IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 1

Evaluating Offshore Electricity Market Design

Considering Endogenous Infrastructure Investments:
Zonal or Nodal?
Michiel Kenis, Vladimir Dvorkin, Member, IEEE, Tim Schittekatte, Kenneth Bruninx, Member, IEEE,
Erik Delarue, Member, IEEE, Audun Botterud, Member, IEEE
arXiv:2405.13169v1 [econ.GN] 21 May 2024

Abstract—Policy makers are formulating offshore energy in- E electrolyzers

frastructure plans, including wind turbines, electrolyzers, and N Nodes
HVDC transmission lines. An effective market design is crucial L AC lines
to guide cost-efficient investments and dispatch decisions. This
paper jointly studies the impact of offshore market design H DC lines
choices on the investment in offshore electrolyzers and HVDC H1 Cross-border DC lines
transmission capacity. We present a bilevel model that incorpo- T Timesteps
rates investments in offshore energy infrastructure, day-ahead Parameters
market dispatch, and potential redispatch actions near real-time M Cg Marginal cost of generator g [e /MWh]
to ensure transmission constraints are respected. Our findings
demonstrate that full nodal pricing, i.e., nodal pricing both Un Utility of consumption as end product at node n [e /MWh]
H2
onshore and offshore, outperforms the onshore zonal combined Ue,t Utility of consumption at electrolyzer e [e /MWh]
with offshore nodal pricing or offshore zonal layouts. While Ce , Ch Unit investment cost of electrolyzer e or line h [e /MW]
combining onshore zonal with offshore nodal pricing can be F̄hDC Maximum capacity of a new DC line h [MW]
considered as a second-best option, it generally diminishes the D̄n,t Demand for electricity as end product at node n [MW]
profitability of offshore wind farms. However, if investment costs H2
of offshore electrolyzers are relatively low, they can serve as D̄e,t Maximum capacity of electrolyzer e [MW]
catalysts to increase the revenues of the offshore wind farms. LFg,t Load factor of generator g [-]
This study contributes to the understanding of market designs nP T DFln ,zP T DFlz Nodal and zonal PTDF [-]
for highly interconnected offshore power systems, offering in- f¯lAC Capacity of AC line l [MW]
sights into the impact of congestion pricing methodologies on RAMl Remaining Available Margin of AC line l [MW]
investment decisions. Besides, it is useful towards understanding
the interaction of offshore loads like electrolyzers with financial N T Ch Net Transfer Capacity of DC line h [MW]
support mechanisms for offshore wind farms. a Redispatch cost mark-up [-]
Ih,n , Ih,z Connection between DC line h and node n or zone z [-]
Index Terms—Multi-level modeling, Market design, Flow-
based market coupling, Offshore wind power, Electrolysis Il,n , Il,z Connection between AC line l and node n or zone z [-]
Variables
T IC Transmission investment cost [e ]
N OMENCLATURE EIC electrolyzers’ investment cost [e ]
Sets St Economic market surplus [e ]
G Generators Rt Redispatch cost [e ]
f¯hDC Capacity of DC line h [MW]
The work of M. Kenis is supported by a Ph.D. grant provided by the Flemish
Institute for Technological Research (VITO). d¯H2
e Capacity of electrolyzer e [MW]
M. Kenis and E. Delarue are with the Division of Applied Mechanics and yg,t Market dispatch of generator g [MW]
Energy Conversion (TME), Department of Mechanical Engineering, KU Leu- dn,t Market dispatch of end consumption at node n [MW]
ven, Celestijnenlaan 300 (box 2421), 3001, Leuven, Belgium. They are also
with the EnergyVille, Thor Park 8310, 3600, Genk, Belgium. M. Kenis is also dH2
e,t Market dispatch of consumption at electrolyzer e [MW]
AC DC
with the Flemish Institute for Technological Research (VITO), Boeretang 200, fl,t ,fh,t Flow captured by the market on line l or h [MW]
B-2400, Mol, Belgium. (e-mail: {michiel.kenis,erik.delarue}@kuleuven.be). r
yg,t Real-time production of generator g [MW]
V. Dvorkin is with the Department of Electrical Engineering and Computer r
Science, University of Michigan, Ann Arbor, MI 48109, USA (e-mail: dn,t Real-time end consumption at node n [MW]
[email protected]). dH2,r
e,t Real-time consumption at electrolyzer e [MW]
T. Schittekatte is with the MIT Sloan School of Management, Massachusetts U D
Institute of Technology, Cambridge, MA 02142, USA, and is with the Florence
yg,t ,yg,t Up/down redispatch of generator g [MW]
School of Regulation, European University Institute, Firenze, FI 50133, Italy dU D
n,t ,dn,t Up/down redispatch of end consumption at node n [MW]
(e-mail: [email protected]). H2,U H2,D
de,t ,de,t Up/down redispatch of electrolyzer e [MW]
K. Bruninx is with the Faculty of Technology, Policy and Management, TU AC,r DC,r
Delft, 2628 Delft, The Netherlands. He is also with the Division of Applied fl,t ,fh,t Real-time flow on line l or h [MW]
Mechanics and Energy Conversion (TME), Department of Mechanical Engi-
neering, KU Leuven, Celestijnenlaan 300 (box 2421), 3001, Leuven, Belgium
(e-mail: [email protected]). I. I NTRODUCTION
A. Botterud is with the Laboratory for Information and Decision Systems,
Massachusetts Institute of Technology, Cambridge MA 02142, USA (e-mail:
[email protected]). T HE total installed offshore wind power capacity world-
wide amounted to 56 GW in 2021 and is expected to
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 2

TABLE I: Overview of market design options. These properties have a significant impact on the resulting
Market design Onshore Offshore
prices, planned production schedule and required redispatch.
Literature on offshore electricity market design is limited
Full Nodal Pricing (FNP) Nodal Nodal
Offshore Nodal Pricing (ONP) Zonal Nodal and share a number of simplifying assumptions [7]–[9]. Firstly,
Offshore Zonal Pricing (OZP) Zonal One offshore zone while the market design determines the efficiency of the dis-
Full Zonal Pricing (FZP) Zonal Nodes part of onshore zone patch of generators, the spatial granularity of the power prices
also impacts investment decisions [10]. The latter is ignored
in existing studies on offshore pricing [7]–[9]. Secondly, flow-
strongly increase in the future [1]. Besides, other technologies based market coupling (FBMC) is the target methodology in
like storage and electrolyzers are expected to be deployed Europe’s zonal electricity markets to calculate and allocate AC
offshore [2], partly as offshore energy hubs [3]. In these hubs, transmission capacity in wholesale electricity markets while
the wind farms are connected with an interconnector, typically current studies typically consider a ‘Net Transfer Capacities’-
using the high-voltage direct current (HVDC) technology [4]. methodology [11], [12]. Inclusion of DC transmission capacity
In addition to government plans, electricity prices guide under FBMC requires, due to operational transmission char-
investments. The market design1 , which differs across regions, acteristics, an adjusted flow-based methodology to calculate
determines how offshore electricity wholesale prices are set. and allocate both AC and DC transmission capacity called
Under nodal pricing, the transmission network is incorporated Advanced Hybrid Coupling [13]. Thirdly, novel technologies
in the market clearing resulting in potentially unique prices per like offshore electrolyzers are not considered in current studies
node as used in, e.g., most U.S. electricity markets. In contrast, on the offshore pricing methodology except in Lüth et al.
under zonal pricing, network constraints are simplified in the [14] who show that offshore bidding zones facilitate offshore
market clearing and one price per predefined bidding zone is hydrogen production. With many projects ongoing to install
calculated. Bidding zone borders typically align with national such technologies offshore, it is important to consider them
borders as in the EU. in the analysis because they serve as the only, or at least
Under the status quo in the European zonal day-ahead a very important, offshore electricity demand [2]. Hence,
electricity marktet, offshore energy hubs would be part of electrolyzers have the potential to have a large impact on
the mainland zone they are (administratively) connected with. offshore prices [15].
This means that the power prices in those hubs would equal This paper addresses the gaps identified above by proposing
mainland prices of the country the hub belongs to. However, an optimization model that considers all these three elements.
the EU strategy for the deployment of offshore infrastructure Specifically, we present a bilevel optimization model that
opts for (an) off-shore market zone(s) to integrate these assets captures four temporal stages: (i) offshore HVDC transmission
in the European electricity market [5]. In contrast, along the investment, (ii) offshore electrolyzer investment, (iii) day-
US East Coast, the natural choice is that offshore nodal pricing ahead market clearing, and (iv) redispatch actions. We assume
will be opted for to integrate future offshore infrastructure into a fixed capacity target for offshore wind generation investment
the wholesale electricity market [6]. There are generally four as these decisions are typically not market-driven. Our results
different market design options to integrate offshore hubs as reveal that ONP and OZP can serve as valuable alternatives
Tab. I summarizes. to FZP. However, these schemes come at the cost of a lower
Literature on onshore market design does not apply to an profitability for existing offshore wind farms due to a lower
offshore context because (i) the offshore market participants average offshore price. The latter benefits the business case
mainly consist of (wind power) generators and (ii) the domi- for potential offshore electricity consumers like, e.g., offshore
nant transmission technology is HVDC as opposed to onshore electrolyzers, which can contribute to decrease the welfare gap
AC technology. The former implies that electricity prices are between ideal FNP, on the one hand, and ONP and OZP on
set by either near-zero marginal cost wind farms or by the the other hand. As a result, policy makers should carefully re-
willingness-to-pay of a very limited set of consumers when consider support mechanisms for, e.g., offshore wind power
the constraints on the exchanges between the onshore and under ONP and OZP.
offshore system are binding in the market clearing mechanism. The paper is structured as follows. Section II discusses the
The latter means that flows on HVDC lines, as well as their existing literature. Section III presents the optimization model,
impact on the AC lines, can be accurately captured in the while Section IV demonstrates the model by applying it to
flow-based market clearing problem via Advanced Hybrid a stylized electricity grid. Finally, Section V discusses the
Coupling, in contrast to their AC counterparts. Note, however, implications of our findings and Section VI concludes.
that by definition only HVDC connections connecting two
market zones are monitored in the market clearing problem.
Intra-market zone offshore HVDC lines are not observed by II. L ITERATURE
the market, whereas congestion on internal AC lines can be
avoided by including them in the calculation of the limitations There are two main literature streams that relate to this pa-
on cross-market zone exchanges, i.e., the flow-based domains. per: literature on (i) existing optimization models that capture
both the investments in and the operation of electricity grids,
1 In this paper, we use the term ‘market design’ to refer to methodologies and (ii) offshore electricity market design with a focus on the
used for pricing and congestion management. spatial granularity of wholesale electricity prices.
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 3

t=1 t=2 t=N-1 t=N t=1 t=2 t=N-1 t=N

Transmission Generation Market clearing for each t Redispatch for each t

investment investment

Fig. 1: Typical composition of optimization models in literature that capture transmission and generation investments as well
as the operation of the considered electricity grid.

A. Capacity expansion models significant attention in the EU after the European Commission
With regards to onshore electricity grids, the existing liter- highlighted its “strategy to harness the potential of offshore
ature has proposed optimization models that encompass trans- renewable energy for a climate neutral future” [5]. Several
mission and generation investments, as well as the operation of studies suggest creating offshore bidding zones to enable
the electricity grid. Figure 1 shows the typical composition of efficient flows of renewable energy to high-cost areas, reduce
these models capturing four stages: transmission investment, redispatch costs, and enhance operational security [7]–[9],
generation investment, market clearings and redispatch actions. [21], [22]. Tosatto et al. [22] incorporate FBMC in their
The structure of these optimization models varies in the models and demonstrate that under offshore bidding zones,
literature based on underlying assumptions. The hierarchical offshore electricity prices would decrease while congestion
or non-hierarchical relationship between agents, each with rents would increase. However, none of the aforementioned
distinct objectives and decision variables, categorizes such studies endogenously model how investments in generation
models. Many papers assume that investors strategically con- and/or transmission assets are impacted by the different price
sider the impact of their investments on the power system’s signals under alternative market designs for congestion man-
operation, resulting in hierarchical models [16]–[18]. Non- agement.
hierarchical models, on the other hand, do not incorporate Hardy et al. [12] are the first to present a capacity expansion
strategic anticipation [12]. and operational market model within the context of offshore
These approaches are, however, not suitable to analyze electricity market design. They show that investors should
offshore power systems, especially when connected to zonal assume that offshore bidding zones will be implemented
electricity markets (onshore) subject to FBMC as in Europe. because the resulting power system (topology as well as
FBMC considers the limited capacity of critical intra-zonal or generating assets) also performs well under changing pricing
inter-zonal AC lines as a constraint in the market clearing methodologies. There are three key differences between their
algorithm in a simplified way [19]. Specifically, the mar- work and our paper. First, in contrast to [12], we assume that
ket representation of the flows on AC transmission network investors in transmission capacity strategically anticipate the
elements is capped by their Remaining Available Margins impact on the market clearing and redispatch actions. Second,
(RAMs), parameters that TSOs provide. The methodology is Hardy et al. [12] do not consider offshore electrolysers. Lastly,
being rolled out in most European countries [11]. To consider their model neglects FBMC when assessing the performance
(offshore) DC lines, an adjusted flow-based market clearing of offshore bidding zones.
algorithm is required because the flow on DC lines depends
on how the AC/DC converter is operated. Two options are III. M ODEL
being considered: Standard Hybrid Coupling (SHC – which is This section presents the high-level structure of the model
currently used) and Advanced Hybrid Coupling (AHC – target that is first organized as a quad-level problem, then refor-
methodology) [13]. AHC, in contrast to SHC, considers the mulated to a bilevel problem. First, we list the assumptions
impact of DC flows on the market representation of AC flows. that underpin our model. Next, we present the mathematical
We refer the reader to Schonheit et al. [11] for an exhaustive formulation of the bilevel model. Finally, we provide a solution
guide on the functioning on FBMC, and to Muller et al. [13] strategy.
for the compatibility with DC lines.
Lete et al. [20] includes FBMC in capacity expansion, but
they do not consider DC lines. To the best of the authors’ A. Assumptions
knowledge, our paper is the first to consider both AC and There are four stages and we assume these stages occur
DC lines in a capacity expansion and operation market model sequentially: HVDC transmission investment (by the TSO),
by applying AHC. Moreover, an additional novelty is that we electrolyzer investment (by a merchant investor), market clear-
optimize investments in both HVDC transmission capacity and ing (by the market operator), and redispatch (by the TSO) (Fig.
offshore electrolysers within the same model. 1). Each level anticipates the impact of its decision on later
stages, while taking the decision variables of all higher levels
as given. We assume perfect foresight in all stages.
B. Offshore electricity market design We assume a fixed capacity for offshore wind generation
The literature on integrating offshore energy infrastructure as such capacity decisions are typically not market-driven, but
into existing onshore zonal electricity markets has gained prescribed in tenders. Similarly, we assume an existing fleet
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 4

deciding on the HVDC transmission capacity. The second

level maximises the net utility of newly added electrolyzers
while deciding on the eletrolysers’ capacities. The third level
maximises the surplus from the day-ahead market with the
generators’ dispatch (and possible load shedding) as decision
variable. Finally, the fourth level minimises redispatch costs
including cost mark-ups assuming cost-based redispatch.
2) From a quad-level to a bilevel problem: We transform
the quad-level problem to a bilevel problem in two steps as Fig.
2 visualises. The assumption of perfect competition between
participants in the wholesale electricity market allows merg-
ing the second (electrolyzers’ investments) and third (market
clearing) level optimization problems into one optimization
problem. The merged problem maximizes the surplus from
the wholesale market less the investment cost of electrolyzers.
The equivalence has been shown in literature for the case
of generators’ investments but is equivalent for electrolyzers’
Fig. 2: From a quad-level problem to a bilevel problem by investments [23].
assuming perfect competition between generators. The colors Secondly, we note that the decision variables of the fourth
map to the different stages in Fig. 1. level (redispatch) are independent from the second (electrolyz-
ers’ investments) or third (market clearing) level optimization
problems, but not from the first (transmission investment)
of conventional generators connected to the onshore nodes. level. As a consequence, the redispatch problem can be con-
Investments are onshore assets is not considered. sidered to be happening on the same hierarchical level as the
As we seek to compare the performance of optimal system combined electrolyzer investment & market clearing problem.
configurations under different market designs, we consider the This is similar to the approach of Pandzic et al. [24] in the
HVDC capacity as a variable. Fixing the HVDC capacity context of reserve capacity markets and balancing markets.
under different market designs may lead to, e.g., over- or
underestimating the benefits for FNP, depending on the grid
topology and capacity of the HVDC elements. In the HVDC B. A bilevel problem
investment stage, we assume that there is no existing HVDC
transmission capacity. We do not consider the lumpiness of We present the mathematical equations of the bilevel prob-
network investments and only investments between predefined lem consisting of two lower level problems. The lower level
nodes are allowed (see Section IV-A). Electrolyzer capacity is representing the combined electrolyzer investment & market
limited to a predefined upper bound per offshore node. clearing problem comes in two forms: either considering full
nodal pricing or full zonal pricing. The cases of ONP and OZP
The market clearing can adopt four pricing methodologies:
are special cases of FZP in which each offshore node or all
FNP, ONP, OZP and FZP. There is perfect competition be-
offshore nodes together represent (a) separate market zone(s).
tween participants in the wholesale electricity market. Only
thermal generation units face a non-zero variable cost M Cg . 1) Upper Level: transmission investment: The decision
The marginal utility of demand for electricity to produce variables are the HVDC transmission capacity f̄hDC of each
hydrogen depends on the hydrogen price and transport cost, candidate line h P H and the transmission investment cost
which are assumed to be fixed. Under zonal pricing, flow- TIC. The decision variables are written in a bold font. We
based market coupling and Advanced Hybrid Coupling are assume no existing HVDC transmission capacity.
considered. We assume that all AC lines are critical and, hence,
the transmission limits of all AC lines are (albeit imperfectly) ÿ” ı
max St ´ Rt ´ EIC ´ TIC (1a)
considered in the market algorithm. We refer the reader to
tPT
Schönheit et al. [11] for an exhaustive guide on flow-based
s.t.
market coupling. ÿ ÿ
H2 H2
Redispatch actions only serve to adjust the generators’ St “ Un dn,t ` Ue,t de,t
dispatch in case their output as cleared in the day-ahead ÿnPN ePE

market would lead to infeasible power flows. This is only ´ M Cg yg,t @t (1b)
required if the day-ahead market does not accurately capture gPG
ÿ “ ‰ ÿ H2
the transmission constraints, i.e., under ONP, OZP an FZP. We Rt “ Un ¨ dn,t ´ drn,t ` Ue,t ¨
assume cost-based redispatch. nPN ePE
ÿ
1) The quad-level problem: Under the assumptions above, ´ dH2,r
“ ‰ “ ‰
dH2
e,t e,t ` M Cg ¨ r
yg,t ´ yg,t @t (1c)
the resulting decision problem can be represented as a quad- gPG
level problem (Fig. 2). The first level, as shown in the left ÿ
EIC “ Ce d¯H2
e @g, @t (1d)
panel of Fig. 2, maximises the overall social welfare while ePE
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 5

ÿ
TIC “ Ch f̄hDC @n, @t (1e) Objective (2a) maximises the surplus from the wholesale
hPH1 market St less the electrolyzers’ investment cost EIC. Equa-
0ď f̄hDC ď F̄hDC @h (1f) tion (2b) defines both terms in the objective. Equation (2c)
puts limitations on the new capacity of offshore electrolyz-
Objective (1a) maximises the overall social welfare. This is ers. Equations (2d)-(2f) limit the dispatch of generators and
reasonable as a transmission investment is typically a strongly consumers to the available power capacities.
regulated process [25]. Social welfare consists of the economic
surplus of the wholesale electricity market St less redispatch
costs Rt , electrolyzer investment cost EIC and transmis- In case of nodal pricing:
” ÿ
sion investment cost TIC. Eqs. (1b)-(1e) define each of the AC
fl,t “
ÿ
nP T DFln yg,t ´ dn,t
terms. Equation (1f) poses constraints to the new transmission nPN gPGpN q
capacity. Investments in transmission capacity only occur if ÿ ÿ ı
the investment cost is outweighed by the operational benefits ´ dH2
e,t ´ DC
fh,t ¨ Ih,n @l, @t (3a)
(higher welfare and/or lower redispatch costs). ePEpnq hPH

The economic surplus of the wholesale electricity market St AC

|fl,t | ď f¯lAC @l, @t (3b)
equals the utility associated with consuming electricity less the DC
|fh,t | ď f¯hDC @h, @t (3c)
operational generation costs. The former splits into the utility ÿ ÿ
of demand for electricity as end-product Un ¨ dn,t and demand yg,t ´ dn,t ´ dH2
e,t
H2
for electricity as a source for electrolysis Ue,t ¨ dH2
e,t . The
gPGpnq
ÿ
ePEpnq
ÿ
marginal utility of demand for electricity to produce hydrogen DC AC
´ fh,t ¨ Ih,n ´ fl,t ¨ Il,n “ 0 @n, @t (3d)
depends on the hydrogen price and transport cost. We further hPH lPL
assume that only thermal generation units face a non-zero Equation (3a) defines the flow on each AC line using
variable cost M Cg . nodal PTDFs. Ignoring the DC approximation, this is an exact
The redispatch cost Rt is the net cost for a TSO to adjust projection of the physical flow as opposed to the case of zonal
the dispatch schedule corresponding the day-ahead market pricing. The final term considers the injection of the DC grid
to prevent line overloading. Cost-based redispatch implies into the AC grid. Ih,n equals 1 (or -1) if the flow on DC-line
that a generator, requested by the TSO to produce more h goes outward (or inward) node n, and is otherwise 0. Eqs.
than scheduled by the market operator, is paid its marginal (3b) and (3c) limits the flow on all AC and DC lines by their
cost M Cg by the TSO while a generator, requested by the physical capacities. Finally, Eq. (3d) presents the nodal power
TSO to produce less than scheduled by the market operator, balance.
pays back its avoided marginal cost to the TSO but keeps
its profit margin. A similar concept holds for consumers: if
the covered demand is lower than scheduled by the market In case of zonal pricing:
H2
” ÿ
operator, consumers are paid their lost utility Un or Ue,t , and
ÿ ÿ ÿ
AC
fl,t “ zP T DFlz yg,t ´ dn,t ´ dH2
e,t
vice versa. Finally, the electrolyzers’ investment cost EIC zPZ gPGpzq nPN pzq ePEpzq
and transmission investment cost TIC scale linearly with the ÿ
DC
ı ÿ
DC
”
nph´q
installed capacity. ´ fh,t Ih,z ` fh,t nP T DFl ¨ Ih,nph´q
hPH hPH1
2) Lower Level 1: electrolyzers’ investment & market clear- ı
np´hq
ing: Problem (2) presents a first part of the first lower level ` nP T DFl ¨ Ih,np´hq @l, @t (4a)
problem in mathematical form. Eqs. (3), in case of nodal AC
|fl,t | ď RAMl @l, @t (4b)
pricing, and Eqs. (4), in case of zonal pricing, present the
DC
remaining constraints. The decision variables are the new |fh,t | ď N T Ch @h P H1 , @t (4c)
capacity of electrolyzers d̄H2
ÿ ÿ ÿ
e as well as the dispatch of yg,t ´ dn,t ´ dH2
e,t
all generators yg,t . Other decision variables are the covered gPGpzq nPZ ePEpzq
demand for electricity as end-product dn,t and electricity for ÿ
DC
ÿ
AC
´ fh,t ¨ Ih,z ´ fl,t ¨ Il,z “ 0 @z, @t (4d)
electrolysis dH2
e,t as well as the power flows on all AC lines
AC DC hPH lPL
fl,t and DC-lines fh,l .
Equation (4a) defines the flow on each AC line using zonal
ÿ PTDFs2 . Zonal PTDFs are parameters, set by TSOs, that are
max St ´ EIC (2a) predictive and imperfect by definition as they aggregate nodal
tPT information. As a consequence, flAC does not capture the
s.t. physical power flow but rather an imperfect heuristic, hereafter
Eq. (1b), Eq. (1d) (2b) referred to as the commercial flow. The impact of the power
flow on a DC line is captured following Advanced Hybrid
0ď d̄H2
e ď D̄eH2 @e (2c)
Coupling [13]. The flow on an AC line l as a result of a
0 ď yg,t ď LFg,t ȳg @g, @t (2d) flow on a DC line h is calculated with nP T DFl
nph´q
and
0 ď dn,t ď D̄n,t @n, @t (2e)
2 Following flow-based market coupling, only the flow on critical AC lines
0 ď dH2 ď d¯H2
e,t e @e, @t (2f) should be monitored and capped. We assume here that all AC lines are critical.
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 6

np´hq
nP T DFl with nph´q and np´hq the starting/ending optimisation problem is to determine the required adjustments
end of DC-line h. Ih,nph´q and Ih,np´hq equal 1 (or -1) if to the market outcome only to avoid line overloadings, not to
the flow on DC-line h is defined as flowing into (or away correct a possibly inefficient dispatch from the market. This is
from) node nph´q or np´hq, and is zero otherwise. Equation in line with real world redispatch actions in zonal wholesale
(4b) limits the commercial power flow on each AC line by markets [26]. Therefore, by imposing cost mark-ups we avoid
the Remaining Available Margin (RAM), i.e., the part of redispatch when there is no congestion. The redispatch cost
the physical transmission capacity that can be traded in the Rt , and thus social welfare, in the upper level problem is the
wholesale market [11]. Equation (4c) limits the flow on each true redispatch costs, i.e., without cost mark-up a.
cross-border DC line by the Net Transfer Capacity (NTC). Eqs. (5b)-(5d) define the final dispatch, after redispatch, of
r
Finally, Equation (4d) presents the zonal power balance. all generators yg,t as well as the final covered demand for
3) Lower Level 2: redispatch: The decision variables are electricity as end-product drn,t and electricity for electrolysis
r H2,r
the final dispatch, after redispatch, of all generators yg,t as de,t . Eqs. (5e)-(5j) provide limits to the upward or down-
well as the final covered demand for electricity as end-product ward adjustment taking the market outcome (from lower level
drn,t and electricity for electrolysis dH2,r
e,t . We also consider 1) as given. Finally, similar to Eqs. (3a)-(3d), Eqs. (5k)-(5n)
explicit variables for the upward and downward adjustment to present the definition of the power flow on each AC line, the
U D H2,U
the wholesale market dispatch yg,t , yg,t , dU D
n,t , dn,t , de,t limitations to the AC and DC power flows, and the nodal
and dH2,D
e,t
AC,r
. Finally, fl,t DC,r
and fh,l are the physical AC and power balance in real-time, i.e., after redispatch.
DC power flow after redispatch.
C. Solution Approach
ÿ ÿ“ ‰ U “ ‰ D
min M Cg ` a ¨ yg,t ` ´ M Cg ` a ¨ yg,t We follow the approach of Kazempour and Conejo [27],
tPT gPG and apply the Benders’ decomposition to a bilevel problem,
considering f¯hDC as the complicating variable. The approach
ÿ“
` a ¨ dH2,U ` a ¨ dH2,D
H2
‰ “ H2 ‰
` ´ Ue,t e,t ` Ue,t e,t
ePE works as follows:
ÿ “
1) Given fixed HVDC transmission investment decisions,
‰ “ ‰ D
` ´ Un ` a ¨ dU
n,t ` Un ` a ¨ dn,t (5a)
nPN we solve two single-level reformulations of the bilevel
problem. Specifically, after fixing the complicating vari-
s.t. able, we first solve the mixed-integer reformulation of
r U D the simplified MPEC which now only has lower-level
yg,t “ yg,t ` yg,t ´ yg,t @g, @t (5b)
decision variables. Subsequently, we solve a continuous
drn,t “ dn,t ` dU D
n,t ´ dn,t @n, @t (5c) yet nonlinear version of the same problem, now for-
dH2,r “ dH2 H2,U
´ dH2,D mulated using the strong duality theorem, wherein the
e,t e,t ` de,t e,t @e, @t (5d)
U variables causing bilinear terms are fixed to the values
0 ď yg,t ď LFg,t ¨ ȳg ´ yg,t @g, @t (5e)
derived from the solution of the first MPEC. This, in
D
0 ď yg,t ď yg,t @g, @t (5f) turn, allows the extraction of sensitivities of welfare w.r.t.
U
0 ď dn,t ď D̄n ´ dn,t @n, @t (5g) HVDC investments — which is the main motivation
behind solving two MPECs at every iteration, as in [27].
0 ď dD n,t ď dn,t @n, @t (5h)
2) The sensitivities obtained in step 1 allow for formulating
H2,U
0 ď de,t ď d¯H2 H2
e ´ de,t @e, @t (5i) a so-called Benders’ master problem whose solution
0 ď dH2,D ď dH2 @e, @t (5j) provides updated HVDC investment decisions.
e,t e,t
AC,r
ÿ ” ÿ 3) Step 1-2 are repeated until no more economic welfare
fl,t “ nP T DFln r
yg,t improvement can be achieved.
nPN gPGpN q
ÿ ÿ ı Our framework resembles that of Kazempour and Conejo
´ drn,t ´ dH2,r
e,t ´ DC,r
fh,t ¨ Ih,n @l, @t (5k) [27], with some minor differences that do not affect the
ePEpnq hPH solution strategy. The operational decisions in our problem are
AC,r
|fl,t | ď f¯lAC @l, @t (5l) not optimized by the upper level investor, as in [27], but rather
DC,r
by a system operator. Another distinction lies in our aim to
|fh,t | ď f¯hDC @h, @t (5m) maximize societal welfare as opposed to minimizing negative
ÿ ÿ H2,r
r
yg,t ´ drn,t ´ de,t ´ profit and the absence of scenarios due to our deterministic
gPGpnq ePEpnq framework. While our model is distinct in these ways, it does
ÿ DC,r
ÿ AC,r not alter the solution strategy proposed in [27].
fh,t ¨ Ih,n ´ fl,t ¨ Il,n “ 0 @n, @t (5n)
hPH lPL
IV. C ASE S TUDY
Objective (5a) minimises the redispatch cost considering a
cost mark-up a to avoid redispatch actions when the outcome A. Set-up
of the day-ahead market does not lead to overloading of Figure 3 presents the network layout. We consider three
network elements. This differs from existing papers that do mainland zones, each consisting of three nodes, and three
not consider a mark-up [16]–[18]. Indeed, the aim of the offshore nodes. We consider 30 time steps with varying
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 7

1 w.r.t. the market surplus under FNP and the lowest electrolyzer
Consumer
Thermal generator unit investment cost. The difference between the market sur-
2 plus and
Wind farm ř overall welfare (Eq. (1a)) consists of the redispatch
AC transmission line cost t Rt , the electrolyzer investment cost GIC, and the
3 H2 Electrolyser transmission investment cost T IC.
DC transmission line
4 H2 We make three observations. Firstly, ONP and OZP out-
perform FZP in terms of welfare despite a higher market
7 surplus and a lower spending on electrolyzer and transmission
H2
5 8
H2
6 capacities under FZP. Specifically, the market surplus under
10 12 FZP amounts to 97.9%, 96.4% and 96.3% of the reference
9
market surplus for the three investment costs of electrolyzers,
11 while it ranges from 87.7% to 96.3% under ONP and from
88.7% to 96.6% under OZP. The market surplus is higher
Fig. 3: Electricity grid. The symbols indicate the location of under FZP compared to ONP and OZP because more grid
generators, consumers and transmission lines with a black constraints are considered in the day-ahead market under
color for existing elements (AC transmission lines, thermal the latter two cases. Besides, the combined investments in
generation units, consumers and offshore wind farms) and a electrolyzer and transmission capacities are consistently lower
yellow color for elements of which the capacity is a decision under FZP (0.7%, 1.5% and 0.1% of the reference market
variable (DC transmission lines and electrolyzers). surplus) compared to other market designs. As a consequence,
the difference in overall welfare between ONP and OZP, on
Perfect monitoring Imperfect monitoring Not monitoring the one hand, and FZP, on the other hand, is entirely driven by
lower redispatch costs Rt due to a better grid representation
in the market clearing problem (see Fig. 4). The redispatch
cost under FZP is 20.9%, 23.3% and 25.4% of the reference
market surplus for the three investment costs of electrolyzers,
(a) FNP (b) ONP (c) OZP (d) FZP while it is 9.7% at most under other market designs.

Fig. 4: Accuracy of monitoring the flows on the AC and DC

Welfare Redispatch Cost
transmission lines in the market clearing algorithm under the Electrolyser Investment Cost Transmission
1 Investment Cost
1 1
four different pricing methods.
-] [
ƩSt, ƩRt, GIC, TIC (normalized)

0,95 0,95
Market Surplus [-]

0,9 0,9
consumption levels and load factors of offshore wind. The
0,85 0,85
size of the case study is similar to related literature [16]–[18].
0,8 0,8
Appendix A presents the assumed values of all parameters in
the algorithm. The algorithm, applying Benders’ decomposi- 0,75 0,75

tion, converges after less than 20 iterations and a total run time 0,7 0,7

of less than 104 seconds for all cases considering a tolerance 0,65 0,65
of 10´1 . FNP ONP OZP FZP FNP ONP OZP FZP FNP ONP OZP FZP
Low
Low Medium High Medium High
We consider three unit electrolyzer investment cost levels Investment Cost of Electrolysers
-] [
Investment cost of electrolyzers [€/MW]
(low, medium and high) to illustrate the possible role of ř
offshore electrolysis under each of the four electricity market Fig. 5: Market surplus t St , normalised w.r.t. the market
designs in Tab. I. surplus under nodal pricing and lowest electrolyzer investment
Figure 4 shows how each pricing method implies a different cost. The overall welfare (Eq. (1a)) is indicated in blue.
monitoring of the flows on the transmission lines in the
market clearing. While FNP perfectly monitors all flows, the
other pricing methods involve (partly) zonal pricing following 3000
FNP
3000

FBMC. Therefore, the flows on the AC lines (L) are imper- ONP
HVDC capacity [MW]

2500 2500
OZP
fectly monitored using zonal PTDFs under ONP, OZP and FZP
2000 2000
FZP. The DC-lines follow the NTC approach and are only
monitored if they are cross-border lines (H1 ). 1500 1500

1000 1000

B. Results 500 500

We discuss market designs implications on (i) the overall 0 0

Low Medium High
welfare, (ii) the profitability of offshore wind farms and (iii) Investment cost of electrolyzers [€/MW]
offshore electricity prices.
1) Implications on welfare: Figure 5 presents the market Fig. 6: HVDC capacity.
ř
surplus t St , aggregated over all time steps and normalised
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 8

1800 1800 2 2
FNP FNP
1600 1600 1,8 1,8
ONP ONP
Electrolyzer capacity [MW]

OZP 1,6 OZP 1,6

1400 1400
FZP 1,4 FZP 1,4
1200 1200

Profit [M€]
1,2 1,2
1000 1000
1 1
800 800
0,8 0,8
600 600
0,6 0,6
400 400 0,4 0,4
200 200 0,2 0,2

0 0 0 0
Low Medium High Low Medium High
Investment cost of electrolyzers [€/MW] Investment cost of electrolyzers [€/MW]

Fig. 7: Electrolyzer capacity. Fig. 8: Operational profit of the offshore wind farms.

FNP (redispatch) ONP (redispatch) OZP (redispatch) FZP (redispatch)

A more accurate representation of the transmission con- FNP (in-market) ONP (in-market) OZP (in-market) FZP (in-market)

straints in the market clearing algorithm allows capturing 0,7 0,7

the contribution of potential HVDC investments to increased 0,6 0,6

Curtailment [-]
welfare and reduced redispatch costs. Recall that intra-zonal 0,5 0,5

HVDC links are not considered in the market clearing problem 0,4 0,4

0,3 0,3
under OZP or FZP. Under ONP, onshore zonal pricing restricts
0,2 0,2
flows on the offshore network elements, limiting the attainable
0,1 0,1
wellfare gains (see also Section IV-B3). In our case study 0 0
a more accurate grid representation results in more HVDC Low Medium High
Investment cost of electrolyzers [€/MW]
capacity (Fig. 6) while a market clearing problem with the
most crude grid representation relies on the possibility for Fig. 9: Curtailment of offshore wind energy, expressed as a
redispatch to avoid congestion. The HVDC capacity under fraction of the available offshore wind energy. We distinguish
ONP, OZP and FZP is lower because the additional market between curtailment as part of the market clearing and the
surplus as a result of installing more HVDC capacity is limited redispatch actions.
under ONP, OZP and FZP compared to FNP. It does, however,
lower the redispatch cost.
Secondly, welfare increases with a decrease in the unit farms. Firstly, the profit is the highest under FZP closely
investment cost of electrolyzers regardless of the market followed by FNP, while profits are significantly lower under
design. This is because a higher electrolyzer capacity, which ONP and particularly OZP. For example, in case of a low
is shown in Fig. 7, allows generating a larger market sur- investment cost of electrolyzers, the profit amounts to 1.72 Me
plus, outweighing the larger electrolyzer investment cost. The under FZP, followed by 1.66 Me (FNP), 1.34 Me (ONP)
market surplus grows with 2.1% (FNP), 8.6% (ONP), 8.0% and 0.60 Me (OZP). Secondly, the profit increases with a
(OZP) and 1.6% (FZP) of the reference market surplus while decrease in investment cost of electrolyzers regardless of the
the change in combined spending in electrolyzers and HVDC market design. Specifically, it grows with 0.10 Me (FNP),
lines is less than 1.4% under all market designs. In addition, 0.79 Me (ONP), 0.32 Me (OZP) and 0.15 Me (FZP) when
a higher electrolyzer capacity allows reducing the redispatch the investment cost decreases from high to low. Thirdly, the
cost, which drops with 1.3% (ONP), 2.6% (OZP) and 4.4% profit increase is the largest under ONP with 0.79 Me . The
(FZP) of the reference market surplus. drivers behind these effects are curtailment and day-ahead
Thirdly, the welfare gain associated with a lower unit invest- offshore electricity prices, which we discuss in the following.
ment cost of the electrolyzers is larger under ONP (10.1% of Figure 9 presents the wind energy curtailment as part of
the reference market surplus) and OZP (11.6% of the reference the market clearing and the redispatch actions, expressed as
market surplus) than under FNP or FZP. Lower offshore a fraction of the available offshore wind energy. Curtailment
wholesale prices under ONP and OZP compared to the other under FZP is at least 6 times higher than under other other
market designs trigger higher investments in electrolyzers (see market designs. This is a consequence of the crude represen-
further) and thus an increasing impact of their investment costs tation of grid constraints in the market as the majority of
on the total welfare. Hence, the electrolyzer capacity is higher the curtailment under FZP occurs as part of the redispatch
under FNP, OZP and ONP than under FZP.3 actions (see also redispatch costs in Fig. 5). The presence of
2) Implications on wind farm & electrolyzer profitability: electrolyzers limits the need for curtailment as it serves as an
Figure 8 presents the operational profit for all offshore wind electricity consumer at the same node at which wind power is
3 Note that the bilevel modeling approach inherently allows multiple optimal
generated. Specifically, we find a negative correlation of -0.52
solutions to the lower level. The upper level problem may ”exploit” this between wind power curtailment and electricity consumption
multiplicity of solutions and select the lower level solution that maximizes the by electrolyzers.
upper level objective. Similarly, this characteristic of bilevel models implies By design, the electrolyzers do not secure a net profit unless
that while the reported welfare and HVDC investment are optimal, this may
be achieved through different electrolyzer capacity investments, day-ahead the optimal electrolyzer capacity equals the assumed upper
market clearing and redispatch actions. limit. In case an electrolyzer’s capacity equals this upper
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 9

bound, its operating profits can outweigh its investment cost drives up the average offshore price under all market designs.
indicating that a higher electrolyzer capacity would increase This is the result of the combination of two drivers: (i) because
social welfare. If the electrolyser capacity does not equal its thermal generation units produce more electricity and therefore
upper limit, operating profits offset the investment costs and can set the price more often, and (ii) because the willingness-
the electrolysers’ net profit is zero. to-pay of electrolyzers Ue,tH2
(27 e /MWh in this case study)
3) Implications on offshore electricity prices: Figure 10 sets the price more often. The increase of the average offshore
presents the day-ahead electricity prices for offshore wind price with decreasing investment costs of electrolyzers is the
farms (node 4, 5 and 6). Offshore electricity prices are, largest under ONP and OZP. Specifically, it increases with
averaged over time and offshore nodes, the highest under 96% (from 31.6 e /MWh to 61.8 e /MWh) under ONP and
FZP. Specifically, the average price amounts to 71.5 e /MWh 81.4% (from 25.8 e /MWh to 46.8 e /MWh) under OZP
under FZP compared to 65.2 e /MWh under FNP at a low compared to 4% (from 62.7 e /MWh to 65.2 e /MWh) under
investment cost, to 69.3 e /MWh under FZP compared to 64 FNP and 3% (from 69.3 e /MWh to 71.5 e /MWh) under
e /MWh under FNP at a medium investment cost and to 69.3 FZP. This is because the electrolyzers drive up the offshore
e /MWh under FZP compared to 62.7 e /MWh under FNP at electricity price such that the level of price convergence (see
a high investment cost. Average prices are the highest under Fig. II) increases the most under these market designs, which
FZP because the flow on HVDC lines between an offshore is reflected in the market surplus that also witnesses the largest
and an onshore node (3-4, 6-7 and 5-12) is not monitored increase under ONP and OZP with decreasing investment costs
nor constrained in the market clearing problem. Hence, price of electrolyzers. As a result, the presence of a consumer at an
convergence with the onshore zones occurs by construction. offshore node in the form of an electrolyzer can contribute to
Contrary, the flows on all HVDC lines between an offshore the profitability of offshore wind farms.
and an onshore node (3-4, 6-7 and 5-12) are monitored and Finally, the offshore HVDC capacity impacts the level of
constrained in the market clearing problem under FNP, ONP electricity prices (10) and price convergence. Table II shows
and OZP because they serve as cross-zonal HVDC lines. As the percentage of time during which a specific amount of
a result, electricity prices in the offshore zone(s) are zero unique prices exists. One unique price implies full price con-
whenever the commercial transmission capacity restricts flows, vergence across all nodes/zones. The largest amount of hours
lowering the average offshore prices under FNP, ONP and OZP in which there are one or two unique prices consistently occurs
compared to FZP. under FNP compared to ONP, OZP and FZP. Specifically, this
Despite that the flows on all HVDC lines between an is during 80%, 86.7% and 76.7% of time for the three levels
offshore and an onshore node (3-4, 6-7 and 5-12) are moni- of investment costs of electrolyzers compared to at most 20%
tored and constrained in the market clearing problem under under any other market design. This is contra-intuitive because
FNP, ONP and OZP, offshore prices are different between FNP allows for up to 12 unique prices. However, FNP signals
these three pricing methods. Specifically, the average price is scarcity of transmission capacity better than any other market
consistently higher under FNP (65.2 e /MWh, 64 e /MWh design leading to the most HVDC deployment, hence, the
and 62.7 e /MWh) than under ONP (61.8 e /MWh, 43.7 highest level of price convergence.
e /MWh and 31.6 e /MWh) which in turn is higher than under
OZP (46.7 e /MWh, 40.8 e /MWh and 25.8 e /MWh). As
V. D ISCUSSION
expected, the electrolyzer capacity is inversely correlated with
the average electricity price per offshore node and timestep. Our results show that FNP is superior to ONP, OZP and FZP
We identify two drivers explaining the price differences in terms of economic welfare. Yet, there exist political hurdles
between FNP, ONP and OZP: (i) the (optimised) HVDC that prevent the implementation of FNP in many regions [28].
transmission capacity, and (ii) not considering offshore internal Instead, FZP is the active pricing methodology in many regions
grid constraints under OZP. Firstly, under FNP, the flows on including, e.g., the EU, UK, Brazil and Australia. Many of
both AC and DC lines are perfectly monitored and limited. these hurdles are rooted in concerns around the redistributional
Contrary, under ONP and OZP, the flows on the onshore impacts for existing consumers and generators. In an offshore
AC lines are imperfectly monitored and limited due to the context, there are no or very few existing generators/loads. We
characteristics of a flow-based market clearing. The zonal would therefore expect less opposition against ONP or OZP,
character of the onshore market implies that the flows on the which our case study reveals to be valuable alternatives to FZP.
offshore DC lines are restricted more than under FNP (see Specifically, despite leading to a lower market surplus and an
Eq. (4a)). This leads to less value for HVDC transmission increased spending on HVDC transmission capacity and elec-
capacity (see Fig. 6). Hence, there is less price convergence trolyzer capacity, these market designs increase overall welfare
between the offshore nodes and the onshore zones. Secondly, by significantly decreasing offshore wind energy curtailment
the average offshore price is lower under OZP than under ONP and redispatch costs, originating either offshore or onshore.
because, under OZP, congestion between an offshore node and The latter is the result of a more accurate representation
the connecting onshore node immediately results in a price of grid constraints in the market clearing algorithm under
difference between all three offshore nodes and the onshore ONP and OZP compared to FZP. Offshore wind developers
zone. This drives down the average offshore price because the remain skeptical of ONP and OZP as they fear lower average
marginal cost of offshore wind farms is assumed to be zero. electricity prices, as well as higher price and volume risks due
A lower unit investment cost of electrolyzers consistently to congestion. Our results indeed show tat the overall increase
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 10

Low Medium High

Investment cost of electrolyzers [C/MW]

Fig. 10: Prices for offshore wind farms. The boxes show the 25th and 75th quantiles and the median while the dashed line
shows the average.

TABLE II: Percentage of time during which a specific amount of unique prices exists. One unique price implies full price
convergence across all nodes/zones.
Low Medium High
FNP ONP OZP FZP FNP ONP OZP FZP FNP ONP OZP FZP
1 unique price 6.7% 3.3% 0% 0% 20% 0% 0% 0% 20% 0% 0% 0%
2 unique prices 73.3% 16.7% 0% 6.7% 66.7% 0% 0% 20% 56.7% 0% 0% 20%
3 unique prices 10% 60% 90% 93.3% 13.3% 43.3% 80% 80% 23.3% 13.3% 50% 80%
4 unique prices 10% 20% 10% - 0% 46.7% 20% - 0% 70% 50% -
5 unique prices 0% 0% - - 0% 10% - - 0% 16.7% - -

in welfare comes at the cost of a lower profitability of existing curtailment and increases the offshore average price, leading
offshore wind farms. The lower profitability is explained by a to an increased profitability of offshore wind farms. Therefore,
lower average offshore price as congestion between onshore there is an important interaction between support mecha-
and offshore nodes is signaled via price differences with a nisms for offshore wind and potential support mechanisms
certain spatial granularity. This effect is stronger under OZP for electrolyzers. The latter would reduce the need for the
compared to ONP. former because of the increasing offshore prices with more
Technically, the challenges with zonal and nodal offshore electrolyzers, especially under ONP and OZP.
market design are similar, and often related to the interaction
with other markets, such as the balancing market. With little
VI. C ONCLUSION
demand and no ”back-up” generators offshore, system oper-
ators structurally have to rely on balancing services provided Different strategies exist, or are part of the ongoing debate,
by neighboring zones or nodes. Notably, the controllability to integrate offshore energy hubs in wholesale electricity
of HVDC technology with regard to power flows offers markets. We develop a bilevel optimization model to simulate
Transmission System Operators (TSOs) more flexibility in the implications of four market designs: full nodal pricing,
managing grid constraints, potentially easing some technical offshore nodal pricing, offshore zonal pricing, and full zonal
challenges. pricing. The model captures four temporal stages: (i) offshore
A different market design than FZP would also impact HVDC transmission investment, (ii) offshore electrolyzer in-
public support mechanisms for offshore wind farms like, vestment, (iii) wholesale electricity market clearing, and (iv)
e.g., Contract-for-Differences (CfDs). Specifically, the cost redispatch actions. Offshore wind generation investment is
for governments would increase under CfDs assuming ceteris exogenously determined.
paribus. Under FZP, support for offshore wind developers We find that full nodal pricing is superior to offshore nodal,
is implicit in the high (socialized) redispatching costs, i.e., offshore zonal and full zonal pricing. Yet, there exist political
wind developers receive artificially high grid-agnostic day- hurdles that prevent the implementation of full nodal pricing
ahead electricity prices but may be curtailed in real-time. in many regions. Our case study reveals that offshore nodal
Alternative to increased support for offshore wind farms, our and offshore zonal pricing can serve as valuable alternatives
results suggest that implementing nodal pricing onshore in to full zonal pricing. Specifically, these market designs could
addition to ONP, which is identical to FNP, also brings the increase welfare as a consequence of decreased offshore wind
profitability of wind farms close to levels under FZP. energy curtailment and decreased redispatch costs despite a
Offshore consumers like, e.g., offshore electrolyzers can lower market surplus. However, the welfare gain comes at the
serve as a catalyst to decrease the remaining welfare gap cost of a lower profitability of existing offshore wind farms due
between FNP, on the one hand, and ONP and OZP on the to congestion between onshore and offshore nodes. The latter
other hand. A lower offshore average price under ONP and benefits the business case for offshore consumers like, e.g.,
OZP compared to FNP could allow offshore electrolyzers to offshore electrolyzers, which, in turn, can have a positive effect
enter the market, which in turn increases the market surplus on the offshore average price. Therefore, if proven profitable in
and reduces redispatch costs. Besides, it reduces wind energy real world power systems, offshore electrolyzers can serve as
IEEE TRANSACTIONS ON ENERGY MARKETS, POLICY AND REGULATION 11

a catalyst to decrease the remaining welfare gap between full [14] A. Lüth, Y. Werner, R. Egging-Bratseth, and J. Kazempour, “Electrolysis
nodal pricing, on the one hand, and offshore nodal and offshore as a flexibility resource on energy islands: The case of the north sea,”
2022, https://www.cbs.dk/files/cbs.dk/hydrogen flexiblity-lueth wp06
zonal pricing on the other hand. Policy makers and regulators final-al-2022 08 29.pdf.
should, hence, carefully coordinate electricity market design [15] R. Green, H. Hu, and N. Vasilakos, “Turning the wind into hydrogen:
with policies supporting the development of offshore energy The long-run impact on electricity prices and generating capacity,”
Energy Policy, vol. 39, no. 7, pp. 3992–3998, 2011, special Section:
infrastructure. Renewable energy policy and development.
Future work could focus on how balancing mechanisms [16] V. Grimm, A. Martin, M. Schmidt, M. Weibelzahl, and G. Zöttl,
should be set-up in under offshore nodal and zonal or full “Transmission and generation investment in electricity markets: The
effects of market splitting and network fee regimes,” European Journal
nodal pricing schemes, in which most balancing energy would of Operational Research, vol. 254, no. 2, pp. 493–509, 2016.
need to be provided by onshore assets in different bidding [17] T. Kleinert and M. Schmidt, “Global optimization of multilevel elec-
zones. Focal points in this discussion are the interaction tricity market models including network design and graph partitioning,”
Discrete Optimization, vol. 33, pp. 43–69, 2019.
with day-ahead electricity markets and transmission capacity [18] J. Egerer, V. Grimm, T. Kleinert, M. Schmidt, and G. Zöttl, “The impact
allocation, the impact on offshore wind farms and the use of neighboring markets on renewable locations, transmission expansion,
of offshore assets such electrolyzers as a source of balancing and generation investment,” European Journal of Operational Research,
vol. 292, no. 2, pp. 696–713, 2021.
energy. [19] K. Van den Bergh, J. Boury, and E. Delarue, “The Flow-Based Market
Coupling in Central Western Europe: Concepts and Definitions,” The
Electricity Journal, vol. 29, no. 1, pp. 24–29, 2016.
R EFERENCES [20] Q. Lété, Y. Smeers, and A. Papavasiliou, “An analysis of zonal electricity
pricing from a long-term perspective,” Energy Economics, vol. 107, p.
[1] WFO, “Capacity of offshore wind turbines in operation and under 105853, 2022.
construction in 2021, by country (in megawatts) [Graph],” 2022, [21] L. Kitzing and M. G. González, “Market arrangements
https://www.statista.com/statistics/264264/capacity-of-offshore-wind- for offshore wind energy networks,” 2020, [WWW
power-plants-worldwide-2010/. Document]. URL. https://orbit.dtu.dk/en/publications/
[2] North Sea Energy, “Energy hubs & transport infrastructure,” 2022, market-arrangements-for-offshore-wind-energy-networks.
https://north-sea-energy.eu/nl/home/. [22] A. Tosatto, X. M. Beseler, J. Østergaard, P. Pinson, and S. Chatzi-
vasileiadis, “North sea energy islands: Impact on national markets and
[3] North Sea Wind Power Hub, “Towards the first hub-and-spoke project,”
grids,” Energy Policy, vol. 167, p. 112907, 2022.
2021, https://northseawindpowerhub.eu/sites/northseawindpowerhub.eu/
[23] V. Grimm and G. Zoettl, “Investment incentives and electricity spot
files/media/document/NSWPH Concept%20Paper 05 2021 v2.pdf, ac-
market competition,” Journal of Economics & Management Strategy,
cessed on 03.02.2022.
vol. 22, no. 4, pp. 832–851, 2013.
[4] ENTSO-E, “ENTSO-E Position on Offshore Development: System Op-
[24] K. Pandžić, K. Bruninx, and H. Pandžić, “Managing risks faced by
eration & Governance,” 2021, https://www.entsoe.eu/2021/07/14/new-
strategic battery storage in joint energy-reserve markets,” IEEE Trans-
entso-e-paper-on-offshore-development-focuses-on-system-operation-
actions on Power Systems, vol. 36, no. 5, pp. 4355–4365, 2021.
governance, accessed on 23.06.2023.
[25] P. Joskow and J. Tirole, “Merchant transmission investment*,” The
[5] European Commission, “An eu strategy to harness the potential of
Journal of Industrial Economics, vol. 53, no. 2, pp. 233–264, 2005.
offshore renewable energy for a climate neutral future,” 2020, [WWW
[26] M. Weibezahl, “Nodal, zonal, or uniform electricity pricing: How to
Document]. URL. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/
deal with network congestion,” Frontiers in Energy, vol. 11, no. 2, pp.
?uri=CELEX:52020SC0273&from=EN.
210–32, 2017.
[6] J. P. Pfeifenberger, J. Delosa, L. Bai, C. Plet, C. Peacock, and R. Nelson,
[27] S. J. Kazempour and A. J. Conejo, “Strategic generation investment
“The benefit and urgency of planned offshore transmission: Reducing
under uncertainty via benders decomposition,” IEEE Transactions on
the costs of and barriers to achieving u.s. clean energy goals,” 2023,
Power Systems, vol. 27, no. 1, pp. 424–432, 2012.
[WWW Document]. URL. https://www.brattle.com/wp-content/uploads/
[28] A. Eicke and T. Schittekatte, “Fighting the wrong battle? a critical
2023/01/Brattle-OSW-Transmission-Report Jan-24-2023.pdf.
assessment of arguments against nodal electricity prices in the european
[7] THEMA Consulting Group, “Market Arrangements for debate,” Energy Policy, vol. 170, p. 113220, 2022.
Offshore Hybrid Projects in the North Sea,” 2020,
https://op.europa.eu/en/publication-detail/-/publication/
28ff740c-25aa-11eb-9d7e-01aa75ed71a1/language-en, accessed on A PPENDICES
24.02.2022.
[8] PROMOTioN - Progress on Meshed HVDC Offshore Transmission Appendix A: Parameters in the case study
Networks, “D12.4 Deployment plan for future European offshore
grid,” 2020, https://www.promotion-offshore.net/fileadmin/PDFs/D12. We assume that each thermal generation unit and each
4 - Final Deployment Plan.pdf, accessed on 24.02.2022. offshore wind farm has an installed capacity ȳg of 2000 MW
[9] The North Seas Countries’ Offshore Grid Initiative, “Discussion
Paper 2: Integrated Offshore Networks and The Electricity Target
and 1000 MW respectively. The operational cost M Cg of each
Model. Deliverable 3.” 2014, https://www.benelux.int/files/4514/0923/ thermal generation unit is 60 e /MWh at node 2, 80 e /MWh
4100/Market Arrangements Paper Final Version 28 July 2014.pdf, at node 9 and 100 e /MWh at node 10, while we assume it
accessed on 24.02.2022.
[10] V. Grimm, B. Rückel, C. Sölch, and G. Zöttl, “The impact of market
is zero for the offshore wind farms. Figure 11 presents the
design on transmission and generation investment in electricity markets,” assumed availability factor LFg,t at each offshore node as
Energy Economics, vol. 93, p. 104934, 2021. well as the price inelastic demand D̄n,t at each node and time
[11] D. Schönheit, M. Kenis, L. Lorenz, D. Möst, E. Delarue, and K. Bruninx, step. There are six peaks in both available wind energy and
“Toward a fundamental understanding of flow-based market coupling for
cross-border electricity trading,” Advances in Applied Energy, vol. 2, p. demand. However, both do not always peak at the same time
100027, 2021. steps. We assume that the utility to consume electricity as end
[12] S. Hardy, A. Themelis, K. Yamamoto, H. Ergun, and D. Van Hertem, product Un is 200 e /MWh and to produce hydrogen Ue,t H2
“Optimal grid layouts for hybrid offshore assets in the north sea under
different market designs,” IEEE Transactions on Energy Markets, Policy is 27 e /MWh. The penalisation parameter in the redispatch
and Regulation, pp. 1–12, 2023. problem a is 1000.
[13] C. Müller, A. Hoffrichter, H. B. Büchel, A. Schwarz, and A. Schnettler, We assume that each AC line has a thermal capacity of
“Integration of hvdc-links into flow-based market coupling: Standard
hybrid market coupling versus advanced hybrid market coupling,” in 700 MW and all have the same impedance. Both the RAMs
CIGRE Symposium, 2017. and NTCs are set identical to the thermal capacity of the
 12

considered line. The zonal PTDFs are derived from the nodal
PTDFs by assuming values for GSKs that are pro rata with
the generation capacity in a zone [19].
We assume three levels for the unit investment cost of
electrolyzers Ce : low (50 e /MW), medium (150 e /MW) and
high (250 e /MW). The maximal electrolyzer capacity D̄e,t
H2

is 600 MW per offshore node. The investment cost of a DC

line Ch is 20 e /MW. The maximal capacity of a new DC line
F̄hDC is 2000 MW.

1 4 7 4 6 5
Demand [MW]

1
Load Factor [-]

1000
0,5
500

0 0
1 6 11 16 21 26 1 6 11 16 21 26
Timestep [-] Timestep [-]

Fig. 11: Availability factors of the offshore wind farms and

price inelastic demand at each node and time step.