System-level power-to-gas energy storage for highpenetrations of variable renewables

B. Lyseng a,*, T. Niet a,b, J. English a, V. Keller a, K. Palmer-Wilson a,B. Robertson a, A. Rowe a, P. Wild a

a Institute for Integrated Energy Systems, University of Victoria, PO Box 1700 STN CSC, Victoria, BC V8W 2Y2,Canadab School of Energy, British Columbia Institute of Technology, 3700 Willingdon Avenue, Burnaby, British ColumbiaV5G 3H2, Canada

a r t i c l e i n f o

Article history:

Received 2 June 2017

Received in revised form

22 November 2017

Accepted 29 November 2017

Available online 26 December 2017

Keywords:

Power-to-gas

Energy storage

Electrolyzer

Curtailment

Power system

Renewable energy

a b s t r a c t

According to outlooks by the IEA and the U.S. EIA, renewables will become the largest

source of electricity by 2050 if global temperature rise is to be limited to 2 !C. However, at

penetrations greater than 30%, curtailment of wind and solar can be significant in even the

most flexible systems. Energy storage can reduce curtailment and increase utilisation of

variable renewables. Power-to-gas is a form of long-term storage based on electrolytic

production of hydrogen. This research models the co-sizing of wind and solar PV capacity

and electrolyser capacity in a jurisdiction targeting 80% penetration of variable renewable

electricity. Results indicate that power-to-gas can reduce required wind and solar capacity

by as much as 23% and curtailment by as much as 87%. While the majority of charging

events last less than 12 h, the majority of the total annual stored energy comes from

longer-term events. Additional scenarios reveal that geographic diversity of wind farms

reduces capacity requirements, but the same benefit is not found for distributing solar PV.

© 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Introduction

Limiting global temperature rise to the 2 !C target requires CO2

emissions from global electricity generation to approach zeroby 2050 according to the IPCC and others [1,2]. However, with67% of global electricity generation currently sourced fromfossil fuels [3], meeting emission targets will require anextensive build-out of low-carbon generation. The Interna-tional Energy Agency's (IEA) 450 Scenario, wherein globalwarming has a 50% chance of being limited to 2 !C, projectsrenewables will contribute nearly 60% of global power

generation in 2040 [4]. Even in the IEA and U.S. Energy Infor-mation Administration reference scenarios, where the 2 !Ctarget is not met, renewables are projected to experience thegreatest growth of all generation options for the coming de-cades, surpassing coal as the largest source of electricityglobally by 2040 [5,6].

Variable renewable electricity (VRE) technologies, namelywind and solar, will likely account for themajority of new low-carbon generation in many jurisdictions because of con-straints on other options. The future of nuclear power is un-

certain amid concerns over reactor accidents, waste disposal,and nuclear proliferation [7]. Coal plants outfitted with carbon

* Corresponding author.E-mail address: bcl@uvic.ca (B. Lyseng).

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier .com/locate/he

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9

https://doi.org/10.1016/j.ijhydene.2017.11.1620360-3199/© 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

capture and sequestration (CCS) technology is a low-carbonbaseload alternative, but according to the IEA “… progress isfar too slow to achieve the widespread commercial deploy-ment envisioned …” [8]. Hydropower, currently the largestsource of renewable electricity, has limited additional poten-tial, particularly in OECD countries [6,9]. Studies exploring therole of bioenergy in future, low-carbon energy systems find itscontribution varies greatly depending on the region,

competing demands, and, as summarised by Rogner et al., itsultimate use will be “… less a question of the available theo-retical potential than of ecological sustainability and socio-economic desirability” [9e11]. The IEA projects that even withbreakthroughs in enhanced geothermal systems, geothermalelectricity will not exceed 3.5% of global power generation in2050 [12]. These factors, taken together with projections fordecreasing costs of wind and solar generation, lead to theconclusion that VRE technologies will contribute significantlyin the future of many jurisdictions.

Emerging policies at various levels of government and

recent studies of Integrated Assessment Models (IAMs)support the expected growth of VRE technologies. Germanyis planning to increase the share of electricity from renew-ables to at least 80% of gross consumption by 2050 [13], andDenmark aims to be 100% renewable in all sectors by 2050[14]. IAMs such as GCAM [15,16], and IMAGE [17,18], includerepresentations of energy systems, economic structures,and climate systems to study global climate change path-ways and policies. Luderer et al. explore the role of renew-able energy in limiting global CO2 concentration to450e550 ppm by comparing results from a range of IAMs

[11]. Even when CCS and nuclear are permitted, most IAMsindicate renewables will generate 30e70% of global annualelectricity by 2050, largely from wind and solar. Subsequentresearch improving the representation (i.e. parameterisationof characteristics) of VRE in IAMs found higher wind and

solar penetrations than previous versions of all considered

models [19].Integrating these high levels of VRE will be challenging

because of the limited ability of existing electrical systems torespond to variation and uncertainty in net load. At timescalesless than 5 min, wind and solar power can decrease systeminertia and increase the need for power quality and regulationservices [20,21]. At longer timescales, system operators can beforced to curtail significant amounts of renewable energygeneration, even at current penetrations of VRE [22].Depending on contractual arrangements, curtailment canincur costs to the system such as constraint payments, or

costs to curtailed generators such as lost energy payments.Expansion of the transmission system to deliver renewablegeneration to load centers is a common solution to thisproblem. However, transmission expansion only reducescurtailment due to transmission constraints, not curtailmentdue to lack of demand.

Curtailmentdue to lackofdemande the focusof this studyeoccurswhenVREplus inflexible baseload generation exceed thesystem demand. Denholm and Hand examine the curtailmentthat occurs at high levels of VRE penetration under varioussystemflexibilities [23]. Definingflexibility factor as the “fractionbelow annual peak [demand] to which conventional generatorscan cycle”, they find that “… achieving 80% of the simulatedsystem's electricity from wind generation only (and withoutstorage) requires a systemflexibility of close to 100%, and resultsinacurtailment rateofmore than43%.” Inthiscase, theeffectivecapacity factor of the last unit of wind installed would be 6%,making its marginal cost over five times the cost when there isno curtailment. In Europe, under the IEA's 450 Scenario, curtail-ment of excess electricity couldoccur up toone-third of the timein 2040 if integration methods such as storage are not imple-mented [4].

Energy storage can facilitate VRE integration and reducecurtailment [24e28]. Candidate technologies capable ofserving long-term energy storage are pumped hydro storage(PHS), compressed air energy storage (CAES), and power-to-gas (PtG) [28e30]. Both PHS and CAES require specific geolog-ical conditions and, therefore, lack flexibility of location. PtG,whereby hydrogen is produced by electrolysis, is less site-specific. Multiple opportunities exist for PtG to leverage thenatural gas grid and integrate with other sectors, as depictedin Fig. 1. The vast majority of hydrogen - a versatile energycarrier - is currently produced from fossil fuels, but PtG offersa means of production with lower emissions. Furthermore,

PtG can be an essential part of a hydrogen economy, which isconsidered by some to be the ultimate future of energy sys-tems [31e33].

Large-scale PtG in high VRE power systems is an activefield of research with many areas yet to be thoroughlyexplored. This study focusses on PtG storage, assuming thenatural gas grid is an unlimited storage reservoir, in anislanded power system targeting 80% VRE penetration. Themodelled system, depicted in Fig. 2, contains wind, solar PV,and combined cycle gas turbines (CCGT) as generators, PtGas the storage technology, and time-varying load in an

hourly model of one year. Electrolyser capacity, chargingbehaviour, curtailment, and hydrogen production are ana-lysed for a range of system configurations that achieve the

Nomenclature

CAES compressed air energy storageCCS carbon capture and sequestrationCCGT combined cycle gas turbine;Ed annual VRE generation used directly to meet

demand [MWh]Eload annual electricity demand [MWh]Est annual energy from storage that serves

demand [MWh]Fvre annual fraction of VRE penetrationLDC load duration curveNvre normalised VRE generation profile;

IAMs integrated assessment modelsPHS pumped hydro storagePtG power-to-gasPd hourly VRE generation used directly [MWh]Pst hourly energy from storage serving load [MWh]Pvre hourly VRE generation potential [MWh]RLDC residual load duration curveVRE variable renewable electricity

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1967

targeted VRE penetration. Sensitivities to geographic di-versity of VRE resources, wind-solar capacity split, andquality of wind year are also explored.

There are many pathways other than storage in Fig. 1 for

PtG hydrogen that are not explored in this study. While thisresearch assumes the natural gas grid to be an infinite storagereservoir, it does not model the injection or interaction withthe natural gas grid. Hydrogen can be further processed tocreate methane before injection into the natural gas grid[34e37]. Alternatively, hydrogen can be used for industrialpurposes, displacing fossil fuel-sourced hydrogen andreducing associated emissions such as carbon dioxide [38e40].The global hydrogen market is expected to be valued atseveral trillion dollars by 2020 [41].

Literature review

Existing researchonPtGcanbedivided into fourbroadcategories:small systems (i.e. <1000 MW), nuclear-based systems, naturalgas-paired systems, and high VRE systems. Depending on thetype of system, the role of PtG can vary from clean hydrogenproduction to storage. This role, aswell as the size of the system,impacts the sizing of PtG and the way in which it is operated.

A recent review of PtG projects reveals that 76% of existinginstallations are stand-alone systems that are not connectedto a bulk power grid [42]. Many of these are pilot projectsinstalled on isolated systems to increase VRE penetration andto reduce reliance on fossil fuels. Some studies assess op-portunities for PtG on these small, weak, or isolated systemswith high wind penetration [43e46]. PtG is shown to signifi-cantly increase wind penetration [46], and provide a cost-competitive storage option for achieving high penetrationsof VRE [45]. Nonetheless, PtG systems would benefit fromincreased electrolyser flexibility and efficiency [43], and

hydrogen production costs are highly sensitive to electrolysersizing [44]. In these studies, VRE and storage are sized forautonomy, system stability, or for reasons not provided. Thesesmall-scale systems have unique generation and demandprofiles that result in storage operation profiles that differfrom the profiles of large-scale systems.

At a larger scale, several studies have explored electrolysisapplications in nuclear-based power systems [47e51,41,52].The focus of these studies is either generation of low-costhydrogen or operation of the electrolyser as a dispatchableload so that the reactors can operate at constant output. A

common finding in these studies is that hydrogen productioncosts are highly sensitive to electrolyser utilisation. Low uti-lisation factorsand/oroperationathighelectricityprices resultin higher hydrogen production costs. Although the focus ofthese studies is not storage of excess VRE, results highlight theimportance of electrolyser sizing and operation in PtG.

Other research has modelled the impacts of PtG on thepower and natural gas sectors [27,53,54]. Qadrdan et al. focuson impacts of PtG in the natural gas system with a high res-olution model of Great Britain with 30% wind penetration.Electrolyser capacity is not sized; rather its operation is con-

strained by maximum allowable hydrogen concentrations inthe gas grid. The analysis, limited to two days representinghigh and low demand, finds that PtG decreases the operatingcosts and emissions of the combined natural gas and powersystem. Vandewalle et al. model the power, natural gas, andCO2 sectors in Belgium, with VRE capacity scaled to generate100% of annual electricity demand, assuming no curtailment.The study's authors find that PtG transfers capacity and flex-ibility issues from the power system to the natural gas system.Electrolyser capacity is determined based on investment costsand full load hours to produce synthetic methane at a cost

competitive with natural gas. Results presented on electro-lyser operation are limited to the annual electricity usage.

Previous studies have investigated PtG in electricity sys-tems with high penetrations of VRE [55e60,35]. Most of thesestudies use cost optimisation models to determine the gener-ationmix, size of storage (MWh), and/or hydrogen production.In a study of a 100% VRE European power grid with multiple

TRANSPORTATION

Electrolyser

Combined Cycle Natural Gas Plant

Local storage

H2 Gas

Electrical Grid

Natural Gas Grid

INDUSTRY

Fuel Cell Vehicles

Natural Gas Vehicles

Electric Vehicles

Fuel cell

Wind &Solar PV

Methanation

CH4

Fig. 1 e Schematic of power-to-gas showing the potentialfor integration of energy carriers and sectors.

Electrolyser

Electrical Grid

Natural Gas

Wind & Solar PVGeneration

PtG Storage

Reservoir

CCGT

Fig. 2 e Schematic representation of the energy system asmodelled.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91968

storage options, PtG accounts for 64% of the installed storagecharging power (MW) and 99% of the storage capacity (MWh)[55]. Another study targeting 100% renewable electricity for asmall region in Germany finds that if the capital cost of PtG isless than 2453 V/kW, it is more cost-effective than Li-ion bat-teries at a cost of 350 V/kW [56]. Some of the aforementioned

studies of a PtG system coupled to an electrical system withhigh penetrations of VRE assess several scenarios withdifferent system mixes; however the relationship betweenelectrolyser capacity and VRE capacity is not examined.Furthermore, the temporal characteristics of PtG operation arenot assessedand,with the exceptionofHeide et al. [43], neitheris the impact of VRE resource mix on system capacities.

In this study we address these gaps by assessing the VRE-electrolyser capacity trade-off and the associated curtail-ment. In addition, we characterise the hourly distribution ofcharging events and discuss its implications for electrolyser

operation. We explore impacts of resource diversity in thisstudy by defining six scenarios representing different wind-solar mixes, locations, and wind years. In addition, a briefassessment of hydrogen concentrations in the natural gas gridis presented for selected systems.

To perform this analysis, we develop a method that jointlydetermines combinations of VRE and electrolyser capacity toachieve 80% VRE penetration. This is done by simulating thehourly operation of hundreds of capacity combinations andselecting ones that have 80% penetration.

Methods

The system investigated in this research represents the elec-tric power system in the Canadian province of Alberta with

80% VRE energy penetration in the year 2050. According to theGovernment of Alberta, “Alberta has one of the most exten-sive natural gas systems in the world…” and exports over 50%of production [61]. This suggests PtG could be broadlydeployed and that the hydrogen produced could be accom-modated by the natural gas infrastructure.

This study uses the simplified representation of the Albertaelectricity system and PtG storage shown in Fig. 2. A time-varying electricity demand is met by wind, solar PV, and/orCCGT generators and delivered by the electrical grid. The PtGstorage system has three main components: charger, reser-voir, and discharger. In our model, the charger is an electro-lyser that operates on excess wind and solar generation toproduce hydrogen. The reservoir is a proxy for the naturalgrid, which we assume to be extensive and thus model itsimply as an unlimited reservoir for hydrogen. The reservoirstorage level increases when hydrogen is produced by the

charger (electrolysis), and decreases when the discharger(CCGT) burns hydrogen from the reservoir. The CCGT can alsoburn natural gas if there is no hydrogen in the storage reser-voir. There are no transmission constraints in themodel of thepower system and, therefore, the effects of siting electrolyserand generation technologies are not assessed.1

Normalised VRE Profile (NVRE)

Scale VRE Genera!on

(PVRE)

VRE genera!on > Load?

No

Direct VRE = VRE genera!on

Direct VRE(Pd)

Residual load > Electrolyser capacity?

Yes

Direct VRE = Load Direct VRE(Pd)

NoPower to storage= Residual load *

electrolyser efficiency

Power from storage= Discharge energy * Discharge efficiency

Yes Power to storage= Electrolyser capacity * electrolyser efficiency

Non-VRE genera!on Non-VRE power

STORAGE

Stored energy available?

Yes Stored VRE(Pst)

No

Ed = ∑Pd

Est = ∑Pst

Fig. 3 e Flow chart illustrating the hourly simulation model.

1 Siting of wind and solar generation is examined based onregional resource profiles, but not with regards to energy systeminfrastructure such as transmission lines.

2 The higher heating value (HHV) is consistent with theassumed efficiency when liquid water is used by the electrolyserand therefore includes the heat of vapourisation.

3 Base map “Canada Alberta relief location map” by Carport isused under CC BY-SA 3.0.

4 https://solaralberta.ca/.5 http://pvwatts.nrel.gov/.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1969

Assuming an infinitely large natural gas grid, the model

does not constrain the rate of hydrogen production (i.e. in-jection rate into the grid), or the amount of hydrogen in thereservoir (i.e. concentration of hydrogen in the grid). None-theless, hydrogen can adversely affect natural gas infra-structure and certain customer end-uses. Therefore we assessthe infinite reservoir assumption by determining averagehydrogen concentrations that result from selected systems(see Section Hydrogen and the natural gas grid).

In order to provide 80% of the annual energy by VRE e

directly and via storage e there must be sufficient VRE ca-pacity and electrolyser capacity. A wide range of combina-

tions of VRE capacity and electrolyser capacity are simulatedto create a contour plot of the VRE penetrations (Fig. 7).Selected VRE and electrolyser capacity combinations thatachieve 80% penetration are then analysed in further detail.

Further explanation of the model is presented in SectionSystem model. Wind, solar, and load data are provided inSection Wind, solar, and load data. Model scenarios aredefined in Section Scenarios, with additional modelling pro-cess and details provided in Section Combinations andconfigurations.

System model

An hourly dispatch model was developed to simulate opera-tion of the system. For each hour, energy flows aremanaged inaccordance with the flow chart shown Fig. 3. Themodel treatswind and solar generation as “must-take” up to the demand inthat hour. Therefore, when VRE generation exceeds demand,the energy in storage increases equal to the higher heatingvalue of the hydrogen generated by an 80% efficient electro-

lyser [36,62,63].2 If VRE generation is not sufficient to serve theload, the model uses energy from storage before using non-VRE generation. Discharge from storage is assumed to occurat 55% efficiency, representing a CCGT [64]. When VRE gen-eration exceeds demand and electrolyser capacity, VRE

generation is curtailed. Fig. 4 illustrates the operation oftechnologies and energy allocation in the model.

To calculate the fraction of annual VRE penetration weinclude VRE generated energy used directly and energy outputfrom storage. The VRE energy used directly, i.e. instanta-neously, to meet hourly demand is denoted as Pd. Hourly en-ergy output from storage, Pst, is also considered VRE-sourcedbecause storage was charged with excess VRE generation.

The fraction of annual VRE penetration, Fvre, is defined inEquation (1), where Eload is the annual energy demand, and Edand Est are the annual sums of Pd and Pst, respectively.

Fvre ¼Ed þ Est

Eload(1)

It is assumed that the power system has a fleet of CCGTplants with a total capacity equal to the peak demand that canserve as both storage discharge and flexible, non-VRE generation.In reality, CCGT plants would burn a mixture of electrolysedhydrogen and natural gas as drawn from a pipeline. The CCGTplants are assumed to have no ramping limits or minimumgeneration level. This duty would more likely be served by acombination of CCGT and open cycle gas turbines in an actualsystem.

It is also assumed that the electrolysers are perfectly flex-ible, able to ramp up and down quickly through their full ca-pacity range with constant efficiency. While electrolysers dohave minimum load constraints and efficiencies are not con-stant, hundreds or possibly thousands of electrolysing unitswould be required for the charging capacities shown here. It is

0

10

20

30

40

0 24 48 72 96

]W

G[ rewo P

Hour

CurtailedElectrolyserNon-VRE generationFrom storageVRE directVRE generationLoad

Fig. 4 e Depiction of how energy is allocated in modelsimulations. Excess generation by VRE up to the maximumelectrolyser capacity can be used to create hydrogen forstorage. The model will dispatch from storage beforerelying on non-VRE generation from natural gas-firedcombined cycle turbines (CCGT).

Wind farmSolar installa!on

ALBERTA

BRITISH COLUMBIA

SASK

ATCH

EWAN

600 km

* Vancouver

Edmonton

Calgary

Medicine Hat

Lethbridge

SESC

SW

N

Fig. 5 e Approximate locations of representative wind andsolar data.3 Abbreviations correspond to the wind regions,see Table 1.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91970

assumed that units are dispatched sequentially and rampquickly so that each unit can be assumed always to be either

off or on at full capacity.

Wind, solar, and load data

Alberta has an area of 662,000 km2, approximately twice thesize of Germany. There are four wind regimes in the province,three in the south and one covering the remainder of theprovince [65]. The diversity in the south results from the

mixed geography of the Rocky Mountains in the west to theprairies in the east. Themajority of wind development to datehas occurred in the southwest region of the province, but anincreasing number of farms are currently in various stages ofdevelopment in other regions.

The hourly wind generation profiles used in the model areconstructed fromhistoric hourly data from one representative

wind farm in each of the four regions, as shown in Fig. 5. Thesewind farms are selected for low cross-correlation of poweroutput with the wind farms in the other regions and highannual capacity factor. Low cross-correlation due togeographic dispersion of generation sites has been shown tobe beneficial for integration of VRE [66,67]. The selected farmshave annual capacity factors of 34e37%, representing thetrend to larger, more productive turbines that will likely beinstalled in the future [68]. Data from the year 2013 is selectedto represent an average wind year based on annual capacityfactors from 2003 to 2015.

As of 2016, Alberta has 10.5 MW of installed solar powercapacity, mainly in the form of rooftop photovoltaic (PV), withthe largest installation being a 2 MW solar farm.4 For thisstudy, PV Watts5 is used to simulate hourly solar PV genera-tion at four locations in Alberta, as shown in Table 2 and Fig. 5.Additional details of the solar installations such as tilt angleand efficiency can be found in the Supplementary Material.

Hourly total system load data from the Alberta ElectricSystem Operator (AESO) for 2013 is used to define the loadprofile. This load profile is scaled to the projected load for 2050,extrapolated from AESO's 2016 Long-term Outlook [69]. The

resulting annual electricity demand for this study is 134 TWh,with an average load of 15.3 GW, and a peak load of 19.3 GW.

Sola

rW

ind

VRE

Normalised VRE profile

Aggregate tech. profileNormalised genera!on profiles Generator

weightsTechnology

weightsAggregate VRE

profile

SW Region fW_SW

fW_SC

fW_SE

fW_N

fW

fS_LB

fS_MH

fS_CG

fS_ED

fS

NVRE

SC Region

SE Region

N Region

Lethbridge

Medicine Hat

Calgary

Edmonton

Fig. 6 e The normalised VRE profile, which defines each scenario, is based on the VRE generation mix.

0.10.1

0.20.2

0.30.3

0.40.4

0.50.5

0.60.6

0.70.7

0.8

1

2

3

0.8

4

0.9

0.9

1

0 1 2 3 4 5 6Normalised VRE Capacity

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Nor

mal

ised

Ele

ctro

lyse

r Cap

acity

VRE fractionSelected configurations

Fig. 7 e Annual energy fractions met by VRE andelectrolyser combinations for the Reference Scenario. Thefour points define different configurations that achieve 80%VRE penetration. Axes are normalised by average systemload (MW).

Table 1 e Representative wind farms for the four regions.Historic hourly generation bywind farm is available fromthe Alberta Electric System Operator (AESO). Listedcapacity factors are for 2013.

Wind region Wind farm Capacity Capacity factor

[MW] [%]

South west (SW) Soderglen 68 36.9South central (SC) Magrath 30 34.8South east (SE) Chin Chute 30 34.4North (N) Wintering Hills 88 36.8

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1971

Scenarios

A scenario is defined by its normalised VRE generation profile. The

normalised VRE generation profile depends on the mix and lo-cations of wind and solar farms, the wind-solar capacity split,and the wind year. Wind and solar generation profiles are nor-malised to their nominal rated capacity before an aggregateprofile is constructed by the method illustrated in Fig. 6. Thevalues defining the scenarios, presented in Table 3, are selectedto represent a plausible range of resource diversities. The sumofthe wind capacity fractions for each of the four wind regions,fW_X, is unity. Similarly, the sumof thecapacity fractions for eachsolar location, fS_X, is unity. The sum of the wind fraction of VREcapacity, fW, and the solar fraction of VRE capacity, fS, is unity.

The Reference Scenario assumes equal wind and solar capac-ities,windcapacity spread equally among the four regions, 2013wind generation data, and solar capacity equally split betweentwo locations in the south of the province. Equally dividing thewind capacity among the four regions results in an aggregatewind generation profile with diversity statistics comparable tothat of Nordic countries [66] (see Supplemental Material).

Three scenarios explore the impact of the wind resource.The Concentrated Wind scenario represents the entire windcapacity being built in the region with the highest capacityfactor in 2013 and uses 2013 wind generation data. 2014 Wind

and 2011 Wind scenarios, which are based on data from thenamesake years, represent a relatively low and a relativelyhigh wind year, respectively. The average capacity factor ofthe four representative wind farms for these years are 31.2%and 41.3%, compared to 35.7% in 2013.

The solar generation profile in most scenarios assumessingle-axis tracking installations with capacity divided evenlybetween Lethbridge and Medicine Hat. These cities are in thesouth of the provincewhere annual solar insolation is highest.The High Solar Fraction scenario assumes that 65% of the VREcapacity is solar and 35% is wind, resulting in more annual

electricity generation from solar PV than wind. Sensitivity to a

large-scale build-out of rooftop PV in Edmonton and Calgary,

the province's two largest cities, is explored in the Rooftop Solarscenario. In this scenario, the capacity is spread evenly amongthe four solar locations.

Combinations and configurations

For each of the six scenarios, 400 combinations of VRE capacityand electrolyser capacity are simulated for one year. Thesecombinations consist of 20 capacities of VRE, ranging from

zero to six times the average load. For each of these VRE ca-pacities, the electrolyser capacity is varied from zero to half ofthe average load in 20 increments. For each combination, thefraction of annual energy that can be provided by VRE isdetermined. This method is used to generate contour plots ofsystem VRE fractions (such as Fig. 7) that are then used toselect configurations for detailed analysis.

A configuration specifies a selected VRE capacity and elec-trolyser capacity combination that achieves exactly 80% VREpenetration. Four configurations are examined in the Refer-ence Scenario and one in each of the other five scenarios.

In the Reference Scenario, the four configurations aredetermined with electrolyser capacities that are 1%, 10%, 20%,and 50% of the average system demand (MW). The corre-sponding VRE capacity for each configuration is that whichresults in 80% VRE penetration. These configurations are usedto examine capacity mix, electricity mix, residual load dura-tion curves, hydrogen production, and electrolyser utilisation.

For the other five scenarios, the electrolyser capacity foreach configuration is set to 20% of the average systemdemand(3.06 GW), as in Configuration 3 of the Reference Scenario. Inthe Concentrated Wind, Rooftop Solar, and High Solar Frac-

tion scenarios, the VRE capacities necessary to reach 80% VREpenetration are determined for each configuration. 2014Windand 2011 Wind scenarios are meant to assess the ability of asystem tomeet the 80%VRE penetration target under differentwind years. Therefore, in addition to electrolyser capacity,these scenarios also have the same VRE capacity as Configu-ration 3 of the Reference Scenario.

Results and discussion

Reference Scenario results

For the Reference Scenario, combinations of VRE capacity andelectrolyser capacity that enable equivalent penetrations of

Table 3 e Parameters defining VRE profile for each scenario. Wind regions and solar locations are described in SectionWind, solar, and load data.

Scenario Windyear

Wind fraction ofVRE capacity

Fraction of wind capacityby region (fW_X)

Solar fraction ofVRE capacity

Fraction of solar capacityby location (fS_X)

(fW) SW SC SE N (fS) LB MH CG ED

Reference Avg. 0.5 0.25 0.25 0.25 0.25 0.5 0.5 0.5 0 0Concentrated Wind Avg. 0.5 1 0 0 0 0.5 0.5 0.5 0 0Rooftop Solar Avg. 0.5 0.25 0.25 0.25 0.25 0.5 0.25 0.25 0.25 0.25High Solar Fraction Avg. 0.35 0.25 0.25 0.25 0.25 0.65 0.5 0.5 0 02014 Wind Low 0.5 0.25 0.25 0.25 0.25 0.5 0.5 0.5 0 02011 Wind High 0.5 0.25 0.25 0.25 0.25 0.5 0.5 0.5 0 0

Table 2e Location and performance of solar installations.

Solar installationdata location

Type Annualgeneration

Capacityfactor

[kWh/kW] [%]

Lethbridge (LB) 1-axis tracking 2565 29%Medicine Hat (MH) 1-axis tracking 2527 29%Calgary (CG) Fixed rooftop 1620 18%Edmonton (ED) Fixed rooftop 1468 17%

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91972

VRE are plotted as contours in Fig. 7. For example, the “0.8” linerepresents the combinations of VRE and electrolyser capac-ities that achieve 80% VRE penetration. With PtG as the onlymodelled storage option, there is a trade-off between VREcapacity and electrolyser capacity at high penetrations;increasing electrolyser capacity decreases required wind andsolar capacity, and vice versa. Storage has negligible impactwhen VRE capacity is less than twice the average load. Thiscan be attributed to the diverse mix of VRE and the fullyflexible non-VRE generation. Four system configurations,indicated by the numbered points on the 80% curve in Fig. 7,are selected for more detailed analysis below.

Configuration 1 has little electrolyser capacity, relyingalmost solely on VRE capacity. Configurations 2, 3 and 4demonstrate that VRE capacity can be reduced with storage,although at a diminishing rate. Configuration 3 has 16% lessVRE capacity than Configuration 1, while Configuration 4 has23% less. The 80% contour line is nearly vertical at Configu-ration 4 indicating that additional electrolyser capacity doesnot significantly decrease VRE capacity.

Combined VRE and electrolyser capacity decreases fromConfiguration 1 to 2, and again from 2 to 3, but is equal in 3 and4 (see Fig. 8(a)). Fig. 8(b) shows how increasing electrolyser

capacity increases the amount of energy fromPtG storage, anddecreases the amount of curtailed and direct VRE energy. Theamount of annual electricity demandmet from storage rangesfrom 0.3% in Configuration 1, to 6.6% in Configuration 4.

Charging of storage occurs when the residual load, i.e.demand less VRE generation, is negative. Sorting hourly re-sidual loads in descending order creates a residual loadduration curve (RLDC), which is helpful in understanding thebehaviour of very high VRE systems with storage. Fig. 9 pre-sents the RLDCs for the four 80%-configurations. When aRLDC is positive, power from dispatchable generation is used

to satisfy demand. This dispatchable generation is fromCCGTs and includes both storage discharge and non-VREgeneration. Negative residual load charges storage up to theinstalled charging capacity, beyond which the energy iscurtailed.

These RLDCs illustrate how storage reduces curtailment,albeit at a diminishing rate with increasing electrolyser ca-pacity. The “tails” (highly negative values) of the RLDCs aresharp so the incremental energy available for storage de-creases as electrolyser capacity increases. This explains why

increasing electrolyser capacity does not proportionally in-

crease energy from storage, as seen in Fig. 8. On the otherhand, increasing electrolyser capacity can greatly reduce theamount of curtailed energy in the tail, a trend also seen inFig. 8. The energy available for storage is significant; howevercharge and discharge efficiencies reduce the amount that canbe delivered back to the power system.

Selection of a preferred PtG system configuration dependson additional factors outside the scope of this study. Theseinclude, but are not limited to, relative costs of VRE and stor-age, resource availability, and additional benefits of storagesuch as balancing services. These factors can be unique to

each system and should be assessed accordingly.

Electrolyser operation

In this section, electrolysis, i.e. charging, events are examinedand implications for the storage technology are discussed. Allof the 80%-configurations in the Reference Scenario experi-ence a similar number of charging events, ranging from 435 inConfiguration 1 to 477 in Configuration 4. This is expected

because all four configurations use the same normalised VREprofile, scaled to different capacities.

Charging events are organised by the number of consecu-tive hours electrolysis occurs. Themajority of charging eventsare less than 12 h in duration, as seen by the histogram forConfiguration 3 shown in Fig. 10 (a). This can be largelyattributed to the nature of the load, wind, and sun in thesummer: system load is lower, winds tend to be lighter, andsolar irradiation is higher. Therefore the VRE generation pro-file is often dominated by solar generation which, on manydays, can exceed demand for a number of the daylight hours.

When these events are organised by the cumulativeamount of hydrogen that is generated, the contribution oflonger duration events is more apparent (Fig. 10 (b)). For theaverage wind year in Configuration 3, only 26% of the eventshave durations of 12 h or more, but these events account for54% of the hydrogen produced. The other wind years shown inFig. 10 (b) support the same trend. This is why long-termstorage technologies like PtG are particularly valuable invery high VRE systems.

Available storage technologies cover a broad range of en-ergy and power specifications and can be roughly divided into

three categories based on energy-to-power ratios [28,70].

58.252.7 49.2 44.6

0.151.53

3.06 7.66

0

20

40

60

1 2 3 4

Capa

city

[GW

]

Configuration

VRE capacity Electrolyser capacity(a)

107 104 102 99

0.4 3.1 5.4 8.9

57.338.2 25.3 7.7

0

60

120

180

1 2 3 4

Ener

gy [T

Wh]

Configuration

VRE direct From storage Curtailed(b)

Fig. 8 e (a) VRE and electrolyser capacity for four configurations in the Reference Scenario. (b) Energy distribution for the fourconfigurations.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1973

Stating energy in units of kWh and power in units of kW, theenergy-to-power ratio is the number of hours a technologycan charge or discharge at its nominal capacity. Short-termstorage, such as capacitors and flywheels, has an energy-to-power ratio of less than 1 h and is used mainly to maintainpower quality for seconds or minutes. Medium-term storage,which includes most batteries, has an energy-to-power ratiobetween 1 and 10 h and is suitable for peak shaving and timeshifting. Long-term storage, such as PtG, has energy-to-power

ratios exceeding 10 h and can be suitable for energy man-agement, seasonal storage or unit commitment operation.

The large range in duration of charging events observed inthis study may be best served by more than one storagetechnology. Existing alkaline electrolysers, the most commonand lowest cost variety, are slow to reach steady operatingtemperature [42,43], have exhibited problems when poweredby a variable source [36,42,43], and cannot turn down opera-tion below approximately 20% of their rated capacity

0 10 20 30 40 50 60Charge duration [hours]

0

10

20

30

40

50

Num

ber o

f eve

nts

(a)0 20 40 60 80 100 120

Charge duration [hours]

0

0.2

0.4

0.6

0.8

1

Frac

tion

of s

tore

d en

ergy

Low (2014)Average (2013)High (2011)

(b)

Fig. 10 e (a) Charge durations in Configuration 3 in an average wind year. (b) Cumulative stored energy by charge duration inConfiguration 3 for three wind years.

Fig. 9 e Residual load duration curves for the four VRE-electrolyser configurations. Demand met directly by VRE, Ed, isrepresented by the positive area between the LDC and the RLDC. Dispatchable generation, provided by combined cyclenatural gas plants, includes both discharge from storage and non-VRE generation.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91974

[42,51,36,71]. Nonetheless, continued development of alkaline

electrolysers may increase their flexibility to sufficientlymanage VRE generation [51,72]. Polymer electrolyte mem-brane (PEM) electrolysers exhibit much greater flexibility interms of response time and minimum load and would besuitable for the medium-term storage events [36,42,62]. PEMelectrolysers are relatively new commercially and likelyrequire cost reductions as well as efficiency and longevityimprovements to compete with battery storage technologiesfor this type of storage duty [36,62]. Storage technologies otherthan PtG for medium-term duty would likely decrease thetotal storage capacity required to meet the 80% target in this

study because of higher round-trip efficiencies. This would bean interesting topic for future research.

Hydrogen and the natural gas grid

Annual hydrogen productions for the Reference Scenarioconfigurations are shown in Table 4 and range from16e408 ktH2. Globally, over 50 million tonnes of hydrogen areproduced annually, mainly from fossil fuels [73]. Electrolysers

account for just 4% of global production, or approximately2 Mt [62]. Configuration 4 of the Reference Scenario wouldgenerate 20% of the current global production of hydrogen byelectrolysis. This represents a significant build-out of elec-trolyser capacity in a relatively small jurisdiction.

One of the principal opportunities of PtG is leveraging thenatural gas grid for storage, but there are limits to how muchhydrogen can be accommodated. Studies suggest concentra-tions between 1 and 20% by volume are possible before theblend adversely impacts end-use devices such as householdappliances, public safety, or durability of the existing natural

gas network [74e76]. Annual average hydrogen concentrationin Alberta's extensive natural gas grid from these configura-tions would be less than 5% and suggests PtG integration withthe gas grid at this scale may be feasible (see Table 4). Detailsof the analysis can be found in Supplementary Material.

The annual average concentrations presented in Table 4suggest feasibility, but there are factors that could lead tohydrogen concentrations that are significantly higher thanthese averages. Firstly, hydrogen production by the electro-lysers will vary throughout the year. Annual utilisation of theelectrolysers for the Reference Scenario configurations is be-

tween 30 and 59%. With the exception of Configuration 4,Fig. 11 shows that most of the hours of operation are at fullpower. This behaviour can also be seen by the “electrolysed”area of the RLDCs in Fig. 9. This means instantaneous

hydrogen production could be two or three times the annualaverage. Secondly, natural gas production and system flowsvary throughout the year, and times of lower flows will causehigher relative concentrations. Dedicated hydrogen storagenear electrolyser plants such as cavern storage could be used

to buffer these injection rates.Another important consideration is the injection points of

hydrogen into the natural gas grid. The listed concentrationsassume injection of hydrogen such that the concentration isuniformly mixed with natural gas. While electrolysers can bedistributed around the electrical and natural gas grids,discrete PtG infrastructure will result in non-uniformconcentration.

Resource diversity impacts

The geographic diversity of wind and solar, and the relativecapacity mix has an impact on the capacities required toachieve high VRE penetration. To compare across the otherscenarios, the electrolyser capacity is fixed at 20% of theaverage system demand (the same as Configuration 3 of theReference Scenario). For each wind and solar scenario the VREcapacity is determined as that which results in 80% VREpenetration. Fig. 12 reveals that all of these scenarios requiremore VRE capacity than Configuration 3 in the Reference

Scenario.Concentrating wind capacity in the location with the best

resource, as simulated in the Concentrated Wind scenario,requires 16% more VRE capacity compared to Configuration 3of the Reference Scenario in which wind is distributed evenlyaround the regions (Fig. 12). Diversity of locations of windfarms smoothes the aggregate generation profile, causingflatter generation duration and RLDC curves, indicating thatmore VRE is captured, both directly and via storage. Thebenefit of wind diversity to system balancing requirementshas been studied previously [66,77]. The current study shows

additional benefits of geographic distribution for systemswithvery high VRE penetrations and storage.

While geographic diversity of wind farms decreasesrequired VRE capacity, geographic diversity of solar PV doesnot have the same benefit. The Rooftop Solar scenario, inwhich 50% of the PV capacity is in fixed rooftop systems,

0

0.25

0.5

0.75

1

0 2000 4000 6000 8000

Frac

tion

of e

lect

rolys

er c

apac

ity

Hours (sorted)

Configuration 1Configuration 2Configuration 3Configuration 4

Fig. 11 e Utilisation duration curves of electrolysersconfigurations in the Reference Scenario.

Table 4 e Hydrogen generation and averageconcentration in Alberta's natural gas grid for the fourconfigurations.

Configuration Annual H2

generationAverage annual H2

concentration

[MtH2] [vol%]

1 0.016 0.17%2 0.142 1.47%3 0.247 2.55%4 0.408 4.21%

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1975

requires 7% more VRE capacity than Configuration 3 in theReference Scenario with the same electrolyser capacity. Thiscan be attributed to the relatively limited East-West distribu-tion of solar sites and to the loss of generation potential fromusing fixed systems compared to single-axis installations.

Solar generation is driven predominantly by its diurnalpattern, and requires greater distances or unique regionalphenomenon to benefit from geographical diversity at thishourly time scale.

The High Solar Fraction scenario (35% wind and 65% solar,by capacity) requires 26% greater total VRE capacity thanConfiguration 3 in the Reference Scenario to achieve 80% VREpenetration. This is due to the low capacity factor of solarrelative to wind, and to the generation profile of solar. Solarhas a steep generation duration curve with many hours at ornear maximum output, resulting in a RLDC with many highly

negative hours. This leads to the high level of curtailmentshown in Fig. 12. These results are consistent with work byHeide et al., who modelled a highly renewable Europe andconcluded that a wind-to-solar energy ratio greater than onereduced storage requirements [60].

The annual resource quality of wind can impact a system'sability to reach its renewable targets. As seen in Table 5,Configuration 3 of the Reference Scenario simulated with aLow wind year does not meet the 80% target, while a Highwind year results in nearly 84% penetration. Table 5 showsthat the relative change in energy from storage between the

scenarios e approximately 10% e is quite significant. Thedifference in stored energy comes from charging events last-ing longer than 12 h, which can be seen in Fig. 10 (b). In theLow wind year, 11% of the annual stored energy comes fromevents lasting longer than 24 h. This is compared to 33% in theHigh wind year, in which one event lasts nearly five days.

Conclusions

This study explores the relationship between VRE and elec-trolyser capacity, and electrolyser operation under highrenewable energy system penetrations. The system investi-gated in this research represents the electric power system inthe Canadian province of Alberta with 80% VRE energy pene-tration in the year 2050. This study uses the simplified repre-sentation of the Alberta electricity system and PtG storage. Atime-varying electricity demand is met by wind, solar PV,and/or CCGT generators and delivered by the electrical grid.

The PtG storage system has three main components: charger,reservoir, and discharger. In our model, the charger is anelectrolyser that operates on excesswind and solar generationto produce hydrogen. The reservoir is a proxy for the naturalgrid, which we assume to be extensive and thus model itsimply as an unlimited reservoir for hydrogen.

We show that, for a system with 80% VRE penetration, PtGstorage can reduce VRE capacity by 23% and curtailment by87% compared to a system without storage. The overall sys-tem capacity benefits of including electrolyser capacity aremaximized with initial deployments; defined by associated

reduction in VRE capacity required to achieve 80% penetra-tion. Incremental electrolyser capacity results in decreasingVRE capacity reductions; however it significantly reducescurtailment.

Analysis of storage operation reveals charging eventslasting from 1 h to several days. Events lasting 12 h and longercontribute most of the annual stored energy, demonstratingthe value of long-term storage such as PtG. On the other hand,the large number of charging events lasting less than 12 hsuggests that medium-term storage technologies such as

Table 5 e System performance of Reference Scenario configuration 3 under different wind years.

Wind year Annualcapacityfactor

Electrolyserutilisation

factor

Chargingevents

Curtailedenergy

Energy fromVRE direct (Ed)

Energy fromstorage (Es)

VRE penetration(Fvre)

[%] [% full load hours] [# events] [TWh] [% annual energy] [% annual energy] [% annual energy]

Low (2014) 31.2% 39.5% 469 22.1 72.3% 3.5% 75.8%Average (2013) 35.7% 45.3% 448 25.3 76.1% 4.0% 80.1%High (2011) 41.3% 50.4% 437 31.6 79.3% 4.4% 83.8%

0

15

30

45

60

75

Reference -Config. 3

Localizedwind

Distributedsolar

High solarfraction

Capa

city

[GW

]

Scenario(a)

Rooftop solar

Commercial solar

Wind region 4

Wind region 3

Wind region 2

Wind region 1

0

15

30

45

60

Reference -Config. 3

Localizedwind

Distributedsolar

High solarfraction

Curta

iled

ener

gy [T

Wh]

Scenario(b)

Fig. 12 e (a) VRE capacity required for 80% VRE penetration with an electrolyser capacity of 3.06 GW. (b) The associatedcurtailed energy in each scenario.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91976

batteries, or flexible PEM electrolysers for PtG, may be more

appropriate for some of the storage duty.The hydrogen generated by the systems in this study is

between 1% and 20% of the current annual global productionof hydrogen by electrolysis; which would be a considerablefeat for a small jurisdiction. If injected into Alberta's naturalgas grid, the annual average hydrogen concentrationwould beless than 5%; below the suggested maximum level of blendingbefore adverse effects. Other jurisdictions may require alter-native PtG possibilities to avoid high hydrogen concentra-tions, such asmethanation, fuel cell vehicles, or industrial useof the hydrogen.

Geographic diversity of wind farms is shown to reduce theVRE capacity required to achieve 80% penetration, however,the same benefit is not found for solar PV. Increasing the solarfraction of total VRE capacity from 50% to 65% requires addi-tional VRE capacity to reach 80% penetration. A systemdesigned to achieve 80% penetration in an average wind yearis shown to reach 76% penetration in a lowwind year, and 84%penetration in a high wind year.

This study has provided a high-level, simplified analysis ofthe opportunity and challenges associated with meeting highrenewable penetrations with PtG. Further research regarding

the temporal concentrations of hydrogen, costs of generationand storage capacity, and alternative or complementarystorage technologies would provide valuable insights.

Acknowledgments

The authors gratefully acknowledge the financial support ofthe Pacific Institute for Climate Solutions (PICS).

Appendix A. Supplementary data

Supplementary data related to this article can be found athttps://doi.org/10.1016/j.ijhydene.2017.11.162.

r e f e r e n c e s

[1] IPCC. Energy systems. In: Climate change 2014: mitigation ofclimate change. Contribution of working group III to the fifthassessment report of the Intergovernmental panel onclimate change; 2014. Cambridge, United Kingdom and NewYork, USA.

[2] Kriegler E, Weyant JP, Blanford GJ, Krey V, Clarke L,Edmonds J, et al. The role of technology for achieving climatepolicy objectives: overview of the EMF 27 study on globaltechnology and climate policy strategies. Clim Change2014;123:353e67.

[3] IEA. Key world energy statistics. 2016. Paris, France.[4] IEA. World energy outlook 2016. 2016. Paris, France.[5] IEA. World energy outlook 2015. 2015. Paris, France.[6] EIA. International energy outlook 2016. 2016. Washington,

DC.[7] Sailor W, Bodansky D, Braun C, Fetter S, van der Zwaan B. A

nuclear solution to climate change? Science 19 May2000;288(5469):1177e8. https://doi.org/10.1126/

science.288.5469.1177 [Online]. Available: http://www.sciencemag.org/content/288/5469/1177.short.

[8] IEA. “Tracking clean energy progress 2013: IEA input to theclean energy ministerial,” organisation for economic co-operation and development. 2013. Paris andWashington, DC.

[9] Rogner H-H, Aguilera RF, Archer C, Bertani R,Bhattacharya SC, Dusseault MB, et al. “Chapter 7-energyresources and potentials,” in global energy assessment -toward a sustainable future. In: Cambridge, UK and NewYork, NY, USA and the International Institute for appliedsystems analysis, Laxenburg, Austria. Cambridge UniversityPress; 2012. p. 423e512.

[10] Rose SK, Kriegler E, Bibas R, Calvin K, Popp A, Van Vuuren DP,et al. Bioenergy in energy transformation and climatemanagement. Clim Change 2014;123:477e93.

[11] Luderer G, Krey V, Calvin K, Merrick J, Mima S, Pietzcker R,et al. The role of renewable energy in climate stabilization:results from the EMF27 scenarios. Clim Change 2014;123:427e41.

[12] IEA. Technology roadmap: geothermal heat and power. 2011.Paris, France.

[13] German federal ministry for economic affairs and energy.Berlin, Germany: Renewable Energy Sources Act - RES Act2014; 2014.

[14] Government of Denmark, “Independent from fossil fuels by2050”. [Online]. Available: http://denmark.dk/en/green-living/strategies-and-policies/independent-from-fossil-fuels-by-2050. [Accessed 14 October 2016].

[15] “GCAMWiki documentation”. [Online]. Available: http://jgcri.github.io/gcam-doc/. [Accessed 02 November 2017].

[16] Fawcett AA, Iyer GC, Clarke LE, Edmonds JA, Hultman NE,McJeon HC, et al. Can Paris pledges avert severe climatechange? Science Dec. 2015;350(6265):1168e9.

[17] MNP. Integrated modelling of global environmental change.An overview of IMAGE 2.4. Bilthoven, The Netherlands:Netherlands Environmental Assessment Agency (MNP); 2006.

[18] van Vuuren DP, Stehfest E, den Elzen MGJ, Kram T, vanVliet J, Deetman S, et al. RCP2.6: exploring the possibility tokeep global mean temperature increase below 2!C. ClimChange Nov. 2011;109(1e2):95e116.

[19] Pietzcker R, Ueckerdt F, Luderer G, Scholz Y, Gils H, Carrara S,et al. Evaluating the capacity of Integrated AssessmentModels (IAMs) to represent system integration challenges ofwind and solar power. Int Energy Workshop 2016;2016.

[20] Lopes JAP, Hatziargyriou N, Mutale J, Djapic P, Jenkins N.Integrating distributed generation into electric powersystems: a review of drivers, challenges and opportunities.Electr Power Syst Res Jul. 2007;77(9):1189e203.

[21] Lund PD, Lindgren J, Mikkola J, Salpakari J. Review of energysystem flexibility measures to enable high levels of variablerenewable electricity. Renew Sustain Energy Rev May2015;45:785e807.

[22] Fink S, Mudd C, Porter K, Morgenstern B. Wind energycurtailment case studies: May 2008eMay 2009. 2010.Columbia, USA.

[23] Denholm P, Hand M. Grid flexibility and storage required toachieve very high penetration of variable renewableelectricity. Energy Policy Mar. 2011;39(3):1817e30.

[24] Bathurst GN, Strbac G. Value of combining energy storageand wind in short-term energy and balancing markets. ElectrPower Syst Res 2003;67(1):1e8.

[25] Zhang G, Wan X. A wind-hydrogen energy storage systemmodel for massive wind energy curtailment. Int J HydrogenEnergy 2014;39(3):1243e52.

[26] Loisel R, Mercier A, Gatzen C, Elms N, Petric H. Valuationframework for large scale electricity storage in a case withwind curtailment. Energy Policy 2010;38(11):7323e37.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1977

[27] Qadrdan M, Abeysekera M, Chaudry M, Wu J, Jenkins N. Roleof power-to-gas in an integrated gas and electricity system inGreat Britain. Int J Hydrogen Energy 2015;40(17):5763e75.

[28] Dıaz-Gonz!alez F, Sumper A, Gomis-Bellmunt O,Villaf!afila-Robles R. A review of energy storage technologiesfor wind power applications. Renew Sustain Energy Rev2012;16(4):2154e71.

[29] de Boer HS, Grond L, Moll H, Benders R. The application ofpower-to-gas, pumped hydro storage and compressed airenergy storage in an electricity system at different windpower penetration levels. Energy 2014;72:360e70.

[30] International Electrotechnical Commission. Electrical energystorage. 2011. Geneva.

[31] Bockris J. The hydrogen economy: its history. Int J HydrogenEnergy Feb. 2013;38(6):2579e88.

[32] Scott DS. Smelling land: the hydrogen defense againstclimate catastrophe. Queen's Printer Publishing; 2008.

[33] Winter C-J. Hydrogen energy d abundant, efficient, clean: adebate over the energy-system-of-change. Int J HydrogenEnergy Jul. 2009;34(14):S1e52.

[34] Dickinson RR, Battye DL, Linton VM, Ashman PJ, Nathan GGJ.Alternative carriers for remote renewable energy sourcesusing existing CNG infrastructure. Int J Hydrogen Energy2010;35(3):1321e9.

[35] Varone A, Ferrari M. Power to liquid and power to gas: anoption for the German Energiewende. Renew Sustain EnergyRev 2015;45:207e18.

[36] G€otz M, Lefebvre J, M€ors F, McDaniel Koch A, Graf F, Bajohr S,et al. Renewable power-to-gas: a technological and economicreview. Renew Energy 2016;85:1371e90.

[37] Guti!errez-Martın F, Rodrıguez-Ant!on LM. Power-to-SNGtechnology for energy storage at large scales. Int J HydrogenEnergy Nov. 2016;41(42):19290e303.

[38] Olateju B, Kumar A. Hydrogen production from wind energyin Western Canada for upgrading bitumen from oil sands.Energy 2011;36(11):6326e39.

[39] Olateju B, Monds J, Kumar A. Large scale hydrogenproduction from wind energy for the upgrading of bitumenfrom oil sands. Appl Energy 2014;118:48e56.

[40] Walker SB, van Lanen D, Fowler M, Mukherjee U. Economicanalysis with respect to power-to-gas energy storage withconsideration of various market mechanisms. Int J HydrogenEnergy 2016;41(19):7754e65.

[41] Naterer G, Fowler M, Cotton J, Gabriel K. Synergistic roles ofoff-peak electrolysis and thermochemical production ofhydrogen from nuclear energy in Canada. Int J HydrogenEnergy Dec. 2008;33(23):6849e57.

[42] Gahleitner G. Hydrogen from renewable electricity: aninternational reviewof power-to-gas pilot plants for stationaryapplications. Int J Hydrogen Energy 2012;38:2039e61.

[43] Ulleberg Ø, Nakken T, Et!e A. The wind/hydrogendemonstration system at Utsira in Norway: evaluation ofsystem performance using operational data and updatedhydrogen energy system modeling tools. Int J HydrogenEnergy 2010;35(5):1841e52.

[44] Genc G, Celik M, Serdar Genc M. Cost analysis of wind-electrolyzer-fuel cell system for energy demand in Pnarbas‚-Kayseri. Int J Hydrogen Energy 2012;37(17):12158e66.

[45] Kaldellis JK, Zafirakis D. Optimum energy storage techniquesfor the improvement of renewable energy sources-basedelectricity generation economic efficiency. Energy2007;32(12):2295e305.

[46] Korpas M, Greiner CJ. Opportunities for hydrogen productionin connection with wind power in weak grids. Renew Energy2008;33(6):1199e208.

[47] Bennoua S, Le Duigou A, Qu!em!er!e M-M, Dautremont S. Roleof hydrogen in resolving electricity grid issues. Int JHydrogen Energy 2015;40(23):7231e45.

[48] Floch PH, Gabriel S, Mansilla C, Werkoff F. On the productionof hydrogen via alkaline electrolysis during off-peak periods.Int J Hydrogen Energy 2007;32(18):4641e7.

[49] Mansilla C, Dautremont S, Shoai Tehrani B, Cotin G, Avril S,Burkhalter E. Reducing the hydrogen production cost byoperating alkaline electrolysis as a discontinuous process inthe French market context. Int J Hydrogen Energy 2011;36(11):6407e13.

[50] Mansilla C, Louyrette J, Albou S, Barbieri G, Collignon N,Bourasseau C, et al. Electric system management throughhydrogen production-a market driven approach in theFrench context. Int J Hydrogen Energy2012;37(15):10986e91.

[51] Mansilla C, Louyrette J, Albou S, Bourasseau C,Dautremont S. Economic competitiveness of off-peakhydrogen production today - a European comparison. Energy2013;55:996e1001.

[52] Cany C, Mansilla C, da Costa P, Mathonniere G. Adapting theFrench nuclear fleet to integrate variable renewable energiesvia the production of hydrogen: towards massive productionof low carbon hydrogen? Int J Hydrogen Energy May2017;42(19):13339e56.

[53] Vandewalle J, Bruninx K, D’haeseleer W. Effects of large-scale power to gas conversion on the power, gas and carbonsectors and their interactions. Energy Convers Manag Apr.2015;94:28e39.

[54] Mukherjee U, Elsholkami M, Walker S, Fowler M, Elkamel A,Hajimiragha A. Optimal sizing of an electrolytic hydrogenproduction system using an existing natural gasinfrastructure. Int J Hydrogen Energy Aug.2015;40(31):9760e72.

[55] Bussar C, St€ocker P, Cai Z, Moraes Jr L, Magnor D, Wiernes P,et al. Large-scale integration of renewable energies andimpact on storage demand in a European renewable powersystem of 2050dsensitivity study. J Energy Storage 2016;6:1e10.

[56] K€otter E, Schneider L, Sehnke F, Ohnmeiss K, Schr€oer R. Thefuture electric power system: impact of power-to-gas byinteracting with other renewable energy components. JEnergy Storage 2016;5:113e9.

[57] Karlsson K, Meibom P. Optimal investment paths for futurerenewable based energy systems-using the optimisationmodel Balmorel. Int J Hydrogen Energy 2008;33(7):1777e87.

[58] Meibom P, Karlsson K. Role of hydrogen in future NorthEuropean power system in 2060. Int J Hydrogen Energy2010;35(5):1853e63.

[59] Jentsch M, Trost T, Sterner M. Optimal use of power-to-gasenergy storage systems in an 85% renewable energyscenario. Energy Proc 2014;46:254e61.

[60] Heide D, von Bremen L, Greiner M, Hoffmann C,Speckmann M, Bofinger S. Seasonal optimal mix of wind andsolar power in a future, highly renewable Europe. RenewEnergy 2010;35(11):2483e9.

[61] Government of Alberta. Natural gas facts and stats. 2015[Online]. Available: http://www.energy.alberta.ca/NaturalGas/Gas_Pdfs/FactSheet_NGFacts.pdf. [Accessed 13September 2016].

[62] Ursua A, Gandia LM, Sanchis P. Hydrogen production fromwater electrolysis: current status and future trends. Proc IEEE2012;100(2):410e26.

[63] Carmo M, Fritz DL, Mergel J, Stolten D. A comprehensivereview on PEM water electrolysis. Int J Hydrogen Energy Apr.2013;38(12):4901e34.

[64] EIA. Updated capital cost estimates for utility scale electricitygenerating plants. 2013. Washington, DC.

[65] Frost W, McCrank D. Wind integration in Alberta: market &operational framework implementation. AESO stakeholderinformation session. 2007. Calgary, Canada.

i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 91978

[66] Holttinen H. Hourly wind power variations in the nordiccountries. Wind Energy 2005;8(2):173e95.

[67] MacDonald AE, Clack CTM, Alexander A, Dunbar A,Wilczak J, Xie Y. Future cost-competitive electricity systemsand their impact on US CO2 emissions. Nat Clim Chang2016;(January):1e6.

[68] Staffell I, Pfenninger S. Using bias-corrected reanalysis tosimulate current and future wind power output. Energy2016;114:1224e39.

[69] AESO. AESO 2016 long-term outlook. 2016. Calgary, Canada.[70] IEA. Technology roadmap: energy storage. 2014. Paris,

France.[71] Sherif SA, Barbir F, Veziroglu TN. Wind energy and the

hydrogen economydreview of the technology. Sol Energy2005;78(5):647e60.

[72] Bertuccioli L, Chan A, Hart D, Lehner F, Madden B, Standen E.Development of water electrolysis in the European Union.2014. Lausanne, Switzerland.

[73] U.S. Department of Energy. Report of the hydrogen productionexpert panel: a subcommittee of the hydrogen & fuel celltechnical advisory Committee. 2013. Washington, DC.

[74] Melaina MW, Antonia O, Penev M. Blending hydrogen intonatural gas pipeline networks: a review of key issues. 2013.

[75] Altfeld K, Pinchbeck D. Admissible hydrogen concentrationsin natural gas systems. 2013.

[76] NaturalHY European Project (FP6). Using the existing naturalgas system for hydrogen. 2004.

[77] Soder L, Hofmann L, Orths A, Holttinen H, Wan Y, Tuohy A.Experience from wind integration in some high penetrationareas. Energy 2007;22(1):4e12.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 4 3 ( 2 0 1 8 ) 1 9 6 6e1 9 7 9 1979