IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO.

2, MARCH 2023 1093

Annealing-Based Quantum Computing for

Combinatorial Optimal Power Flow
Thomas Morstyn , Senior Member, IEEE

Abstract—This paper proposes the use of annealing-based development of linear approximations [5] and convex relax-
quantum computing for solving combinatorial optimal power ations [6], which are accurate under specific conditions and
flow problems. Quantum annealers provide a physical comput- allow OPF problems to be solved in polynomial time. With the
ing platform which utilises quantum phase transitions to solve
specific classes of combinatorial problems. These devices have rise of DERs, there has been significant work to consider the
seen rapid increases in scale and performance, and are now particular features of OPF relevant at the distribution system
approaching the point where they could be valuable for indus- level, including unbalanced voltages, losses and reactive power
trial applications. This paper shows how an optimal power flows [7], [8], [9].
flow problem incorporating linear multiphase network modelling, Combinatorial OPF is a more challenging class of problem,
discrete sources of energy flexibility, renewable generation place-
ment/sizing and network upgrade decisions can be formulated which emerges when an OPF needs to be solved alongside
as a quadratic unconstrained binary optimisation problem, additional discrete decisions, such as when resources have dis-
which can be solved by quantum annealing. Case studies with crete flexibility [10], as well as where resource placement and
these components integrated with the IEEE European Low network investment decisions need to be made accounting for
Voltage Test Feeder are implemented using D-Wave Systems’ fixed costs and limited sizing options [11]. In practice, many
5,760 qubit Advantage quantum processing unit and hybrid
quantum-classical solver. DERs only offer flexibility in discrete increments, including
EV chargers [12] and heat-pumps [13] with on/off control,
Index Terms—Distribution network, D-Wave, electric vehicle, and schedulable appliances with fixed operating cycles [14].
optimal power flow, power system planning, quantum annealing,
quantum computing. smart charging. Also, even when power converters allow continuous con-
trol of DERs, low operating power is often associated with
low efficiency, making it desirable to impose a minimum
turn-on power [15]. Combinatorial OPF is directly relevant
I. I NTRODUCTION for distribution system operators (DSOs) seeking to increase
HE OPTIMAL deployment and operation of new sources the hosting capacity for clean energy technologies through
T of generation and flexibility is critical for achieving
a low-cost transition to reliable and decarbonised electrical
a combination of targeted network reinforcements and active
management of DERs [16].
power systems [1]. The scope for optimised planning and Combinatorial optimisation problems can be solved using
operation has expanded significantly due to the emergence exhaustive search and dynamic programming, but the curse of
of distributed energy resources (DERs), including small and dimensionality means that the computational burden increases
medium scale renewables and flexible loads, combined with exponentially with the number of decision variables [17].
the new availability of substation- and customer-level sensing Mixed Integer Linear Programming (MILP) can be applied in
and communications [2]. However, the vast potential scale of cases where the objective and constraints can be formulated
the resulting coordination challenge has created concern over as linear functions of the discrete variables [18]. Significant
future computational requirements [3]. progress has been made towards solving large MILPs to
Optimal power flow (OPF) problems involve finding set- reasonable levels of accuracy, but in general they remain com-
points for controllable power sources which meet demand at putationally intensive [19]. Lagrangian Relaxation [20] and
minimum cost, while satisfying resource and network con- Surrogate Lagrangian Relaxation [21] are iterative approaches
straints [4]. In general, the nonlinear characteristics of power suited to problems that can be decomposed into a set of sim-
networks makes this challenging, which has motivated the pler subproblems by relaxing a limited number of coupling
constraints. Combinatorial problems can also be solved using
metaheuristic methods including genetic algorithms [22], par-
Manuscript received 17 February 2022; revised 10 June 2022 and 14 August
2022; accepted 17 August 2022. Date of publication 22 August 2022; date ticle swarm optimisation [23], tabu search [24] and simulated
of current version 20 February 2023. This work was supported by the U.K. annealing [25]. However, scalability remains a challenge as
Engineering and Physical Sciences Research Council (EPSRC) under Project the convergence time of metaheuristic methods also increases
EP/S000887/2, Project EP/S031901/1, and Project EP/T028564/1. Paper no.
TSG-00227-2022. with the problem dimension [26].
The author is with the School of Engineering, The University of Edinburgh, Over the last 20 years, there has been significant progress
Edinburgh EH8 9YL, U.K. (e-mail: [email protected]). in the development of quantum devices which offer a funda-
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TSG.2022.3200590. mentally new computing architecture compared with classical
Digital Object Identifier 10.1109/TSG.2022.3200590 digital silicon-based computers. A major milestone towards
1949-3053 
c 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
1094 IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO. 2, MARCH 2023

this was the recent achievement of quantum supremacy with a for a larger scale problem, which highlights the value of
54-qubit device, i.e., the practical demonstration of a quan- co-optimising distribution network upgrades and renewable
tum computer solving a problem that would be infeasible generation investment with operational flexibility.
for classical computers [27]. For power systems, gate-based The rest of the paper is organised as follows: Section II
quantum computing algorithms are presented for generator presents a brief overview of D-Wave’s implementation of
unit commitment in [28], [29]. However, significant challenges quantum annealing and its application to QUBO problems.
remain for scaling up universal gate-based quantum comput- In Section III, the proposed QUBO formulation for the com-
ers to the point where they could be widely used for industrial binatorial OPF is developed. Case study results are presented
applications [30]. in Section IV. Section V concludes the paper.
The challenges of scaling up gate-based quantum comput-
ers has motivated the development of more scalable quantum II. Q UANTUM A NNEALING
hardware architectures aimed at specific computing problems.
This section provides an overview of quantum annealing, as
Quantum annealers are currently the largest quantum com-
implemented by D-Wave’s quantum processors. As mentioned,
puting devices and are capable of solving a specific class
a key application is to solve QUBO problems, which can be
of combinatorial optimisation, namely unconstrained quadratic
described by
binary optimisation (QUBO) problems [31]. Quantum anneal-  
ers incorporate a lattice of qubits which can be controllably min Qij xi xj + ci xi , (1)
biased and coupled. Based on the quantum adiabatic theorem, (i,j)∈E i∈X
the qubit lattice is controlled so that it physically evolves to a
where xi ∈ {0, 1}, i ∈ X := {1, . . . , X} are binary decision
low-energy state which represents the solution to an optimisa-
variables, E := {(i, j)|i, j ∈ X , i = j}. Qij ∈ R, (i, j) ∈ E are
tion problem. This is somewhat analogous to the process that
the quadratic QUBO objective function coefficients and ci ∈
is replicated by simulated annealing, but with physical quan-
R, i ∈ X are the linear QUBO objective function coefficients.
tum fluctuations replacing simulated thermal ones [32]. Also,
The QUBO problem can be equivalently expressed as
it should be noted that before the development of annealing-
an Ising model minimisation problem, through a change of
based quantum processors, quantum annealing was used to
variables yi = 1 − 2xi for i ∈ X [31], giving
refer to a variation of simulated annealing with simulated  
quantum fluctuations [33]. min Jij yi yj + hi yi ,
Theoretically demonstrating when noisy quantum anneal- (i,j)∈E i∈X
ing has a definitive advantage over classical alternatives is ⎛ ⎞
1 1⎝ 
challenging [34], but a performance advantage has been Jij = − Qij , hi = − ci + Qij ⎠, (2)
demonstrated for specific applications [35], [36]. Moreover, 4 2
j∈X
quantum annealing hardware is still in its infancy, and is
rapidly improving in terms of the number of qubits and noise where the spin values yi ∈ {−1, 1}, i ∈ X .
level [37]. This has motivated investigations into a range appli- Quantum annealing is based on the natural behaviour of
cations including protein folding [38], machine learning [39], coupled qubits to seek a ground state (lowest-energy state).
and wireless base station decoding [40]. The opportunity for The quantum annealing process can be described by a time
quantum annealing to be applied to power system applica- varying Hamiltonian H(s) [34]
tions is noted in [41], but without a detailed investigation. H(s) = A(s)HI − B(s)HP , (3)
In [42], quantum annealing is demonstrated for generator unit
commitment, but power flow modelling and network con- where A(s) and B(s) are annealing path functions, which are
straints are not considered. Other power system applications of defined in terms of the normalised annealing time s = t/ta .
quantum annealing include grid partitioning [43] and phasor These are designed so that initially A(0) = 1 and B(0) = 0,
measurement unit placement [44]. and after annealing A(1) = 0 and B(1) = 1.
The novel contribution of this paper is to propose and The initial Hamiltonian HI is selected so that it has a known
demonstrate the use of annealing-based quantum computing ground state which is easy to prepare, for example [34]
for combinatorial OPF. Given the still relatively limited scale 
and developing nature of quantum annealing hardware, our HI = σix , (4)
i
focus is on its applicability to power systems rather than
the potential for speed-up with current hardware. Towards where σix
is the Pauli-x operator applied to qubit i. The
this, a novel QUBO formulation is presented for a linear problem Hamiltonian HP is given by [34]
multiphase OPF problem with controllable on/off EV charg-  
HP = Jij σiz · σjz + hi σiz . (5)
ing, non-dispatchable renewable generation placement/sizing
i,j i
and network upgrade decisions, which can be solved using
quantum annealing. Case studies are implemented on D-Wave where σiz
is the Pauli-z operator applied to qubit i. The eigen-
Systems’ 5,760 qubit Advantage quantum processor to inves- vectors of this Hamiltonian correspond to the solutions of the
tigate how the number of required qubits scales with the Ising model (2).
number of EVs and network constraints. D-Wave’s hybrid The quantum annealer first initialises the superposition state
quantum-classical binary quadratic model solver is then used of a qubit lattice so that H(0) = HI . The qubit couplings are

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
MORSTYN: ANNEALING-BASED QUANTUM COMPUTING FOR COMBINATORIAL OPTIMAL POWER FLOW 1095

Fig. 2. A 5 qubit logical network representing a simple QUBO problem,

min (x1 + x2 + x3 + x4 + x5 )2 , and a 6 qubit embedding on a section of the
Pegasus topology.

which can solve QUBO models with up to a million

variables [48]. The solver code is proprietary, and is described
as making use of parallel computation on classical CPUs and
GPUs, which send queries to a quantum processor to help
guide the exploration of the solution space [50].
Fig. 1. A 3 by 3 cell example (144 qubits) Pegasus topology. D-Wave’s
Advantage processor is 16 by 16 cells (5,760 qubits).
III. P ROBLEM F ORMULATION
This section presents the proposed QUBO formulation for
then manipulated over the annealing time so that the system a combinatorial OPF problem. To provide a concrete setting,
evolves towards the problem Hamiltonian. Dwave’s device the formulation focuses on a DSO which aims to optimally
uses radio frequency superconducting quantum–interference schedule EVs with controllable on/off charging, while making
device (rf-SQUID) qubits [45]. The underlying physical mech- decisions about the location and sizing of non-dispatchable
anisms used by D-Wave’s quantum processors are described renewable generation and distribution network upgrades. Note
in more detail in [46]. According to the adiabatic theorem of that with fairly minor modifications the formulation could be
quantum computing, if the annealing time is sufficiently long updated to address transmission networks (e.g., by replacing
the time varying Hamiltonian will remain in the ground state the distribution-focused multiphase linear power flow model
throughout. The problem Hamiltonian has classical eigenval- used here [9] with a DC linear power flow model [51]) and
ues, and thus the spin values at H(1) = Hp will be classical a broader range of time-coupled flexible loads, such as smart
values (i.e., yi ∈ {−1, 1}) and these will correspond with the heating and schedulable appliances [52].
optimal solution of the Ising model. First, a constrained nonlinear binary formulation is
An added complexity is that the physical qubit lattices presented for the combinatorial problem, which is then used to
within D-Wave’s quantum processors are sparsely connected. develop the proposed QUBO formulation, which can be solved
In September of 2020, D-Wave released its ‘Advantage’ pro- using quantum annealing. Consider a distribution network with
cessor, with 5,760 qubits connected in a Pegasus topology, a set of nodes N = {0, . . . , N}, where node 0 is the point
which has most qubits connected to 15 neighbours. Fig. 1 of connection with the main grid. The network has phases
shows a 144 qubit version of the Pegasus topology [47]. The  = {a, b, c}. The set of intervals in the optimisation horizon
Advantage processor allows for significantly larger problems is T = {1, . . . , T}, where each interval has duration τ . For
than its predecessor, the D-Wave 2000Q, which had 2,048 interval t ∈ T , the price of energy at the grid connection point
qubits arranged with most qubits connected to 6 neighbours. is λ0t . V = {1, . . . , V} is the set of EVs, G = {1, . . . , G} is the
The limited connectivity means that an Ising model must set of potential renewable generation sites and U = {1, . . . , U}
be translated into an equivalent model that is compatible with is the set of mutually exclusive potential plans for network
the processor’s physical qubit lattice. The translation process upgrades.
is called minor embedding, and involves representing each The combinatorial OPF includes long-term investment deci-
logical qubit of the original model with either a single physi- sions ahead of operation and operational scheduling over the
cal qubit or a chain of strongly coupled physical qubits [48]. optimisation horizon. Computational limits generally require
A example is shown in Fig. 2. Optimal minor embedding is that the optimisation horizon considers a shortened period rep-
itself a computationally intensive problem, but heuristic tools resentative of longer-term operation (e.g., one or more days).
suitable for large problems have been developed [49]. Also, Here, only one set of investment decisions are considered, but
embeddings can be reused between problems with the same a potential extension is to consider multiple stages of invest-
coupled decision variables, even if the linear and quadratic ment and operation (e.g., yearly). Due to the difference in
weights are different. timescales, asset investment costs need to be discounted and
For problems that are too large for current quantum pro- adjusted based on the duration of the optimisation horizon
cessors, D-Wave makes available cloud-based hybrid solvers, relative to their lifetime (see, e.g., [53]). The equivalent cost

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
1096 IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO. 2, MARCH 2023

  g g
cequ for the optimisation horizon of an investment with upfront − τ ηiev uev
i ρi xit +
ev ev
cjs xjs
investment cost cinv is given by i∈V t∈Ti j∈G s∈Sj
 
 

+ cuk xku +
p
τ λ0t kt0 xku , (7a)
T r
cequ = cinv y y , (6) k∈U t∈T
T 1 − (1 + r)−TL 
s.t. E0i +
ev
τ ηi ρi xit
ev ev ev
≤ Ēiev for i ∈ V, (7b)
where T y is the number of optimisation intervals τ in one year, t∈Ti
y
TL is the asset lifetime in years and r is the discount rate. 1  
For EV i ∈ V, Ti ⊆ T is the subset of intervals when vω ≤ ṽωt + ρiev
ev ∈ {0, 1} |i |
ev
the EV is plugged in and available for charging, xiv i∈V ψ∈i ev

represents the on/off charging decision for t ∈ Ti , uev i is the  

utility for energy charged, ηiev is the charging efficiency, ρiev is · Kψωt + ψωkt xk xitev +
K u
vωkt xku
the rated charging power, E0i ev is the energy upon arrival, and k∈U k∈U

ev
Ēi is the maximum energy. DERs and loads may have single  1  g 
or multi-phase connections, but for ease of presentation it is − g ρjs Kψωt + ψωkt xk
K u
|j |
assumed that they are wye connected. iv is the set of node– j∈G g
ψ∈j k∈U
phase pairs which EV i is connected at (e.g., if connected at  g g
· xjs p̂jt ≤ vω for ω ∈ , t ∈ T , (7c)
phases a and b of node n then iv = {(n, a), (n, b)}).
s∈Sj
Each potential renewable generation site j ∈ G, has a set  
g
of sizes Sj at which generation can be installed, which deter- xjs ≤ 1 for j ∈ G, xku ≤ 1, (7d)
g g
mine the rated power ρjs and cost cjs , s ∈ Sj (discounted and s∈S k∈U
g
adjusted based on the duration of the optimisation horizon rel- xitev , xjs , xku ∈ {0, 1}. (7e)
ative to the lifetime). The decision to install generation of size
g g
s is indicated by xjs ∈ {0, 1}. It is assumed that the renewables The decision variables are xitev , xku , xjs , which are all binary
sources are non-dispatchable (e.g., solar or wind not controlled valued.
by the DSO during operation). Therefore, each source oper- The objective (7a) is to minimise the net system cost, which
ates with a normalised generation profile over the time horizon, includes the cost/revenue of buying/selling energy upstream,
g g g g g
p̂j = (p̂j1 , . . . , p̂jT ), so the total output power vector is ρjs p̂j . the utility obtained from EV charging, the cost of renew-
Using the linear multiphase power flow model from [9] able generation investment and the cost of network upgrades.
and a set of nominal operating points over the time horizon, Constraint (7b) limits the maximum energy levels of the EVs.
time dependent coefficients Aψωt can be obtained relating real Constraint (7c) limits the maximum and minimum voltage
power injections at node–phase pair ψ to the change in net magnitude of each node–phase pair ω ∈ . Constraints in (7d)
real power imports at the slack node. Similarly, a coefficient specify that a single installation size can be selected for each
Kψωt can be obtained which relates the impact of a power renewable generation site and that at most a single network
injection at node–phase pair ψ to the voltage magnitude of upgrade plan can be selected. Constraint (7e) specifies the
another node–phase pair ω. For node–phase pair ω, let ṽωt decision variables are binary.
be the voltage magnitude at the nominal operating point, with The problem must be reformulated as a QUBO for it to
upper and lower allowed limits vω and vω . be solved using quantum annealing. The objective function
The selection of network upgrade plan k ∈ U is indi- of the combinatorial OPF problem (7a) is made up of linear
cated by xku ∈ {0, 1}. Plans are mutually exclusive, each and quadratic terms of the binary decision variables and can
being associated with a specific set of upgrades to lines and therefore be directly incorporated into a QUBO formulation.
transformers, resulting in new power flow model coefficients, However, since constraints cannot be directly incorporated, the
Aψkt and Kψωkt . The discounted and time horizon adjusted objective must instead include equivalent penalty terms which
cost of upgrade k is cuk . Let Aψkt = Aψkt − Aψt and are high when constraints are violated and zero for feasible
K solutions. A general linear inequality constraint of form Ax ≤
ψωkt = Kψωkt − Kψωt . Also, for time t, let the impact of
upgrade k on the net import power at the nominal operat- b can enforced through an equivalent penalty term [54],
p
ing point be given by kt0 and let the impact on the voltage
2
magnitude at node–phase pair ω be given by vωkt . 
Y−1
s
P Ax − b + δ 2l yl , with δ = . (8)
The combinatorial optimisation problem can be formu- 2Y − 1
l=0
lated as

 1    P is a penalty scalar, which will ensure the constraint is satis-

min Aψt + ψkt xk
A u
fied at optimality if sufficiently large, although an overly large
|iev |
i∈V ψ∈i t∈Ti ev k∈U penalty can increase the solution time. Trade-offs and rules-
 1   of-thumb for penalty selection are discussed in [54]. Auxiliary
·τ λ0t ηiev ρiev xitev − g slack variables yl ∈ {0, 1}, l ∈ {0, . . . , Y − 1} are introduced to
|j |
j∈G ψ∈j t∈T
g
enforce inequality, as opposed to equality. These are arranged

  g g g
as a binary expansion, which efficiently enforces the constraint
Aψt + Aψkt xku xjs τ λ0t ρjs p̂jt with conservativeness no greater than δ, given that the required
k∈U s∈Sj slack quantity is no greater than s.

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
MORSTYN: ANNEALING-BASED QUANTUM COMPUTING FOR COMBINATORIAL OPTIMAL POWER FLOW 1097

For specific linear constraints, simpler equivalent penalty Equivalent penalty terms for the minimum voltage magni-
terms are available which make use of the properties of binary tude constraints in (7c) are given by
 for xi ∈ {0, 1}, i ∈ N = {1, . . . , N},
variables. In particular, ⎛
a linear constraint i∈N xi ≤ 1 can be enforced with the
  1  
Pv ⎝ṽωt − vω +
following penalty [54], |iev |
 t∈T ω∈ i∈V ψ∈iev t∈Ti
P xi xj (9)

 u,ev

i,j∈N ,i=j ρt Kψωt xit +
ev ev
ψωkt xkit
K
+ vωkt xku
k∈U k∈U
Due to the network upgrades, (7) also has quadratic inequal-

ity constraints. To integrate these into a QUBO, the following  1   g

 u,g
− g Kψωt xjs + K
ψωkt xkjs
penalty term is proposed to enforce the relationship z = xy |j |
j∈G g
ψ∈j s∈Sj k∈U
between x, y, z ∈ {0, 1},
Y v −1

2
P(xy − 2zx − 2zy + 3z) (10)  g g
 v
· ρjs p̂jt − δωt
v
2l yωt (14)
This penalty will be 0 if z = xy, and greater than or equal to t∈T l=0
P otherwise. v
yωtl ∈ {0, 1}, l ∈ {0, . . . , Y v − 1} are the auxiliary slack vari-
Using (8), the inequality constraints enforcing the EV max-
ables. Since the maximum and minimum voltage limits have
imum energy levels (7b) can be incorporated into the QUBO
similar characteristics, the same penalty scalar, number of aux-
problem using equivalent penalty terms given by
iliary variables, and conservativeness parameter are used for
⎛ ev −1
⎞2 both.
  Y
Pev ⎝E0iev
− Ēiev + τ ηiev ρiev xitev + δiev 2l yev
il
⎠ . (11) Finally, using (9), equivalent penalty terms for the con-
i∈V t∈Ti l=0 straints in (7d) specifying that a single installation size can be
selected for each renewable generation site, and that a single
Pev is the penalty scalar and yev il ∈ {0, 1}, l ∈ {0, . . . , Y − 1}
ev
network upgrade plan can be selected, are given by
are binary auxiliary slack variables which enforce the con-  g g 
straint as an inequality. δiev is the maximum conservativeness Pg xjs xjs + Pu xku xku , (15)
of the constraint. j∈G s,s ∈S ,s=s k,k ∈U ,k=k
The maximum voltage magnitude constraints for each node–
phase pair in (7c) can be similarly enforced using equivalent where Pg and Pu are the penalty scalars.
penalty terms given by Bringing together the original objective function (7a)
⎛ and the equivalent penalty terms for constraint enforcement
  1   (11)–(15), the proposed equivalent QUBO formulation for (7)
Pv ⎝ṽωt − vω + is given by
|i |
ev
t∈T ω∈ i∈V ψ∈iev t∈Ti

min (7a) + (11) + (12) + (13) + (14) + (15).
 u,ev

ρt Kψωt xit +
ev ev
ψωkt xkit
K
+ vωkt xku g
u,ev
k∈U k∈U
The QUBO problem has decision variables xitev , xku , xjs , xkit ,

u,g
xkjs , yev , yv , yv , which are all binary valued.
 1   g
 u,g
il ωtl ωtl
− g Kψωt xjs + ψωkt xkjs
K The proposed formulation focuses on DERs with discrete
|j | flexibility, which require combinatorial optimisation. The for-
j∈Gg
ψ∈j s∈Sj k∈U
v −1
2 mulation could be extended to also include continuously
 g g
Y
controllable DERs based on a small discrete control step-
· ρjs p̂jt + δωt
v
2l yvωtl . (12) size and continuous output powers approximated by multiple
t∈T l=0
binary decision variables (arranged in binary expansions). For
Pv is the penalty scalar, yvωtl ∈ {0, 1}, l ∈ {0, . . . , Y v − 1} are example, consider a controllable generation source with output
auxiliary slack variables which enforce the constraint as an power pct , limited by 0 ≤ pct ≤ p̄c . Let the control step-size
inequality, and δωt
v is the maximum constraint conservative- be δ c and the additional binary decision variables be xtm c ∈
u,ev u,g
ness. xkit and xkjs are binary auxiliary variables, which are {0, 1}, m ∈ {0, . . . , N −1}, where N = ceil{log
c c
N c −12 (p̄m /δ
c c )}. The
introduced to manage the quadratic terms in (7c). Using (10), DER’s output power would be pct = δ c m=0 2 xtm c , which
u,ev u,g g
the required relationships xkit = xku xitev and xkjs = xku xjs can could be incorporated into the objective and penalty terms of
be enforced by introducing additional penalty terms given by the QUBO problem.
   u,ev u,ev u,ev
Pu,ev xitev xku − 2xitev xkit − 2xku xkit − 3xkit
k∈U i∈V t∈Ti
IV. C ASE S TUDIES
   g g u,g u,g u,g
 In this section, case studies are presented demonstrating
+Pu,g xjs xku − 2xjs xkjs − 2xku xkjs − 3xkjs , (13)
the implementation of the proposed QUBO formulation. The
k∈U j∈G s∈S j
case studies make use the IEEE European Low Voltage Test
where Pu,ev and Pu,g are the penalty scalars. Feeder [55], shown in Fig. 3, with controllable on/off EV

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
1098 IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO. 2, MARCH 2023

TABLE I
A DDITIONAL C ASE S TUDY PARAMETERS

number of decision variables (EV charging decisions over the

time horizon, EV energy constraint slack variables, and aux-
Fig. 3. The distribution feeder used for the case studies, showing the location
iliary variables associated with network upgrades), and adds
of the main grid connection, 55 domestic loads, 30 of which may have EV new quadratic relationships between variables, which may also
charging, and 3 sites where PV generation can be installed. add qubits to enable minor embedding on the Pegasus topol-
ogy. Adding additional voltage constraints similarly increases
the number of decision variables and quadratic relationships.
charging, photovoltaic (PV) generation placement/sizing and When implemented with a single EV and voltage limits
network upgrade decisions. on 3 node–phase pairs without including network upgrades,
The network has 55 single phase domestic loads, and up to the QUBO formulation has 746 decision variables. This
30 EVs with single phase 7.2 kW chargers and batteries sizes increases to 792 when network upgrade decisions are included.
of 75, 85 or 100 kWh. The EV arrival and departure times are Given voltage limits on 3 node–phase pairs, without network
normally distributed. The mean arrival time is 6 pm and mean upgrades each additional EV adds on average 17 decision vari-
departure time is 8 am, both with a standard deviation of 2 h. ables to the problem, while with network upgrades extra EVs
The EVs’ energy upon arrival is uniformly distributed between add an average of 44 decision variables (note that the number
10% and 30%. The network has 3 PV generation sites which of extra decision variables varies depending on the number of
have three phase connections and can support 25 or 50 kWp. intervals each EV is available for charging). Given a single
Two network upgrade plans are available, which respectively EV, enforcing voltage limits at an additional node–phase pair
reduce the impedance of the transformers and lines by half increases the number of decision variables by 240, regardless
or three quarters. The case studies consider a 24 hour optimi- of whether or not network upgrades are considered.
sation horizon from 12 pm to 12 pm the next day, with 1 h Fig. 4a shows the number of assigned qubits for different
duration intervals. numbers of EVs, given voltage limits are enforced at 3 node–
Here, operation over a single day with 1 h intervals has been phase pairs, and Fig. 4b which shows the number of assigned
chosen since the focus is demonstrating the implementability qubits for a single EV and different numbers of node–phase
of the proposed QUBO formulation, but it could be refined pairs with voltage limits. For each problem size, the range of
by considering multiple representative days and more granu- assigned qubits provided by D-Wave’s heuristic minor embed-
lar scheduling (at the cost of increased computational burden). ding tool for 20 runs is shown, varying from the mean by
The upstream energy price is assumed to follow a standard at most ±9.4%. Without including network constraints, the
U.K. ‘Economy 7’ tariff, with an energy price of £0.15 / number of assigned qubits increases approximately linearly
kWh from 6 am to 11 pm and £0.07 / kWh from 11 pm up to 9 EVs and 8 node–phase pairs with voltage limits,
to 6 am [56]. Smart meter data from the U.K. Customer- beyond which feasible embeddings are not regularly found.
Led Network Revolution project is used for the domestic A similar relationship is seen with network upgrades, but fea-
loads [57]. Scale-dependent PV installation costs are calcu- sible embeddings are not regularly found above 5 EVs and 7
lated using data from [58], assuming a 25-year lifetime and node–phase pairs with voltage limits. Linear scaling indicates
5% discount rate. The network upgrade costs are calculated that the proposed formulation should meaningfully benefit as
using data from [59], assuming a 35-year lifetime and 5% larger annealing-based quantum processors become available.
discount rate. The voltage magnitude limits are assumed to However, the current maximum problem size is restrictive even
be 0.95 and 1.05 pu. Additional case study parameters are for relatively small-scale applications.
provided in Table I. To investigate the performance of quantum annealing com-
pared with classical computing, the proposed QUBO problem
A. Quantum Processor Implementation was solved using D-Wave’s Advantage processor and with
The proposed QUBO problem is implemented on D-Wave’s simulated annealing, which is a state-of-the-art classical
5,760 qubit Advantage quantum processor to investigate how method for combinatorial optimization problems [60]. For
the number of required qubits scales with the problem size. quantum annealing, the problem was solved using 100 samples
To do this, the problem is implemented for different numbers and an annealing time of 100 μs. For simulated annealing, the
of EVs and node–phase pairs where the upper and lower volt- problem was implemented using the dwave-neal Python pack-
age limits are enforced. Adding additional EVs increases the age [61], and solved using a 2.3 GHz 8-core Intel Core i9

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
MORSTYN: ANNEALING-BASED QUANTUM COMPUTING FOR COMBINATORIAL OPTIMAL POWER FLOW 1099

TABLE II
C OMPARISON B ETWEEN THE AVERAGE C OMPUTATION T IME , AVERAGE
N ET U TILITY AND AVERAGE QUBO E NERGY, W HEN THE QUBO
P ROBLEM I S S OLVED 10 T IMES U SING S IMULATED A NNEALING (SA)
AND Q UANTUM A NNEALING (QA). T HE QUBO P ROBLEM WAS
F ORMULATED W ITH D IFFERENT N UMBERS OF EV S , VOLTAGE L IMITS AT
3 N ODE –P HASE PAIRS AND W ITHOUT N ETWORK U PGRADE D ECISIONS

Fig. 4. The number of assigned qubits on D-Wave’s Pegasus topology for

Fig. 5. The net real import power for case studies with 30 EVs, with and
(a) different numbers of EVs, given voltage limits at 3 node–phase pairs, and
without network upgrade decisions, using D-Wave’s hybrid binary quadratic
(b) voltage limits enforced at different numbers of node–phase pairs, given a
solver with a time limit of 120 s.
single EV, with and without network upgrade decisions.

processor and 16 GB of RAM. To allow comparison between

the computation time of the two methods, simulated annealing
was implemented with one repetition and two sweeps, which
was found to result in solutions of similar energy (i.e., the
QUBO objective including penalty terms). Table II compares
the average computation time, average net utility and average
energy of the two methods, when the problem is formulated
with different numbers of EVs, voltage limits at 3 node–phase
pairs and without considering network upgrade decisions. The
Fig. 6. The average EV energy for case studies with 30 EVs, with and
averages were obtained by solving each formulation 10 times without network upgrade decisions, using D-Wave’s hybrid binary quadratic
using each method. As shown, for a single EV, quantum solver with a time limit of 120 s.
annealing has a lower average computation time, higher aver-
age net utility and lower average energy. Also, under quantum
annealing, the computation time only increases slightly from demand and network voltages. Upper and lower voltage limits
49.9 ms with 1 EV, to 52.2 ms with 9 EVs, while the compu- are specified for 12 node–phase pairs spread throughout the
tation time for simulated annealing increases from 56.4 ms to network. Case studies with and without the potential for
121.5 ms. The average net utility is also consistently higher network upgrades are compared to show the additional value
using quantum annealing. Each method yields similar average of co-optimising operational flexibility and network investment
QUBO energies, driven mainly by slight constraint violations. decisions. The hybrid solvers are heuristic, iteratively making
However, for simulated annealing the average energy is more use of both classical and quantum computation for a specified
consistent for different problem sizes, while using quantum time limit, where upon the solution giving the lowest total
annealing there is an upward trend with increasing problem energy (QUBO objective) is returned.
size, which could indicate a future scaling challenge. First, case studies with and without network upgrades are
completed with a time limit of 120 s. Fig. 5 shows the net
power imported from the main grid for each case, and Fig. 6
B. Hybrid Solver Implementation shows the average energy of the 30 EVs. Fig. 7 shows the
D-Wave’s hybrid quantum-classical binary quadratic model range of the voltage magnitudes across the three phases. In
solver allows larger problems of practical interest to be solved. the case where network upgrades are allowed, the solution
To demonstrate this, case studies are presented with 30 EVs, returned by the hybrid solver specifies the moderate network
which together have a significant impact on local power upgrade plan (reducing impedances by half) and 100 kW of

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
1100 IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO. 2, MARCH 2023

Fig. 8. Results from solving the QUBO problem 30 times for a range hybrid
solver time limits, with and without network upgrade decisions. Distributions
are shown for (a) the QUBO energy and (b) net utility values. The box
Fig. 7. The voltage magnitude range across the nodes for each phase for plots show the median (centre line), interquartile range (box), 1.5 times the
case studies with 30 EVs, using D-Wave’s hybrid binary quadratic solver interquartile range above/below the box (whiskers), and outliers (circles).
with a time limit of 120 s. The voltage limits are 0.95 pu and 1.05 pu. The
voltage magnitude ranges are shown for (a) the case without network upgrade
decisions and (b) the case with network upgrade decisions.
from slight improvements in constraint satisfaction, and from
Fig. 8b, it can be seen that the average net utility is fairly
PV generation (split between the three potential sites). Without stable above 30 s. The heuristic nature of the hybrid solver is
network upgrades, only 50 kW of PV is installed, at the site clear from the variability of the net utility, but it can be seen
closest to the main grid. that even accounting for outliers, the net utility for cases where
As shown in Fig. 5, allowing network upgrades results in network upgrades are allowed is consistently higher than when
greater maximum import and export of power, and from Fig. 6 they are disallowed.
it can be seen that the EVs reach a higher final average energy
level (78.6% compared with 56.8%). With network upgrades V. C ONCLUSION
enabled, the overall net utility over the day is £452, which Annealing-based quantum computing offers a new comput-
is 78% higher than the net utility without network upgrades ing hardware platform with the future potential to efficiently
(£254). For the node–phase pairs where the voltage limits are solve large-scale combinatorial optimisation problems. This
explicitly enforced, the lowest voltage magnitude reached is could be highly valuable for the power sector, particularly
0.958 pu for the case without network upgrades, and 0.951 for for network operators aiming to integrate DER flexibility into
the case with network upgrades. As seen in Fig. 7(a) and 7(b), network planning and investment decision making. To demon-
there are slight violations at nodes where the limits are not strate this opportunity, a novel QUBO formulation which can
enforced, with 0.947 pu the lowest voltage magnitude reached be solved with quantum annealing was developed for a lin-
in both cases. ear multiphase OPF problem, with controllable on/off EV
Next, to show the impact of the hybrid solver time limits, the charging, renewable generation placement/sizing and network
QUBO problem is solved 30 times for a range limits between upgrade decisions. Case studies based on the IEEE European
15 s and 120 s, with and without network upgrade decisions. Low Voltage Test Feeder were implemented on D-Wave’s
Note that D-Wave’s hybrid solver imposes a minimum time 5,760 qubit Advantage quantum processor to show how
limit based on its assessment of the problem complexity, which the problem size impacts the required number of qubits.
varied around 12 s, so 15 s was selected as a consistent start- Although it was found that the quantum processor is too small
ing point. The box plots in Fig. 8a and Fig. 8b show the for distribution-scale applications, the number of qubits was
distributions of the QUBO energy (returned objective func- observed to grow linearly with the number of EVs and the
tion value) and net utility. As shown in Fig. 8a, increasing the number of network voltage constraints, indicating that there
time limit reduces the QUBO energy, with diminishing returns is a promising future opportunity given the rate of techno-
starting to be seen above 60 s. Most of this reduction results logical development of annealing-based quantum processors.

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
MORSTYN: ANNEALING-BASED QUANTUM COMPUTING FOR COMBINATORIAL OPTIMAL POWER FLOW 1101

D-Wave’s hybrid quantum-classical binary quadratic model [16] D. Apostolopoulou, S. Bahramirad, and A. Khodaei, “The interface
solver was also used to solve larger case studies with 30 of power: Moving toward distribution system operators,” IEEE Power
Energy Mag., vol. 14, no. 3, pp. 46–51, Jun. 2016.
EVs, where EV charging flexibility has a significant impact [17] P. S. Georgilakis and N. D. Hatziargyriou, “Optimal distributed gener-
on distribution network power flows. In this case, combina- ation placement in power distribution networks: Models, methods, and
torial co-optimisation of EV flexibility with generation and future research,” IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3420–3428,
Aug. 2013.
network investment decisions was shown to offer substantial [18] P. You, Z. Yang, M.-Y. Chow, and Y. Sun, “Optimal cooperative charging
value. The paper has focused on deterministic combinatorial strategy for a smart charging station of electric vehicles,” IEEE Trans.
OPF as a first step, but in practice network planning and Power Syst., vol. 31, no. 4, pp. 2946–2956, Jul. 2016.
[19] L. Liberti, “Undecidability and hardness in mixed-integer nonlinear
scheduling applications may involve significant uncertainty programming,” RAIRO Oper. Res., vol. 53, no. 1, pp. 81–109, Jan. 2019.
due to the weather-dependence of renewable generation and [20] X. Zhao, P. B. Luh, and J. Wang, “Surrogate gradient algorithm
the behaviour-dependence of flexible loads. An important area for lagrangian relaxation,” J. Optim. Theory Appl., vol. 100, no. 3,
pp. 699–712, Mar. 1999.
for future work is to investigate how methods for robust opti- [21] M. L. Fisher, “The Lagrangian relaxation method for solving integer
misation (see, e.g., [62]) can be implemented within the qubit programming problems,” Manage. Sci., vol. 50, no. 12S, pp. 1861–1871,
limitations of quantum annealers. Another important area for Dec. 2004.
[22] S. Zhou et al., “The combinatorial optimization by genetic algorithm
future work is the optimal selection of penalty terms, which and neural network for energy storage system in solar energy elec-
are necessary for constraint handling within the proposed for- tric vehicle”, Proc. World Congr. Intell. Control Autom. (WCICA),
mulation, but may affect the computation time and solution pp. 2838–2842, Jun. 2008, doi: 10.1109/WCICA.2008.4593375.
[23] J. Soares, Z. Vale, B. Canizes, and H. Morais, “Multi-objective parallel
quality if chosen inappropriately. particle swarm optimization for day-ahead vehicle-to-grid schedul-
ing,” in Proc. IEEE Symp. Comput. Intell. Appl. Smart Grid, 2013,
R EFERENCES pp. 138–145.
[24] H. Mori and S. Sudo, “Strategic tabu search for unit commitment
[1] L. N. Ochoa, F. Pilo, A. Keane, P. Cuffe, and G. Pisano, “Embracing an in power systems,” IFAC Proc. Vol., vol. 36, no. 20, pp. 485–490,
adaptable, flexible posture: Ensuring that future European distribution Sep. 2003.
networks are ready for more active roles,” IEEE Power Energy Mag., [25] N. Deeb, “Simulated annealing in power systems,” in Proc. IEEE Int.
vol. 14, no. 5, pp. 16–28, Sep./Oct. 2016. Conf. Syst. Man Cybern., Jan. 1992, pp. 1086–1089.
[2] A. Dimeas et al., “Smart houses in the smart grid: Developing an [26] S. Chen, J. Montgomery, and A. Bolufé-Röhler, “Measuring the curse of
interactive network,” IEEE Electrific. Mag., vol. 2, no. 1, pp. 81–93, dimensionality and its effects on particle swarm optimization and differ-
Mar. 2014. ential evolution,” Appl. Intell., vol. 42, no. 3, pp. 514–526, Apr. 2015.
[3] F. Alexander et al., “Exascale applications: Skin in the game,” Philos. [27] F. Arute et al., “Quantum supremacy using a programmable supercon-
Trans. Royal Soc. A, Math. Phys. Eng. Sci., vol. 378, no. 2166, ducting processor,” Nature, vol. 574, no. 7779, pp. 505–510, Oct. 2019.
Mar. 2020, Art. no. 20190056. [28] F. Feng, P. Zhang, M. A. Bragin, and Y. Zhou, “Novel resolution
[4] H. W. Dommel and W. F. Tinney, “Optimal power flow solutions,” IEEE of unit commitment problems through quantum surrogate Lagrangian
Trans. Power App. Syst., vol. PAS-87, no. 10, pp. 1866–1876, Oct. 1968. relaxation,” IEEE Trans. Power Syst., early access, Jun. 10, 2022,
[5] J. Momoh, R. Adapa, and M. El-Hawary, “A review of selected optimal doi: 10.1109/TPWRS.2022.3181221.
power flow literature to 1993. I. Nonlinear and quadratic program- [29] N. Nikmehr, P. Zhang, and M. A. Bragin, “Quantum distributed unit
ming approaches,” IEEE Trans. Power Syst., vol. 14, no. 1, pp. 96–104, commitment,” IEEE Trans. Power Syst., vol. 37, no. 5, pp. 3592–3603,
Feb. 1999. Sep. 2022.
[6] J. Lavaei and S. H. Low, “Zero duality gap in optimal power flow [30] J. Preskill, “Quantum computing in the NISQ era and beyond,”
problem,” IEEE Trans. Power Syst., vol. 27, no. 1, pp. 92–107, Quantum, vol. 2, p. 79, Aug. 2018.
Feb. 2012. [31] P. Hauke, H. G. Katzgraber, W. Lechner, H. Nishimori, and W. D. Oliver,
[7] S. Gill, I. Kockar, and G. W. Ault, “Dynamic optimal power flow for “Perspectives of quantum annealing: Methods and implementations,”
active distribution networks,” IEEE Trans. Power Syst., vol. 29, no. 1, Rep. Progr. Phys., vol. 83, no. 5, May 2020, Art. no. 54401.
pp. 121–131, Jan. 2014. [32] S. E. Venegas-Andraca, W. Cruz-Santos, C. McGeoch, and
[8] L. Gan, N. Li, U. Topcu, and S. H. Low, “Exact convex relaxation of M. Lanzagorta, “A cross-disciplinary introduction to quantum
optimal power flow in radial networks,” IEEE Trans. Autom. Control, annealing-based algorithms,” Contemp. Phys., vol. 59, no. 2,
vol. 60, no. 1, pp. 72–87, Jan. 2015. pp. 174–197, 2018.
[9] A. Bernstein, C. Wang, E. DallAnese, J.-Y. Le Boudec, and C. Zhao, [33] T. Kadowaki and H. Nishimori, “Quantum annealing in the transverse
“Load Flow in multiphase distribution networks: Existence, uniqueness, ising model,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. Interdiscip.
non-singularity and linear models,” IEEE Trans. Power Syst., vol. 33, Top., vol. 58, no. 5, pp. 5355–5363, Nov. 1998.
no. 6, pp. 5832–5843, Nov. 2018. [34] C. C. McGeoch, “Theory versus practice in annealing-based quantum
[10] M. Khonji, S. C. K. Chau, and K. Elbassioni, “Combinatorial computing,” Theor. Comput. Sci., vol. 816, pp. 169–183, Jan. 2020.
optimization of AC optimal power flow with discrete demands in radial [35] T. Albash and D. A. Lidar, “Demonstration of a scaling advantage for a
networks,” IEEE Trans. Control Netw. Syst., vol. 7, no. 2, pp. 887–898, quantum annealer over simulated annealing,” Phys. Rev. X, vol. 8, no. 3,
Jun. 2020. Jul. 2018, Art. no. 31016. [Online]. Available: https://link.aps.org/doi/
[11] D. Ernst, M. Glavic, G. Stan, S. Mannor, and L. Wehenkel, “The cross- 10.1103/PhysRevX.8.031016
entropy method for power system combinatorial optimization problems,” [36] A. D. King et al., “Scaling advantage over path-integral Monte Carlo
in Proc. IEEE Lausanne Power Tech, Jul. 2007, pp. 1290–1295. in quantum simulation of geometrically frustrated magnets,” Nat.
[12] B. Sun, Z. Huang, X. Tan, and D. H. K. Tsang, “Optimal scheduling Commun., vol. 12, no. 1, pp. 1–6, Feb. 2021.
for electric vehicle charging with discrete charging levels in distribution [37] E. J. Crosson and D. A. Lidar, “Prospects for quantum enhancement with
grid,” IEEE Trans. Smart Grid, vol. 9, no. 2, pp. 624–634, Mar. 2018. diabatic quantum annealing,” Nat. Rev. Phys., vol. 3, no. 7, pp. 466–489,
[13] M. Zhang, Q. Wu, T. B. H. Rasmussen, X. Yang, and J. Wen, “Heat Jul. 2021.
pumps in Denmark: Current situation of providing frequency con- [38] A. Perdomo-Ortiz, N. Dickson, M. Drew-Brook, G. Rose, and
trol ancillary services,” CSEE J. Power Energy Syst., vol. 8, no. 3, A. Aspuru-Guzik, “Finding low-energy conformations of lattice protein
pp. 769–779, May 2022. models by quantum annealing,” Sci. Rep., vol. 2, pp. 1–7, Aug. 2012.
[14] D. Papadaskalopoulos and G. Strbac, “Nonlinear and randomized pricing [39] J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and
for distributed management of flexible loads,” IEEE Trans. Smart Grid, S. Lloyd, “Quantum machine learning,” Nature, vol. 549, no. 7671,
vol. 7, no. 2, pp. 1137–1146, Mar. 2016. pp. 195–202, 2017.
[15] C. Crozier, M. Deakin, T. Morstyn, and M. McCulloch, “Incorporating [40] M. Kim, D. Venturelli, and K. Jamieson, “Leveraging quantum anneal-
charger efficiency into electric vehicle charging optimization,” in Proc. ing for large MIMO processing in centralized radio access networks,”
IEEE PES Innov. Smart Grid Technol. Europe (ISGT-Europe), Sep. 2019, in Proc. Conf. ACM Special Interest Group Data Commun., 2019,
pp. 1–5. pp. 241–255.

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.
1102 IEEE TRANSACTIONS ON SMART GRID, VOL. 14, NO. 2, MARCH 2023

[41] R. Eskandarpour, K. J. B. Ghosh, A. Khodaei, A. Paaso, and L. Zhang, [54] F. Glover, G. Kochenberger, and Y. Du, “Quantum bridge analytics I: A
“Quantum-enhanced grid of the future: A primer,” IEEE Access, vol. 8, tutorial on formulating and using QUBO models,” 4OR, vol. 17, no. 4,
pp. 188993–189002, 2020. pp. 335–371, Dec. 2019.
[42] A. Ajagekar and F. You, “Quantum computing for energy systems [55] “European low voltage test feeder.” Accessed: Feb. 1, 2022. [Online].
optimization: Challenges and opportunities,” Energy, vol. 179, no. 607, Available: cmte.ieee.org/pes-testfeeders
pp. 76–89, Jul. 2019. [56] “A complete guide to economy 7 and how it works.” Accessed:
[43] D. Wang, K. Zheng, Q. Chen, Z. Li, and S. Liu, “Quantum annealing Feb. 1, 2022. [Online]. Available: uswitch.com/gas-electricity/guides/
computing for grid partition in large-scale power systems,” in Proc. IEEE economy-7
5th Int. Electr. Energy Conf. (CIEEC), 2022, pp. 2004–2009. [57] “Enhanced profiling of domestic customers with solar photovoltaics.”
[44] E. B. Jones et al., “On the computational viability of quantum Customer-Led Network Revolution. Accessed: Feb. 1, 2022. [Online].
optimization for PMU placement,” in Proc. IEEE Power Energy Soc. Available: networkrevolution.co.uk
Gen. Meeting (PESGM), 2020, pp. 1–5. [58] D. Feldman, V. Ramasamy, R. Fu, A. Ramdas, J. Desai, and R. Margolis,
[45] M. Kjaergaard et al., “Superconducting qubits: Current state of play,” “U.S. solar photovoltaic system and energy storage cost benchmark: Q1
Annu. Rev. Condens. Matter Phys., vol. 11, no. 1, pp. 369–395, 2020,” Nat. Renew. Energy Lab., Golden, CO, USA, Rep. NREL/TP-
Mar. 2020. 6A20-77324, Sep. 2021.
[46] R. Harris et al., “Experimental demonstration of a robust and scal- [59] D. Cartlidge, Spon’s Mechanical and Electrical Services Price Book.
able flux qubit,” Phys. Rev. B, Condens. Matter Condens. Matter Mater. Boca Raton, FL, USA: CRC Press, 2018.
Phys., vol. 81, no. 13, pp. 1–20, Apr. 2010. [60] S. V. Isakov, I. N. Zintchenko, T. F. Rønnow, and M. Troyer, “Optimised
[47] K. Boothby, P. Bunyk, J. Raymond, and A. Roy, “Next-generation topol- simulated annealing for Ising spin glasses,” Comput. Phys. Commun.,
ogy of D-wave quantum processors,” D-Wave Technical Report Series, vol. 192, pp. 265–271, Jul. 2015.
D-Wave Syst., Burnaby, BC, Canada, Rep. 14-1026A-C, 2020. [Online]. [61] “Dwave-neal documentation.” D-Wave Systems. 2022. [Online].
Available: dwavesys.com/learn/publications Available: docs.ocean.dwavesys.com
[48] C. McGeoch and P. Farré, “The D-Wave advantage system: [62] D. Bertsimas, D. B. Brown, and C. Caramanis, “Theory and applica-
An overview,” D-Wave Technical Report Series, D-Wave Syst., tions of robust optimization,” SIAM Rev., vol. 53, no. 3, pp. 464–501,
Burnaby, BC, Canada, Rep. 14-1049A-A, 2020. [Online]. Available: 2011.
dwavesys.com/learn/publications
[49] Y. Sugie et al., “Minor-embedding heuristics for large-scale annealing
processors with sparse hardware graphs of up to 102,400 nodes,” Soft
Comput., vol. 25, no. 3, pp. 1731–1749, 2021. Thomas Morstyn (Senior Member, IEEE) received
[50] W. Bernoudy, C. Mcgeoch, and P. Farr, “D-Wave hybrid solver service the B.Eng. degree (Hon.) in electrical engineering
+ advantage: Technology update,” D-Wave Technical Report Series, D- from the University of Melbourne in 2011, and
Wave Syst., Burnaby, BC, Canada, Rep. 14-1048A-A, 2020. [Online]. the Ph.D. degree in electrical engineering from the
Available: dwavesys.com/learn/publications University of New South Wales in 2016.
[51] B. Stott, J. Jardim, and O. Alsaç, “DC power flow revisited,” IEEE He is a Lecturer of Power Electronics and
Trans. Power Syst., vol. 24, no. 3, pp. 1290–1300, Aug. 2009. Smart Grids with the School of Engineering,
[52] L. Barth, N. Ludwig, E. Mengelkamp, and P. Staudt, “A comprehen- University of Edinburgh. He is also the Deputy
sive modelling framework for demand side flexibility in smart grids,” Champion of Energy Distribution and Infrastructure
Comput. Sci. Res. Develop., vol. 33, nos. 1–2, pp. 13–23, Feb. 2018. for the Scottish Energy Technology Partnership. His
[53] C. Huang, C. Wang, N. Xie, and Y. Wang„ “Robust coordination research interests include multi-agent control and
expansion planning for active distribution network in deregulated retail market design for integrating distributed energy resources into power system
power market,” IEEE Trans. Smart Grid, vol. 11, no. 2, pp. 1476–1488, operations. He is an Associate Editor of IEEE T RANSACTIONS ON P OWER
Mar. 2020. S YSTEMS.

Authorized licensed use limited to: Tsinghua University. Downloaded on March 02,2023 at 06:08:52 UTC from IEEE Xplore. Restrictions apply.