative information is not supplied by the user or not exploited by the algorithm. A static solution procedure usually assumes a fixed gas demand for a single pathway. The static approach is simple and therefore widely adopted, but has significant restrictions when it is applied for analysis of gas pressure drop within a pipeline segment. Indeed it disregards changes of gas pressure in pipe segments. The static approach, with or without optimization, is inaccurate because it ignores the financial and technical effect of evolving factors (such as gas demand and techno-economic parameters of technologies) on the supply side. Dynamic approaches consider the dynamic changes in the infrastructure over time and how transport from one pathway to another should take place as market conditions change. The dynamic resource allocation problem is an extension of the static problem for finding the optimal allocation of a specific resource to a certain demand at each time period. Since natural gas network is composed of various systems and related energy recovery technologies are novel, a dynamic analysis of the longterm development of gas supply system is vital. Based on the above conclusion, model GNM (Gas Network Model) has been developed in the present research work which shall be described in the present paper. GNM is a dynamic and linear optimization-model which avails itself to long-run planning of the development of natural gas supply system. All supply and

S. Safarian et al. / Energy Policy 61 (2013) 65–77

demand nodes and midstream processes are depicted in the aforementioned model. The midstream technologies include GPRSs, GPCSs among others. Objective function of the model is minimization of total discounted costs of the gas supply system subject to meeting the natural gas demand. The specific feature of the model includes planning of capacities of new technologies and investment strategies at the beginning of each period. In addition the GNM provides means of studying both potential of power recovery in GPRS using turbo expander system and thermal energy recovery in GPCS utilizing ORC. At the next pathway the generated electricity can either be sold to the power grid or consumed in a local process or used to produce hydrogen by electrolyzer. The model GNM has been applied for studying natural gas supply system in Iran and the results of the case study shall be discussed to demonstrate the potentials of the application of the model.

2. An overview of GNM 2.1. Concepts and structure of GNM GNM is a techno-economic optimization model of natural gas supply system based on a quasi-dynamic linear programming. It was developed as an extension of ESM (energy system model) (Saboohi, 2002). The specific application of the model includes natural gas supply system in Iran over a time span between 2011 and 2030. GNM minimizes the total discounted costs of an energy supply system (which includes Capital, Operation, Maintenance and Resource costs) over planning time horizon and it represents the flow of natural gas from resources to the end users. The energy and material balances of the model ensure that the energy demand in various nodes will be covered. Natural gas demand is exogenous to the model and it is obtained by regression analysis of historical data on demand and its growth rates. The objective function of the model is Z ¼ minðPresent Value total costÞ ¼ Present Value tot; inv þPresent Value O&M

ð1Þ

67

Present Value tot; inv ¼

n

t¼1

n

Present Value O&M ¼ ∑

∑

t¼1

Kt  Yt ð1 þ iÞt

M t ðY t þ H t Þ Ot  X t  ηt þ ð1 þ iÞt ð1 þ iÞt

ð2Þ ð3Þ

where Kt is the capital costs per unit capacity of the technology at time point t, Yt is the new capacity build up of the technology at time point t, Mt is the maintenance costs per unit capacity of the technology at time point t, Ot is the operation costs per unit main output of the technology, Ht is the historical capacity that can be operated at time t, Xt is the input to the technology at time point t, ηt is the efficiency of technology at time point t, i is the discount rate, and n is the plant service life. 2.2. Natural gas flow diagram Gas supply system is segregated into many control volumes that are identified as sub systems and elements of the system at aggregated level. The flow of energy between various control volumes is a reflection of the relationship between sub systems and it defines the structure of the system, which is shown as flow diagram (Fig. 1). Resource, processing, transport, motive-power, electrolyzer, conversion and demand are the main levels of natural gas flow diagram that each level is decomposed into many technologies with a certain structured data format. A technology is the basic element in the systematic approach of the GNM. The whole structure of the model has been developed as a composition of technologies and the interrelationships of the control volumes are identified by the input/output structure of technologies. Input/ output of each technology shows the flow of energy and it is defined on the basis of mass and energy balances. The mentioned levels of diagram are explained in further sectors. The model structure is shown in Fig. 2, which represents the input/output data to/from the model. Input data of the model has been defined in 4 categories of system boundary, system technologies, resources and system surroundings. System boundary is envisaged to define the time dimension of the system and it contains time periods and base year. A system is defined as aggregation of interconnected control

Fig. 1. Natural gas supply flow diagram.

68

S. Safarian et al. / Energy Policy 61 (2013) 65–77

The amount of primary resources and profile of historical natural gas flow rate are the necessary information for the resource level that is obtained from the Iran Hydrocarbon Balance Book (2009). Flow of energy into the extraction technology originates from fields of resources and it may be satisfied according to Rf t −∑ P f tϕ ¼ 0

ð5Þ

ϕ

where Pftφ is the input of energy into extraction technology φ at time point t. Flow through extraction technology is limited by the capacity constraint and it may be formulated as inequality

Fig. 2. General GNM structure.

volumes as sub systems. Each sub system is then segregated into various technologies that define the functionalities of the control volumes. The data of each technology is structured as below:












Input to technology Output of technology Capital phasing of investment Plant factor and unit costs Production in base year Profile of historical capacities Efficiency of each technology Bound on flow Bound on build up of new capacity

Resources are input to the whole system and it contains the amount of primary resources, profile of historical flow rate, costs of resource or value of by product and upper bound on production. Moreover, system surrounding includes data related to the presentation of the interactions of the system with the surrounding and its elements are: discount rate, annual and regional demand and prices. Output data has been determined in 5 categories of solution status, investment strategies, balance of demand/supply, the amount of flow rate and new capacity build up. Solution of the model indicates the quantities of all state and control variables of the system at the optimum point. Costs indicate the extent of use of resources and production factors in the gas supply system. They are considered as major criteria in identifying the optimal development path of the system. In addition, the output data of processing/conversion technologies show functionality of the system. Data show the state of the technologies at the optimum point. The main data categories of each technology are: flow rate, capacity and costs. Transport network is an integrated combination of many segments and flow in each segment shows the operation state of the transport network. Information on the state of each segment at the optimum point is also indicated by 3 categories of flow rate, capacity and costs.

2.2.1. Resources and processing Gas reserves are categorized as exhaustible energy resources and they are depleted as production of gas proceeds. Total production from a certain field cannot exceed total proved reserves. This point may be represented by the equation of depletion and it is envisaged in the model according to the constraint T

∑ Rf t ≤Rf

t¼1

ð4Þ

where Rft is the production of gas from filed f at time point t, Rf is the total proved reserves of field f at initial time point, and T is the planning time horizon.

t P f ϕt  ηϕt − ∑ ΔL ω ¼ ðt−PLÞ

or ω ¼ 1

Y f ϕω  PF f ϕt ≤

b

∑

θ ¼ b−ðPL−tÞ

H f ϕθ  PF f ϕt

ð6Þ

if ω o1

where ηφt is the efficiency of extraction technology φ that makes input energy to output energy at time point t, Yfφω is the new build capacity of technology φ for producing primary energy at time point ω, PFfφt is the plant factor of technology φ for producing energy at time point t, Hfφθ is the historical capacity of technology φ for producing primary energy at time point θ, where θ is a point from (b−θ) to b and b is the base year and PL is the plant service life, and Δl is the time length of load zone l. Processing plant is supposed to take primary energy from fields and it produces secondary energy which is delivered into the transport network. Input into a processing plant is the sum of production of energy fields connected to a specific processing plant. The relationship between energy production and feeding into the network is indicated in ∑ ∑ P f tϕ  ηϕt −∑ Aτt ¼ 0 f

ð7Þ

τ

ϕ

where Aτt is the input of energy into processing technology τ for producing secondary energy at time point t. Flow through processing technology is limited by the capacity constraint and it may be formulated as inequality t Aτt  ητt − ∑ ΔL ω ¼ ðt−PLÞ

or ω ¼ 1

Y τω  PF τt ≤

b

∑

θ ¼ b−ðPL−tÞ

H τθ  PF τt

ð8Þ

if ωo 1

where τ is processing technology. The necessary data such as plant life, efficiency and profile of historical capacities of processing technologies are retrieved from the Detailed Statistics of the National Iranian Gas Company (2009).

2.2.2. Transport Pipeline throughput (flow rate) is a key parameter that depends upon the gas pressure drop through the pipeline and other physical parameters such as gas properties, pipe diameter and length, gas pressure and temperature. Therefore, the natural gas pressure drop is an important state variable that is determined in the transport level. Several equations are available to estimate the pressure drop in a pipeline at a given flow rate, such as: AGA, Colebrook–White, Panhandle, and Weymouth equations (Shashi Menon, 2005). If we start with a fixed upstream pressure in a pipe segment at a given flow rate, these equations will enable prediction of different downstream pressures. A review of the application of the aforementioned equations indicates that some equations overestimate the pressure drop for the same flow rate in a specific segment than the others. The highest pressure drop is predicted by the Weymouth equation. Therefore to be on the safe side, Weymouth equation has been used in the present work because it is the most conservative flow equation that overestimates

S. Safarian et al. / Energy Policy 61 (2013) 65–77

pressure drop. Weymouth equation is as follows: Q ¼ 3:7435  10−3 E


Tb Pb

"

P 1 2 −es P 2 2 GLe ZT ave

#0:5 D2:667

ð9Þ

where Le ¼ L

ðes − 1Þ s

s ¼ 0:0684 G

ð10Þ


 H 2 −H 1 ZT f

ð11Þ

Q is the gas flow rate, standard m3/day, Tb is the base temperature, K, Pb is the base pressure, kPa, Tf is the average gas flow temperature, K, P1 is the upstream pressure, kPa, P2 is the downstream pressure, kPa, Z is the gas compressibility, Le is the equivalent length of pipe segment, km. Weymouth equation represents a non-linear relationship between flow rate and pressure drop. But the natural gas network model of GNM is based on LP formulation in order to be able to identify the global optimal solution. Therefore, inclusion of Weymouth equation in GMN model could lead to a non-linear formulation of the gas network model. Since the model GNM is of

large scale type and it reflects the complex interaction of different levels of gas network, identification of global optimal solution of a non-linear problem shall be rather impossible. Therefore, transformation of the non-linear formulation to a convex programming technique could facilitate the numerical solution. It is, therefore, necessary to identify a linear equation that would represent the numerical approximation to the Weymouth equation over the pressure range that is dominant in the natural gas network. Hence, several pipeline segments with different lengths, diameters and heights have been considered in order to be able to identify the correlation between flow rate and pressure drop. Then, the flow rate in each differential segment is estimated using the Weymouth equation. Variation of pressure drop in the pipeline is then plotted and the functionality of pressure over flow rate is estimated by means of regression analysis. Fig. 3 shows the variations of pressure drop against flow rate. The regression analysis of the data yields the following relationship: Pressure drop ¼ 109:9Q −2078

ð12Þ

The cost of gas pipeline delivery depends on the installed capital cost of the pipeline, and operating and maintenance costs (O&M costs). The total capital cost of the pipeline is a function of diameter and distance; the O&M costs are also functions of capital cost (Najibi et al., 2009). Size of the pipeline is determined according to the gas demand; however, the design capacity of a pipeline segment is higher than the average flow rate to account for time variations in flow. Therefore, pipelines with diameters of 42 to 56 in. are considered for gas transport and diameters of 10 to 16 in. are assumed for gas distribution. Fig. 4 displays a simple schematic of a cyclic and tree network. Considering the flow conservation equation at each node is a vital element of gas network model (O'neill et al., 1979; Borraz-Sanchez and Haugland, 2011): ∑ Q ij ¼ ∑ Q ji þ Q i

ði;jÞ∈A

Fig. 3. Variations of pressure drop against flow rate.

69

ði;jÞ∈A

ð13Þ

At a supply node i, the gas flow Qi, must remain within limitations that are usually specified in the supply contracts. A gas supply contract indicates an average daily quantity to be taken by the transmission network from the supply nodes. Depending on the flexibility of the contract, the transmission network has the possibility of lifting a quantity ranging between a lower and an upper bound on the average contracted quantity. Such a bound

Fig. 4. Simple schematic of a cyclic and tree network.

70

S. Safarian et al. / Energy Policy 61 (2013) 65–77

may be represented by relationship (14) as below: Q i;min ≤Q i ≤Q i;max

ð14Þ

Also, at each exit point, the demand must be satisfied at a guaranteed low pressure. On the other hand, the gas transmission company cannot take gas at a pressure higher than insured maximal permissible pressure: P i;min ≤P i ≤P i;max

ð15Þ

In addition to the mention equations, the constraints related to energy and material balance should be satisfied. Energy input into the network takes place in the region where processing or conversion plant exists. Flow of the output of the processing or conversion plant into the transport network is represented with the help of ∑ Aτt  ητt −∑ Q rkt ¼ 0 τ

ð16Þ

k

where Qrkt is input of energy into transport segment in region r delivering into region k at time point t. Capacity constraint of transport segment is based on inequality Q rkt  ηrkt þ Q krt  ηkrt − ΔL ω

t

∑

¼ ðt−PLÞ

or ω ¼ 1

≤

b

∑

θ ¼ b−ðPL−tÞ

H rkθ  PF rkt

Fig. 5. Variations of compressor power against flow rate.

The following linear Eq. (19) has been obtained through regression analysis of the estimated cross sectional data

Y rkω  PF rkt

W ¼ 5:626Q −234:8

if ω o 1

ð17Þ

The 1st term of inequality Eq. (17) reflects the possibility of transport of gas in two directions with the same capacity. In other words, it is possible to transport gas in a certain segment of the transport network in a load zone in one direction, i.e. from r to k, and in the other load zones in opposite direction, i.e. from k to r. Inclusion of inequality Eq. (17) provides appropriate means of studying the impact of flow in reverse directions on the load curve of energy.

2.2.3. Compression as a level of transformation Compression is assumed to be a compressor in GPCS. According to natural gas flow diagram, the low pressure gas and power are two input flows to the level that represents the first transformation of sensible potentials of natural gas. The required power for driving the compressors at the first transformation level is the most important parameter and it is related to the level of pressure and flow rate as follows (Woldeyohannes and Abd Majid, 2011):


 γ hz 1 þ z 2 i 1 P 2 γ−1 W ¼ 4:0639T 1 Q ð Þ γ −1 ð18Þ γ−1 ηa P 1 2 where W is the required compression power, kW, γ is the ratio of specific heat capacities of gas, dimensionless, Q is the gas flow rate, Nm3/day, T1 is the gas suction temperature, K, P1 is the gas suction pressure, kPa, P2 is the gas discharge pressure, kPa, Z1 is the gas compressibility at suction conditions, dimensionless, Z2 is the gas compressibility at discharge conditions, dimensionless, and ηa is the compressor adiabatic (isentropic) efficiency, decimal value. The employed equation in the first-conversion level is nonlinear too. Therefore, it is necessary to identify a linear relationship that would represent the behavior of a compressor and it would avail itself to estimating the driving force of the compression. Inclusion of such a linear relationship in the GNM could provide a means of transforming it to a linear programming problem. Consequently for a given segment, the compressor power is calculated for various operating conditions based on assumed flow rate of natural gas using the above equation. The results of calculation may be observed in Fig. 5.

ð19Þ

2.2.4. Motive power 2.2.4.1. The compressors drivers. In order to represent the alternative technologies for driving the compressors, a level indicating the motive power supply to the compressors has been considered in the reference system of natural gas network (Fig. 1). The compression of gas is usually driven by gas turbines or electric motors. Gas turbines are more often used in the natural gas network since electric power may rarely be available in remote areas. However, there are two main energy losses in current compression stations based on the utilization of gas turbines: 1. In some cases, such as Iranian gas network, it is estimated that 3–5% of the transported natural gas is consumed in the turbines of GPCSs in order to compensate for the lost pressure of natural gas along the gas pipeline (Wu et al., 2000; Borraz-Sanchez and Rios-Mercado, 2005). 2. Utilization of traditional gas turbines is involved with high temperature exhaust gases from gas turbines which are released into the environment and it leads to serious environmental pollution. However, the sensitive heat of exhaust gas from turbine can alternatively be used to produce additional electric power via a bottoming cycle (Desai et al., 2009; Polyzakis et al., 2008). The above mentioned technologies and issues related to the energy losses in the gas turbine are explicitly represented in the natural gas flow diagram which provides the conceptual basis for GNM. Input flow to electric motor is electricity. The required electric power can be supplied by two options. First option would be that the electricity is purchased from the power distribution grid and the next option might be the electricity generated in an ORC which may utilize the sensible heat of exhaust gas from gas turbine. Power input into an electric motor would be linear function of motor efficiency and the power required in the process of compression Pe ¼

W ηm

ð20Þ

Pe is electric power feed into the motor (kW), W is shaft work (kW) required in the process of compression and ηm is electric motor efficiency.

S. Safarian et al. / Energy Policy 61 (2013) 65–77

If a gas turbine technology is utilized as a driver for compressor, a fraction of pipeline throughput should be consumed as fuel to the gas turbine. The fuel required by the gas turbine is also a linear function of turbine thermal efficiency and the shaft work required by the compressor. The required fuel for gas turbine may be estimated with the help of the relationship (Ouwerkerk and de Lange, 2006): Q f uel ¼

W turb ηth

minlet gas ¼

Q f uel LHV NG

ð21Þ ð22Þ

Qfuel is extractable energy from hot gas (kW), Wturb is shaft work (kW) demanded by the compressor and ηth is thermal efficiency of gas turbine power plant. Moreover, minlet gas in Eq. (22) is the required amount of inlet gas to turbine, which is consumed as inlet fuel and LHVNG is lower heat value of natural gas (kJ/kg). In addition to the power output by the gas turbine, the hot flue-gas from turbine is also an energy carrier which could be recovered and used further. The potential energy of flue-gas can be taken into account with the help of energy balance equation around the gas turbine power plant as follows: Q f uel ¼ W turb þ Ef luegas

ð23Þ

2.2.4.2. Organic rankine cycle system. The interest for low and medium grade heat recovery grew in the past three decades when the scarcity of primary energy resources becomes a major issue. ORC has been the most widely applied solution among many proposed technological alternatives. The basic flow diagram of ORC is represented in the natural gas flow diagram of GNM where electric power is considered as the useful outlet flow. The generated electricity can either be delivered to the power distribution grid or consumed as inlet flow into electric motor. Electric power generated by ORC may be considered as a linear function of thermal efficiency (ηth, ORC) and the sensible heat of flue-gas as follows (Wei et al., 2007; Quoilin et al., 2010; Papadopoulos et al., 2010): Electricity ¼ ηth:ORC  Ef luegas

expander (Maddaloni and Rowe, 2007). Fig. 6 shows a simple schematic of a turbo expander system. The main advantage of turbo expander system over the expansion valve is power generation together with delivery of very low pressure gas. Based on natural gas flow diagram, generated electricity can be delivered to the power distribution grid as an alternative. The turbo expander outlet power and the required fuel for preheating are two principal variables in the second conversion level. The turbo expander system is represented in the GNM with the help of following relationships W exp ¼ mNG ðh2 −h3 Þηexp

ð26Þ

Q preheat ¼ mNG ðh2 −h1 Þ

ð27Þ

mf uel ¼

Q preheat LHV NG ηHEX

ð28Þ

mfuel is the amount of fuel consumed in the preheater (kg/s), h is enthalpy (kJ/kg), ηexp is turbo expander efficiency and mNG is gas flow through the turbo expander system (kg/s). It is important to note that the specific enthalpies in Eqs. (26) and (27) are non-linear functions of temperature and pressure. Hence it would be necessary to identify alternative linear relationships that would represent the function of turbo expander and could be embedded in the model GNM. The operation of an integrated system of turbo expander and heating system has been simulated and the output power and heat duty are computed for various assumed gas flow rates at different operating conditions. Fig. 7 shows the power outlet and the required heat duty against gas flow rate. The regression analysis of simulated data provides a means of substituting the above two non-linear equations with the following linear relationship between output power of turbo expander

ð24Þ

2.2.5. Electrolysis of water system The electrolyzer converts the electrical energy received from both ORC system in GPCSs and turbo expander in GPRSs to chemical energy splitting water into oxygen and hydrogen. The produced hydrogen is linear function of electrolyzer efficiency and inlet electricity Hydrogen ¼ ηelectrolyzer  Electricity

71

Fig. 6. Simple schematic of turbo expander system.

ð25Þ

2.2.6. Second conversion In accord with natural gas flow diagram the second conversion level includes GPRSs. Natural gas leaving the compression stations flows into the distribution grid in the natural gas demand centers. The pressure of natural gas must be reduced at the demand points. The traditional technical facilities for reducing pressure are expansion valves. But the potential work associated with the high pressure of natural gas is destroyed in them. Therefore, turbo expanders may be used as substitutes to the expansion valves and part of the potential work could be recovered. The most important operational problem of turbo expander is hydrate formation that occurs due to the presence of slight amount of humidity in gas composite. Two factors that intensify hydrate formation are low temperature and high pressure. Therefore, gas should pass through a heater before entering the turbo

Fig. 7. Variations of turbo expander system outlet energies against gas flow rate.

72

S. Safarian et al. / Energy Policy 61 (2013) 65–77

and the rate of flow of natural gas W exp ¼ 1044 Q þ 926:7

ð29Þ

Q preheat ¼ 1164 Q þ 1041

ð30Þ

2.2.7. Energy conversion Energy conversion plant is supposed to take primary or secondary energy to produce secondary energy in other forms. According to natural gas flow diagram, compressor, motive power, electrolyzer and GPRS are considered as conversion technologies. Input into an energy conversion plant is the sum of delivery through network. The relationship between energy production and feeding into the network is indicated in ∑ Eeδt  ηδt −∑ C ejst ¼ 0

ð31Þ

s

δ

where Eδt is the input of energy carrier e to technology δ connecting to conversion technologies at time point t, ηδt is the efficiency of technology δ at time point t, and Cst is the input of energy e into conversion technology s for converting into energy j at time point t. Demand of energy at conversion level (e.g. level of compressor) is the sum of external demand (i.e. transport of energy) and delivery to the conversion system. Flow through conversion technology is limited by the capacity constraint and it may be formulated as inequality t C ejst  ηejst − ∑ ΔL ω ¼ ðt−PLÞ

or ω ¼ 1

Y ejsω  PF ejst ≤

b

∑

θ ¼ b−ðPL−tÞ

H ejsθ

if ω o1

 PF ejst

ð32Þ

ηejst is the efficiency of conversion technology s that converts energy e into secondary energy e at time point t, Yejsω is the new build capacity of technology s for producing energy j at time point ω, where ω is a point from 1 to t, PFejst is the plant factor of technology s for producing energy j at time point t, and Hejstθ is the historical capacity of technology s for producing energy j at time point θ, where θ is a point from (b−θ) to b and b is the base year and PL is the plant service life.

where Dɣt is the input of gas to distribution technology ɣ (second conversion) connecting to demand points at time point t, ηɣt is the efficiency of distribution technology ɣ at time point t, and Ut is the demand for gas at time point t. 3. Case study The GNM has been developed on the basis of gas flow diagram and considering the physical and technical constraints that predominate the flow of natural gas within an interconnected network of transformation and distribution. The application of the model has then been demonstrated with the help of a case study. The case study includes three provinces in the southern part of Iran where natural gas resources are mainly located and demand for natural gas is considerable in these regions. 3.1. Natural gas demand Demand is the last level in natural gas flow diagram and it is assumed as given and it is an exogenous parameter to the model. In the GNM, the demand level is divided into two sectors of Very Low Pressure Gas (VLPG) and Low Pressure Gas (LPG). 1. VLPG is the GPRS outlet gas for the demand points which are located within the case study region. In the present study, 3 states of Esfahan, Yazd and Kerman and also city of Shiraz are assumed as the VLPG demand nodes in the network. 2. It is further assumed that there would be export of natural gas from the region to the other provinces in the country which are not included in the case study. Demand for natural gas outside the region being studied is considered as LPG type.

2.2.8. Demand constraints Balancing of supply and demand is a characteristic of the model and it is guaranteed through the demand constraint ∑ Dγt  ηγt ≥U t

ð33Þ

γ

Fig. 8. The LPG and VLPG demand.

Fig. 9. Spatial dimension of the model.

S. Safarian et al. / Energy Policy 61 (2013) 65–77

The LPG and VLPG demand centers of the case study are shown in Fig. 8. Demand for natural gas at the level of end use is assumed to follow the trend that has been estimated by National Iranian Oil Company, (Aghayan, 2008).

73

of hydrogen which is sold to industries in Iran is an average 8000 cents per kilogram ($80/kg) (Qadrdan et al., 2008).

4. Results and discussions 3.2. Definition and assumptions As depicted in Fig. 9, the considered study location is southern region of Iran which extends to Esfahan province and it consists of 3 gas processing plants (Asalouyeh, Valiasr (Fajr) and Bid Boland 1), 7 main gas pipelines and 40 GPCSs. The main assumptions considered in the application of GNM in the present case study are as follows. The inlet and outlet gas composition of all gas processing plants would be similar. Molecular composition of methane in natural gas at the inlet and outlet has been assumed to be 85% and 88%, respectively. Other thermodynamic assumptions used in the present analysis are reported in Table 1 and they are taken from relevant references (Maddaloni and Rowe, 2007; Chacartegui et al., 2011; Shengjun et al., 2011; Ghazikhani et al., 2011; Kovalev et al., 2003; Jomard et al., 2008). The study time horizon is 2011 to 2030 with hot and cold seasons as two load zones and discount rate is assumed to be 10% throughout the planning horizon. The costs of natural gas and electricity supply from power grid have been assumed 35 cents per cubic meter (0.35 US $/Nm3) and 5 cents per kilowatt hour (0.05 US $/kWh), respectively. The price

Table 1 Thermodynamic assumptions. Turbine thermal efficiency ORC thermal efficiency Compressor isentropic efficiency Electric motor efficiency Electrolyzer efficiency Turbo-expander isentropic efficiency Heat exchanger pressure drop

35% 10% 88% 90% 75% 80% 2%

The application results of GNM provides information on the flow of natural gas through different levels of reference system, capacity expansion of alternative technologies, required capital investment that would support the development of gas network. The flow of natural gas from each processing plant is displayed in Fig. 10. It can be observed that the production in Asalouyeh processing plant rises sharper than Valiasr and Bid Boland processing plants. In addition, the costs of supplying natural gas i.e. capital costs, current (excluding the costs of natural gas used as fuel) and resource costs (costs of electricity and natural gas from grid) for existing units which are shown in Table 2. 4.1. Gas compressor stations Fig. 11 depicts the trend of total capacity of compressors according to the assumed increasing trend of demand in the present case study. The GPCSs consist of several series and parallel compressors. When the demand for natural gas rises the demand for compression intensifies. Hence, an expansion of compression capacity is anticipated. It may be observed in the natural gas flow diagram that there are various technological options available at the level of motive power. These options for supplying motive power include following cases: 1. Traditional gas turbine which releases exhaust gases into the environment 2. Electric motor that uses electric power from the power grid 3. Combination of traditional gas turbine and electric motor that uses power from grid 4. Gas turbine together with ORCs and delivering electric power to the grid 5. Gas turbine and electric motor which uses electric power generated by ORCs 6. Gas turbine together with ORCs and electrolyzer and delivering hydrogen Different technological options for supplying motive power to compressors are summarized in Table 3. The results of the model indicate that combination of gas turbine, ORC and electrolyzer presents the optimal alternative (Fig. 12). On the grounds of the high price of hydrogen in Iran and being electrolysis of water as a very efficient process, option 6 has a major impact on the total cost. Although the electricity is purchased from electric grid in option 2, the use of electric motor is cheaper than turbine (option 3). It is due to the relatively high efficiency and low capital cost of electric motor. So the usage of turbine is not recommended except along with ORC (option 4). Of course the results from GNM suggest installing ORCs in existing network which is based on using turbine. Application of ORC has considerable impact on

Fig. 10. Gas supply from the processing plants.

Table 2 Costs of natural gas network including capital, current and resource costs, (MM denotes million). Cost Investment cost Current cost Resource cost

MM. $ MM. $ Billion $

2011

2012

2013

2014

2015

2020

2025

2030

2303 589 47.02

143 620 49.74

205 668 53.16

330 718 57.15

6.19 767 61.84

1246 1012 82.67

3268 1273 105.6

1553 1633 136.1

74

S. Safarian et al. / Energy Policy 61 (2013) 65–77

minimizing total costs which can be understood from comparison of options 1 and 4. From the above statements it can be concluded the combination of options 2 and 4 i.e. option 5 is the best choice after scenario 6.

4.2. Organic rankine cycle and electrolyzer systems Information obtained with the help of applying GNM indicates that the generated power and hydrogen by using of ORC and electrolyzer in GPCSs of case study gas network would increase (see Fig. 13). Changes in power and hydrogen generation are due to the fact the natural gas flow rises which stems from increasing demand for natural gas. High flow of natural gas requires high rate of compression and the exhaust heat from gas turbine in the compressor station increases consequently. Total electric power capacity of ORC includes the existing plants and the incremental addition of new capacities. The development of incremental new capacities is shown in Fig. 14. Since there exists no capacity of ORC in the present system, build up of large capacity of ORC and its integration in the existing system is preferred. Therefore, build up of large amount of new capacity of ORC is observed in the results of the model which indicates high potential of energy recovery in the actual natural gas network. Thereafter, incremental addition of new ORC system is observed. It can also be seen that further expansion of ORC in the later periods (after 2015) would be part of an optimal plan. It has also been found that delivery of generated electric power with the help of ORC to the power grid could provide an appropriate means of integrated resource management.

4.3. Gas pressure reduction stations The demand nodes inside the case study are assumed to be of VLPG type. Three alternative technologies have been considered in the supply chain of VLPG end users as described below: 1. Supplying VLPG through expansion valves 2. Generation of electric power by means of turbo expander in the supply chain of VLPG 3. Combination of expansion valve and turbo expander and delivering electricity to the grid 4. Turbo expander together with electrolyzer and delivering hydrogen According to economic evaluation of GNM (Fig. 15), the results present combination of turbo expander and electrolyzer as the optimal option. Owing to the fact that the price of hydrogen is much higher than electricity in Iran; the GNM does not recommend the usage of turbo expander except along with electrolyzer (comparison of scenarios 3 and 4). In spite of higher capital cost of turbo expander than expansion valve, option 2 is more economical than option 1. It is due to delivering power generated by turbo expander. In order to explain the prevailing trends of utilization of turbo expander, the supply profile of VLPG demand is shown based on option 3 (Fig. 16). It may be seen in Fig. 16 that a high fraction of VLPG demand is met by flow of natural gas through expansion valves in the initial and immediate years. Such a finding indicates that the economies of scale have considerable impact on the penetration of technologies with high capital costs. Since demand of natural gas is relatively low in the initial period utilization of capital intensive turbo expander with low electricity prices is not preferred. Increase in demand for natural gas facilitates the utilization of large scale turbo expander systems. Development of electric power and hydrogen generation with the help of turbo expanders and electrolyzer are shown in Fig. 17. The slopes of the profiles are changed after 2015 where a similar pattern in demand for natural gas is observed. Rising trend of electricity generation is due to the utilization of turbo expander that is preferred to the use of pressure reduction valves. The results also indicate that entire natural gas flows through turbo expanders in the distribution systems in year 2030.

5. Conclusion A linear optimization-based model for the long-range planning of natural gas supply system has been developed and introduced in the present research work. The objective function of the model is minimization of discounted value all related operating and investment costs which may occur in the planning horizon. Enhancement of mechanical potentials and destruction of potential of mechanical work are indispensible features of a natural gas supply network. Therefore, utilization of energy recovery

Fig. 11. Development of required capacity of compression.

Table 3 Different technological options of supplying the required motive power of gas compressors. Option #

Gas turbine

1 2 3 4 5 6

■ ■ ■ ■ ■

ORCs

Electric motor

Electrolyzer

Delivery of power

■ ■ ■ ■ ■

Purchasing electricity

Delivery of hydrogen

■ ■ ■ ■

■ ■

■

S. Safarian et al. / Energy Policy 61 (2013) 65–77

technologies in the natural gas supply network is technically feasible and they may contribute to the economical efficiency of natural gas supply system. It is, hence, essential to embed energy recovery options in the analytical tools. Natural gas flow rate is a function of mechanical potentials, i.e. pressure, which leads to complicated formulation of technical constraints on the gas supply system. The aim of the present work has been to utilize the concept of convex programming and to reformulate the non-linear relationship between gas flow rate and

pressure in the form of a set of simultaneous linear functions. Such formulation of the relationship between gas flow rate and pressure has enabled application of linear programming in developing a model of natural gas supply system. Global optimal point of gas supply system is therefore identified when the natural gas supply model of GNM is solved.

Fig. 15. Total cost of natural gas supply system for different options.

Fig. 12. Estimated total costs of natural gas supply system for assumed options in Table 3.

Fig. 13. Total generated electricity and hydrogen by ORC and electrolyzer.

75

Fig. 16. Supply profile of VLPG based on option 3.

Fig. 14. New capacity expansion of ORC.

76

S. Safarian et al. / Energy Policy 61 (2013) 65–77

Fig. 17. Total generated electricity and hydrogen by turbo expander and electrolyzer.

Application of the model has been demonstrated with the help of a case study. The results of application of the model provide information on optimal development of gas supply system which include investment strategies, capacity expansion of alternative technologies, penetration of energy recovery and hydrogen production systems, alternative routes for transmission of natural gas and capacity build up of gas processing and distribution systems in the planning horizon. The results of the model indicate that utilization of gas turbines as driving systems for compressors together with the ORC energy recovery system and delivering hydrogen produced by employment of electrolyzer are the most appropriate technological alternatives that should be considered in the gas compressor stations. Based on the optimal option, the total produced hydrogen in GPCSs is 1170 million kg in the period 2011–30. Demand nodes are identified as Very Low Pressure Gas (VLPG) and four alternative options of supplying natural gas to demand nodes are considered in the model which included use of pressure reduction valves, turbo expanders and combination of both. The results show that utilization of turbo expander is the most economical option after 2015 when the natural gas flow through distribution system rises. In addition, based on minimizing total cost the GNM recommends the employment of turbo expander together with electrolyzer and delivering the produced hydrogen. The total produced hydrogen in GPRSs is approximately 167 million kg in the period 2011–30.

References Aghayan, H., 2008. The Oil Products Consumption Information. National Distribution Company of Oil Products, Tehran. Azadeh, A., Asadzadeh, S.M., Ghanbari, A., 2010. An adaptive network-based fuzzy inference system for short-term natural gas demand estimation: uncertain and complex environments. Energy Policy 38, 1529–1536. Borraz-Sanchez, C., Haugland, D., 2011. Minimizing fuel cost in gas transmission networks by dynamic programming and adaptive discretization. Computers & Industrial Engineering 61, 364–372. Borraz-Sanchez, C., Rios-Mercado, R.Z., 2005. A hybrid meta-heuristic approach for natural gas pipeline network optimization. Lecture Notes in Computer Science 3636, 54–65. Chacartegui, R., Sanchez, D., Munoz, A., Sanchez, T., 2011. Real time simulation of medium size gas turbines. Energy Conversion and Management 52, 713–724. Cheesman, A.P., 1971. Understanding origin of pressure is a key to better well planning. Oil and Gas Journal 69 (46), 146. Cobos-Zaleta, D., Rios-Mercado, R.Z., 2002. A MINLP Model for a Minimizing Fuel Consumption on Natural Gas Pipeline Networks. Memorias del XI Congreso

Latino Iberoamericano de Investigacion de Operaciones (CLAIO), 27–31 de Octubre de. Desai, Y., Wang, J., Gao, L., 2009. Parametric optimization and comparative study of organic Rankine cycle (ORC) for low grade waste heat recovery. Energy Conversion and Management 50, 576–582. Detailed Statistics of the National Iranian Gas Company, 2009. Department of Management Information Systems. Dinca, C., Rousseaux, P., Badea, A., 2007. A life cycle impact of the natural gas used in the energy sector in Romania. Journal of Cleaner Production 15, 1451–1462. Flanigan, O., 1972. Constrained derivatives in natural gas pipeline system optimization. Journal of Petroleum Technology 24, 549–556. Ghazikhani, M., Passandideh-Fard, M., Mousavi, M., 2011. Two new high performance cycles for gas turbine with air bottoming. Energy 36, 294–304. Graham, G.E., Maxwell, D.A., Vallone, A., 1971. How to optimize gas pipeline networks. Pipeline Industry, 41–43. Grigoriev, S.A., Porembsky, V.I., Fateev, V.N., 2006. Pure hydrogen production by PEM electrolysis for hydrogen energy. International Journal of Hydrogen Energy. 31 (2), 171–175. Iran Hydrocarbon Balance, 2009. Institute for International Energy Studies (IIES), Tehran. Jomard, F., Feraud, J.P., Caire, J.P., 2008. Numerical modeling for preliminary design of the hydrogen production electrolyzer in the Westinghouse hybrid cycle. International Journal of Hydrogen Energy 33, 1142–1152. Kabirian, A., Hemmati, M.R., 2007. A strategic planning model for natural gas transmission networks. Energy Policy 35, 5656–5670. Kovalev, L.K., Ilushin, K.V., Kovalev, K.L., Penkin, V.T., Poltavets, V.N., Koneev, S.M.A., Akimov, I.I., Gawalek, W., Oswald, B., Krabbes, G., 2003. High output power electric motors with bulk HTS elements. Physica C 386, 419–423. Larson, R.E., Wong, P.J., 1968. Optimization of natural gas system via dynamic programming. Industrial and Engineering Chemistry AC 12, 475–481. Liao, Z., Wang, J., Yang, Y., Rong, G., 2010. Integrating purifiers in refinery hydrogen networks: a retrofit case study. Journal of Cleaner Production 18, 233–241. Mah, R.S.H., Schacham, M., 1978. Pipeline network design and synthesis. Advances in Chemical Engineering 10, 126–205. Maddaloni, J.D., Rowe, A.M., 2007. Natural gas exergy recovery powering distributed hydrogen production. Hydrogen Energy 32, 557–566. Marshall, A., Børresen, B., Hagen, G., Tsypkin, M., Tunold, R., 2007. Hydrogenproduction by advanced proton exchange membrane (PEM) water electrolysers— reduced energy consumption by improved electrocatalysis. Energy 32, 431–436. Martch, H.B., McCall, N.J., 1972. Optimization of the Design and Operation of Natural Gas Pipeline Systems. Paper No: SPE 4006, Society of Petroleum Engineers of AIME. Miltner, A., Wukovits, W., Proll, T., Friedl, A., 2010. Renewable hydrogen production: a technical evaluation based on process simulation. Journal of Cleaner Production 18, S51–S62. Najibi, H., Rezaei, R., Javanmardi, J., Nasrifar, Kh., Moshfeghian, M., 2009. Economic evaluation of natural gas transportation from Iran's South-Pars gas field to market. Applied Thermal Engineering 29, 2009–2015. O'neill, R.P., Williard, M., Wilkins, B., Pike, R., 1979. A mathematical programming model for allocation of natural gas. Operations Research 27, 857–873. Ouwerkerk, H., de Lange, H.C., 2006. Technical and economical feasibility of the Rankine compression gas turbine (RCG). Applied Thermal Engineering 26, 413–420. Papadopoulos, A.I., Stijepovic, M., Linke, P., 2010. On the systematic design and selection of optimal working fluids for Organic Rankine Cycles. Applied Thermal Engineering 30, 760–769. Polyzakis, A.L., Koroneos, C., Xydis, G., 2008. Optimum gas turbine cycle for combined cycle power plant. Renewable Energy Conversion and Management 49, 551–563. Pozivil, J., 2004. Use of expansion turbine in natural gas pressure reduction stations. Acta Montanistica Slovaco 9, 258–260. Qadrdan, M., Saboohi, Y., Shayegan, J., 2008. A model for investigation of optimal hydrogen pathway, and evaluation of environmental impacts of hydrogen supply system. International Journal of Hydrogen Energy 33, 7314–7325. Quoilin, S., Lemort, V., Lebrun, J., 2010. Experimental study and modeling of an Organic Rankine Cycle using scroll expander. Applied Energy 87, 1260–1268. Rios-Mercado, R.Z., Kim, S., Boyd, E.A., 2006. Efficient operation of natural gas transmission systems: a network-based heuristic for cyclic structures. Computers and Operations Research 33, 2323–2351. Saboohi, Y., 2002. ESM, Technical Report, Seri. Shashi Menon, E., 2005. Gas Pipeline Hydraulics Handbook. Taylor & Francis Group (CRC Press), New York. Shengjun, Z., Huaixin, W., Tao, G., 2011. Performance comparison and parametric optimization of subcritical Organic Rankine Cycle (ORC) and transcritical power cycle system for low-temperature geothermal power generation. Applied Energy 88, 2740–2754. Wang, E.H., Zhang, H.G., Fan, B.Y., Ouyang, M.G., Zhao, Y., Mu, Q.H., 2011. Study of working fluid selection of organic Rankine cycle (ORC) for engine waste heat recovery. Energy 36, 3406–3418. Wei, D., Lu, X., Lu, Z., Gu, J., 2007. Performance analysis and optimization of organic Rankine cycle (ORC) for waste heat recovery. Energy Conversion and Management 48, 1113–1119.

S. Safarian et al. / Energy Policy 61 (2013) 65–77

Woldeyohannes, A.D., Abd Majid, M.A., 2011. Simulation model for natural gas transmission pipeline network system. Simulation Modelling Practice and Theory 19, 196–212. Wrana, N., Sparling, R., Cicek, N.B., Levin, D., 2010. Hydrogen gas production in a microbial electrolysis cell by electrohydrogenesis. Journal of Cleaner Production. 18, S105–S111.

77

Wu, S., Rios-Mercado, R.Z., Boyd, E.A., Scott, L.R., 2000. Model relaxations for the fuel cost minimization of steady-state gas pipeline networks. Mathematical and Computer Modeling 31, 197–220. Zeng, K., Zhang, D., 2010. Recent progress in alkaline water electrolysis for hydrogen production and applications. Progress in Energy and Combustion Science 36, 307–326.