Professional Documents
Culture Documents
An Evaluation of A New Pricing Technique To Integrate Wind Energy Using Two Timescales Scheduling
An Evaluation of A New Pricing Technique To Integrate Wind Energy Using Two Timescales Scheduling
An Evaluation of A New Pricing Technique To Integrate Wind Energy Using Two Timescales Scheduling
Smart Grids
using two
timescales
Mutaz Tuffaha
Supervisor: Roger Petterson
Linnaeus University-Vaxjo
Part I
Pricing in
Smart Grids
using two This thesis is based on the article: Multiple Timescale Dispatch
timescales
and Scheduling for Stochastic Reliability in Smart Grids with
Wind Generation Integration. By M. He, S. Murugesan and J.
Zhang. arXiv:1008.3932v2 [cs.SY], 13 Sep 2010.
My objectives are to:
present this model with detailed proofs of the main
results.
demonstrate its performance by simple simulations and
experiments.
study the influence of the increase of the wind energy on
the profits.
discuss some of the advantages and drawbacks of this
model
Flow chart
Pricing in
Smart Grids
using two
timescales
Model description
Pricing in
Smart Grids
using two The target is to maximize the profits of the Energy
timescales
Management System (EMS) which acts as an agent
between the energy producing company and the end user
by exploiting the wind energy.
The users are classified according to their nature into:
a Traditional users who do not have smart meters or
appliances.
b Opportunistic users who have smart meters or appliances
which provide 2-way communication channel with the
(EMS). They also are divided into persistent and
non-persistent.
The (EMS) is supposed to declare the real-time price for
the opportunistic users and the day-ahead dispatch
schedule for the energy producing company and price for
traditional users.
Costs
Pricing in
Smart Grids
using two
timescales
Pricing in
Smart Grids
using two
timescales
The day is divided into M T1 -slots. One day ahead the (EMS)
has to decide for each T1 -slot:
1 S: The base-load energy generation scheduled.
2 u: The day-ahead retail price per unit energy assumed to
be u ≤ ucap .
Real-time processes
Pricing in
Smart Grids
using two
timescales
Pricing in
Smart Grids
using two
timescales
Wind model
Pricing in
Smart Grids
using two
timescales
Pricing in
Smart Grids
using two
timescales
Dt = E[Dt ] + t = αt u γt + t ,
where:
t : accounts for the the uncertainty of Dt which can be
modeled as white noise t ∼ N (0, σ2t )
αt : a normalizing constant (which is not defined in the
article).
γt : the price elasticity at the corresponding T1 -slot and it
is defined by:
Pricing in
Smart Grids
The demand of the opportunistic users Do is modeled by a
using two
timescales
Gaussian distribution N (qo (v ), σo2 (v )) with:
qo (v ) = κ1 αo v γo Eo
σo2 (v ) = κ1 αo v γo Eo2 ,
where:
κ1 = λo T2 : The mean number of users arriving to the grid
in a T2 -slot.
Eo : The amount of energy consumed by each opportunistic
user in a T2 -slot, it is assumed to be constant.
−γo
αo : a normalizing constant given by vmin .
γo : the price elasticity at the corresponding T2 -slot and it
is defined by:
dE[Do ]/E[Do ]
γo = .
dv /v
Pricing in
Smart Grids
using two
timescales
Part II
Pricing in
Smart Grids
using two
timescales
The target is to maximize the profit of the (EMS) which can
be formulated as:
RT
Pnon-pst : max R l (ψ l , s, u, v ),
v
with
where
= W + s − (Dt + Do ),
Real-time Problem II
Pricing in
Smart Grids
using two
timescales
Pricing in
Smart Grids
Wind generation is not sufficient to meet the total energy
using two
timescales
demand in the system. Under this assumption χ{A} = 0, and
hence χ{B} + χ{C } = 1. Thus the solution would be:
v ∗ = vcap ,
when the opportunistic user is relatively inelastic, i.e.,
−1 ≤ γo < 0. And:
γo (c1 −cp ) γ (c1 −cp )
1+γo if Y > qo ( o 1+γ o
)
v∗ = γo c2
1+γo
γo c2
if Y < qo ( 1+γ ) ,
o
−1
qo (Y ) otherwise
when the opportunistic user is relatively elastic, i.e., γo < −1,
where:
Y = s + W − Dt .
Day-Ahead Problem
Pricing in
Smart Grids
using two
timescales The target is to maximize the profit in a T1 -slot which can be
formulated as:
K
X
DA
Pnon-pst : max EWk EuDt max[R l (ψkl , s, u, vk )].
S,u k vk
k=1
Pricing in
Smart Grids
using two
timescales
ucap if − 1 ≤ γt < 0
∗
u =
γt
1+γt c1 if γt < −1
∗
S = arg max{(c2 − c1 )S + K EvW ,D [(v − c2 )Do
S
−cχ{B}(S − K (αt u ∗γt + t + Do − W ))]},
Part III
Pricing in
Smart Grids
using two
timescales
I assumed a vector of means of wind generation for 24 hours as
follows:
Pricing in
Smart Grids
using two
timescales
Simulation the day-ahead price
Pricing in
Smart Grids I assumed the following values for the parameters used:
using two
timescales
ucap = 100, c1 = 60, c2 = 75, cp = 40.
Pricing in
Smart Grids
The day-ahead prices where found and hence the traditional
using two
timescales
users’ demands could be estimated by the following time series:
Design of the experiment
Pricing in
Smart Grids
using two
timescales
An experiment was designed to study the influence of the
increase of the wind generation on the dynamics of this model
as follows: Two time intervals were considered:
1 02:00-03:00 during which the users can be considered
elastic (or sensitive) to prices because it can be considered
as an off-peak hour.
2 19:00-20:00 during which the users can be considered
inelastic (or insensitive) to prices because it can be
considered as a peak hour.
The mean wind energy was increased gradually.
Values
Pricing in
Smart Grids
using two
timescales
Pricing in
Smart Grids
using two
timescales 1 The scheduled dispatch S was calculated for each mean wind
energy using the values above.
2 The profits of the (EMS) were calculated as follows:
R = Gt + Go − C ,
K
X K
X
Gt = uk∗ Dtk , Go = vk∗ Dok
k=1 k=1
and
K
X c1 s ∗ if Yk − Dok ≥ 0
C= .
c1 s ∗ + c2 (Dtk − Yk ) if Yk − Dok < 0
k=1
Results during an off-peak hour
Pricing in
Smart Grids
using two 2:00-3:00 (γo = −1.1, γt = −1.1)
timescales
W (MWh) S ∗ (MWh) R(e)
0.5 2.999 1807.1
0.55 2.9360 1827.4
0.6 2.8719 1828.3
0.7 2.7433 1833.1
0.8 2.6143 1845.8
0.9 2.4850 1859.1
1.0 2.3556 1845.1
1.1 2.2260 1858.5
1.3 1.9665 1862.1
1.5 1.7068 1856.9
1.7 1.4469 1903.2
Pricing in
Smart Grids
using two
timescales
Suggested solution
Pricing in
Smart Grids
using two
timescales
γt
ucap = c1 1+γcap
t cap
⇐⇒ ucap + γtcap ucap = c1 γtcap
⇐⇒ γtcap (c1 − ucap ) = ucap
ucap
⇐⇒ γtcap = c1 −ucap
Pricing in
Smart Grids 19:00-20:00 (γo = −0.3, γt = −0.3)
using two
timescales W (MWh) S ∗ (MWh) R(e)
0.5 7.2329 333.4
0.55 7.1690 338.3
0.6 7.1049 330.7
0.7 6.9764 346.7
0.8 6.8475 352.0
0.9 6.7183 358.5
1.0 6.5889 367.0
1.1 6.4593 372.8
1.3 6.1999 376.1
1.5 5.9402 390.6
1.7 5.6804 397.6
1.9 5.4204 411.5
2.1 5.1603 419.0
Incremental profits
Pricing in
Smart Grids
using two
timescales To study the effect of the increase of the wind energy on the
profits, I suggested calculating the change of profits with
respect to change of the wind energy as follows:
K
X K
X K
X
∆R = v ∗ ∆Dok +c1 norminv(cp /c) ∆σZ ,k +c1 ∆θm,k ,
k=1 k=1 k=1
Pricing in
Smart Grids
using two
From the experiments above I could conclude the following:
timescales
1 The prices in the off-peak hours increase drastically due to
the behavior of the price function when the elasticity
γo,t → −1− and I suggested a solution for that problem.
2 The profits increase with the wind generation and with the
increase of the opportunistic users’ demands which can be
increased by increasing the price elasticity γt , i.e., by
encouraging the opportunistic users to increase their
consumption when the wind energy increases by lowering
the price.
3 Compared with the day-ahead schedules this model may
guarantee more profits for the (EMS) because the
real-time price is designed to discourage the opportunistic
users to use electricity during the peak hours.
Further reading
Pricing in
Smart Grids
using two
timescales For further reading the reader is advised to read:
M. He, S. Murugesan and J. Zhang. Multiple Timescale
Dispatch and Scheduling for Stochastic Reliability in Smart
Grids with Wind Generation Integration. arXiv:1008.3932v2
[cs.SY], 13 Sep 2010.
M. Roozbehani, M. Dahleh and S. Mitter. Dynamic Pricing
and Stabilization of Supply and Demand in Modern
Electric Power Grids. IEEE 978-1-4244-6511-8, 2005.
S. Caron and G. Kesidis. Incentive-based Energy
Consumption Scheduling Algorithms for the Smart Grid.
IEEE 978-1-4244-6511-8, 2010.