You are on page 1of 6

1

Operation Optimization for a


Virtual Power Plant
M. Amin Salmani, S.M.Moghaddas Tafreshi , and Hosein Salmani

 Pexi Total active powers exported to ith inter connected


Abstract-- In the 21st century the concept of electrical power
system or imported from ith inter connected
supply is morphing gradually from centralized into decentralized
system, indicated by increasing the Installation of distributed system (if Pexi was negative) to the VPP area;
generation on the main grid. With emerging of advanced
Communication and information technology, the aggregation of U exi Price of imported or exported energy from ith
several small size, environmentally compatible, generators close to inters connected system;
the loads can be developed as virtual power plant in order to Cuk Penalty factor for un-served energy for loads;
provide added-value to the electric Power system.
In this paper an operation optimization algorithm is proposed to Puk Total un-served energy for loads;
integrate two DGs and a micro grid into a VPP control area,
containing several loads, which will be able to generate and sell PDGI Total active power purchased from DGI;
electrical energy to loads and power market. Suitable models are
proposed to simulate electric production, and the energy exchange
PMaxDGI Maximum active power that DGI is offered
between VPP and Grid. The optimization problem is here to provide;
formulated as a non-liner maximization algorithm with U DGI Active power price purchased from DGI;
constraints, where the objective function is profit of the VPP with
supply of electric energy for all the loads. PDGDC Total active power obtained from the DGDC;
Solving the VPP optimization problem in two different
scenarios with GAMS software will complete the paper. C DGDC Producing cost of DGDC;
Pmic Total active power purchased from micro grid;
Index Terms-- Distributed generation, Operational
optimization, Virtual Power Plant, Profit maximization, Control PMaxmic Maximum active power that micro grid is
Coordination Center, and Local Controller. Offered to provide;

I. NOMENCLATURE U mic Active power price purchased form micro grid;

PLoad 1 Total active power provided for sensitive loads;


U Load 1 Active power price provided for sensitive loads; II. INTRODUCTION
The VPP uses the billing system of the Local Distribution
PLoad 2 Total active power provided for non-sensitive Company (LDC) which is defined as low voltage, retail
loads; distribution, to aggregate the electrical output of several
U Load 2 Active power price provided for non-sensitive distributed generators located close to the loads so that the
VPP appears to investors as one unit power plant, rather than
loads;
several individually financed appliances [1, 2].
The primary objective of operation optimization algorithm can
vary. For example, economic optimization can either aim to
minimization the cost of producing energy or maximizing the
profit of the VPP [3, 4].
In this paper operational optimization is mainly focused on
M.Amin Salmani is with the MAPNA Co., Tehran-Iran
providing electrical energy for loads in the VPP control area
(E-mail: amin.salmani@gmail.com). with maximum profit achieved from selling of energy, in
S.M.Moghaddas Tafreshi is with the Department of electrical engineering, addition the VPP can also trade with power market to increase
K.N.Toosi University of technology, Tehran-Iran total profit, but in this case some limitations should be
(E-mail: tafreshi@eetd.kntu.ac.ir).
Hosein Salmani is with the Department of electrical engineering, Shahid
considered, e.g. capacity of distribution lines[5].
Beheshti University, Tehran-Iran Considering these facts, virtual power plants are an interesting
(E-mail:salmani@ee.kntu.ac.ir). option for the future to substitute conventional power plants.
This Paper is sponsored by MAPNA Co.
2

B. VPP's Technical Coordination


The VPP must handle the technical coordination among the
DisCo, the interconnected power grid, the micro grid, DGI,
III. VIRTUAL POWER PLANT (VPP) DGDC and Loads. Fig. 2 summarizes these interrelations.
In this section, a general formulation of the VPP model will be
described. It should be considered that VPPs are divided into two
categories according to the control topology [9]:
A. General Structure in Power Market 1. The decentralized VPP: Each DG unit is locally controlled
The general model considers two different DGs, a micro grid by local controller (LC).
and several loads inside a VPP control area. 2. The centralized VPP: The DG unites are centrally controlled
by control coordination center (CCC).
Wholesale Energy market In this paper centralized VPP (CCC) is used.

DisCo
Interconnected
system

Micro
DG1
grid

VPP
Buyer

DG
Seller I

VPP
DG2 Load 20KV/400v

Public Grid

Public Grid
DG
DC CCC Micro
Grids

400V/20KV
Loods

Fig. 1."connections between vpp control area and power grid"

In Fig.1 the VPP has shown in dashed line as part of a more communication
system
general power market structure. The definitions of a market
agent are as followed [1, 6]. Power transmission
system
•VPP: the aggregation of two DG units, a micro grid and loads
which are connected with power grid. Fig. 2."Power grid in vpp's control area"
• DisCo: Distribution Company that operates the facilities and
purchases energy from the power market or VPP. In Fig. 2, two LV substations between the public grid and the
• Micro Grid: Distributed generation unit inside the VPP VPP control area are shown [10]. For this model following
Control area. This unit is managed or owned independently coordination requirements are identified:
from the VPP. • DGI unit offers: DGI unit must send energy offers (price and
• DGI: Distributed generation unit inside the VPP Control quantity) to the CCC to inform the VPP about its energy
area. This unit is managed or owned independently from the producing schedules.
VPP. From the market point of view DGI sends energy offers • Micro grid offers: micro grid must send energy offers (price
(price and quantity) to VPP controller. and quantity) to the CCC to inform the VPP about its energy
Note: The difference between DGI unit and micro grid is the producing or requirement schedules.
loads which exist inside the micro grid control area. Because It should be noted that in micro grid offers the negative
of this feature some times micro grid purchasing energy from amount of energy show that micro grid is willing to buy energy
VPP to provide energy for loads. from VPP instead of selling energy to VPP.
• DGDC unit offers: In DGDC unit offers instead of the price,
• DGDC: Distributed generation unit inside the VPP Control the cost of producing energy is mentioned.
area, which are managed or owned by the VPP. • Load predictions: It identifies Load contracts, with detailed
• Loads: Customers which VPP has to provide them energy. cost Information for un-served energy ranges, which must be
In this model have two different types of loads are considered submitted to the CCC.
[7, 8]: • Market predicted prices: CCC has to predict the prices in
1. Sensitive loads: which is obligatory to provide them each node that VPP is connected to the main Grid.
energy completely. All the above information is supplied to the CCC in order to
2. Non-sensitive loads: Also defined as Cut-able load, take a decision on amount of energy that each generator can
are those non critical loads that incase energy generate.
shortage occurs, the system can forsake their needed
power, partially or completely, but financial penalty IV. VPP’S MODELING
is the consequence.
VPP is organized as a day-ahead market with daily offers from The information described in previous section can be
DGI, DGDC, micro grid, DisCo, and the wholesale market. In formulated for each time period as a mathematical
addition, VPP manages the cost of un-provided energy for all optimization problem (OP).
of the Loads.
3

A. General Structure VI. PRODUCING COST OF DGDC


OP in previous section represents the general structure of the In this paper the features of a 90 KW fuel cell unit [10, 11] is
proposed model [1]: used for the DGDC unit. Producing cost function for this unit
is shown:
C DGDC D  E PDGDC  J PDGDC
2
(11)
Max Z=f(x)
s.t. g(x)=0 (1)
h(x) ”0 The parameters of this function are different for each case
xmin < x <xmax study, for the first case study:
This structure corresponds to a maximization problem. The D ($ / h) 0 , E ($ / MWh) 30.4 and J ($ /( MW ) 2 h) 0.01
objective function f(x) represents the VPP revenues for
supplying its demand and trading with power market. And for the second case study:
B. Objective Function
D ($ / h) 0 , E ($ / MWh) 30.4 and J ($ /( MW ) 2 h) 0.05
The resulting objective function for CCC is as Followed:

VII. SOLVING THE OPTIMIZATION PROBLEM


24
The General Algebraic Modeling System (GAMS) is a high-
f ( x) ¦
i 1
PLoad 1i U Load 1i  ( PLoad 2i  Puki ) U Load 2i  Pex1i U ex1i  Pex 2i U ex 2i
level modeling system for mathematical programming
Cuk Puk  C DGDC PDGDC  Pmic U mic  PDGI U DGI problems. It consists of a language compiler and a variety of
integrated high-performance solvers. Solving the VPP
(2)
optimization problem with the GAMS will discussed in the
following part. [12, 13, 14, 15, 16]
V. CONSTRAINETS
In this paper the solutions for two different case studies will be
1. Power Balance: These are the nodal active power balances presented. The inputs of both cases are the same, only some
[1, 10]. Its general form is changes in general assumptions of modeling made these two
cases different.
In the first case study this assumptions are considered valid:
PLoad 1i  ( PLoad 2i  Puki )  Pex1i  Pex 2i  PBisi PDGDCi  Pmici  PDGIi  PBipi
• U Load 1 = 68.5 $/MW
(3)
• U Load 2 = 55.5 $/MW
2. Transmission Line Limits: These constraints represent the • Cuk = 10.566 $/MW
power flow limits on transmission lines. Two alternative
You can see the inputs below:
models are proposed [1]:
TABLE I
MICRO GRID'S BID:
• By using active power flow transmission limit:
Pij d PijMax (4)
Hour
U mic ($ / MW ) PMaxmic (kW )
1 31.234 240
2 28.475 250
• By using current limit: 3 28 240
2 2 4 26.5 230
I ij d I ijMax (5) 5 26.45 220
3. LV substations limit: This constraint represent the power 6 29.865 230
7 30.465 210
flow limit on LV substations: 8 45.861 210
S T d S T max (6) 9 49.5 190
10 53 180
11 55.741 190
4. Bounds: All optimization variables are subject to the 12 56.1 170
following operational constraints: 13 56.99 130
Vmin i d Vi d Vmax i (7) 14 55.274 130
15 62.5 120
Vmin 0.95 P.u. 16 64.8 110
17 64.719 -20
Vmax 1.05 P.u. 18 60.01 -30
Pmin d P d Pmax (8) 19
20
56.399
48.26
-30
150
Qmin d Q d Qmax (9) 21 45.5 180
22 40.1 180
5. Unserved energy limitation: Total unserved energy must be 23 35.55 230
less than 10% of total loads: 24 33.8 230
Puk d 0.1( PLoad 1  PLoad 2 ) (10)
4

TABLE II
THE DGI'S BID
Hour
U DGI ($ / MW ) PMaxDGI (kW ) TABLE IV
MARKET FORECASTED PRICES:
1 30.34 180
2 26.82 180 Hour
U exi ($ / MW )
3 27.2 180 1 29.255
4 27.652 180 2 26.397
5 28.655 180 3 22.47
4 21.07
6 29.399 190
5 23.163
7 31.45 190 6 30.863
8 44.3 190 7 31.556
8 47.39
9 48.68 190
9 49.7
10 52.251 200 10 52.1
11 54.479 200 11 55.35
12 55.01 200 12 55.5
13 57.01
13 56.978 200 14 54.42
14 55.3 200 15 63.12
15 63.715 200 16 65.59
17 67.24
16 63.99 200
18 63.87
17 65.267 200 19 55.61
18 59.89 200 20 52.55
21 47.55
19 49.865 190
22 39.69
20 47.65 190 23 37
21 45.348 190 24 30.51
22 38.69 180
23 36.1 180
24 31.724 180

TABLE III The GAMS out puts for the first case study are shown in Fig. 3:
LOADS FORECASTING 
Hour
PLoad PLoad 1 PLoad 2

1 365 300 65
2 350 305 45
3 300 265 35
4 300 265 35
5 350 300 50
6 400 350 50
7 400 330 70
8 460 375 85
9 485 385 100
10 505 405 100
11 511 396 115
12 545 425 120 Fig. 3." The GAMS out puts for first case study"
13 535 415 120
14 545 425 120 We used these information for second case study’s input, In
15 550 415 135 addition for the second case study, some assumptions are
16 555 420 135 changed, for instance energy prices. The assumptions are
17 570 405 165 considered valid [10, 11]:
18 540 380 160
19 510 360 150 • U Load 1 = 77.5 $/MW
20 485 360 125 • U Load 2 = 63.5 $/MW
21 465 365 100
22 410 310 100 • Cuk = 12.5 $/MW
23 385 300 85
24 380 305 75
5

respect of the network constrains, such as allowable line


And the GAMS out puts for the second case study can be loading and bus-bar voltage levels.
observed in Fig.4: New constraints are also taken into accounts, which are related
to lines capacity and voltage level at each bus of the network.
This approach involves many innovative aspects concerning
the management of distribution systems, especially related to
the planning strategies for active networks or micro grids.

IX. REFERENCES
[1] Rodrigo Palma-Behnke, Senior Member, IEEE, José Luis Cerda A., Luis
S. Vargas, Senior Member, IEEE, and Alejandro Jofré," A Distribution
Company Energy Acquisition Market Model With Integration of
Distributed Generation and Load Curtailment Options", IEEE
TRANSACTIONS ON POWER SYSTEMS, VOL. 20, NO. 4,
NOVEMBER 2005
Fig. 4." The GAMS out puts for second case study"
[2] Robertoo Caldon, Andera Rossi Patria and Roberto Turri,"optimization
algorithm for a virtual power plant operation".
It can be derived that the amount of power that is produced by
each generator, especially amount of power produced by the [3] K.Dielmann, Alwin van der Velden,"VIRTUAL POWER PLANTS (VPP)
- A NEW PRESPECTIVE FOR ENERGY GENERATION?" 2003
DGDC unit, is totally different in case studies. This difference IEEE.
is occurred because of increasing energy prices and producing
cost of DGDC. [4] D. Pudjianto, C. Ramsay and G. Strbac," Virtual power plant and system
integration of distributed energy resources", The Institution of
In addition total profit for the VPP in first case study was Engineering and Technology 2007
506.046 $ for a day and for the second one was 715.6974 $ for
a day and it is an effective proof for this model. [5] Christian Schulz, Gerold Roder, Michael Kurrat Technical University
Braunschweig, Institute of High-Voltage and Electric Power Systems,
Germany," Virtual Power Plants with combined heat and power micro-
units".

[6] J.Oyarzabal, J.Jimeno,J.Ruela,A Engler and C.Hardt, "Agent based Micro


grid management system " Digital Object Identifier: 10.1109/FPS. 2005.
204230, Current Version Published: 2006-03-06

[7] Heiko Englert, "Informationstechnische Vernetzung der DEA in


Virtuellen Kraftwerken", presented at the Workshop Virtual Power
Plants, Unna, Germany, 2005.

[8] T. Ackermann, G. Andersson, and L. Söder, “Electricity market


regulations and their impact on distributed generation,” in Proc. Conf,
Electric Utility Deregulation and Restructuring and Power Technologies
2000, London, U.K., Apr. 4–7, 2000, pp. 608–613.
Fig. 5." The VPP’s hourly profit for each case study”
[9] Eko Adhi Setiawan. "Concept and Controllability of Virtual Power
The profit for each hour of both strategies can be compared in Plant."Kassel University Press GmbH, Kassel , Diss. 2007,ISBN: 978-3-
Fig. 5. 89958-309-0.

VIII. CONCLUSIONS
[10] D. Pudjianto, C. Ramsay and G. Strbac," Microgrids and Virtual power
plants: concept to support integration of distributed energy resources",
The formulation of the algorithm enables to easily deal with IMechE 2008.
different forms and types of production units, each
characterised by its own cost function and constrains. [11] R. Caldon, A. Rossi Patria, R. Tum. “Optimal Control of a Distribution
System with a Virtual Power Plant”, Bulk Power System Dynamics and
The proposed approach has demonstrated to be flexible and Control - VI, Cortina d’Ampezzo (ITA), 22-27 August 2004.
suitable to manage a VPP with more than three generating
systems. [12] Michael R. Bussieck, Michael C. Ferris, Alexander Meeraus. "Grid-
Enabled Optimization with GAMS", September 2008.
The ever greater diffusion of small generating systems are
expected to affect distribution systems, since each VPP, or [13] Federico Milano. "A Graphical and Open Source Matlab-GAMS
single producer, manages its own plant in an independent interface for Electricity Market Models."
manner. Economic optimisation should thus be subject to the [14] João Tomé Saraiva. "Markets and Regulation". Doctoral Programme in
Electrical and Computer Engineering at FEUP.
6

[15] DIDIER HENRION. "Global Optimization Toolbox for Maple", IEEE


CONTROL SYSTEMS MAGAZINE » OCTOBER 2006.

[16] Yohan Sutjandra. "Mathematical Programs with Equilibrium


Constraints".

X. BIOGRAPHIES

M. Amin Salmani was born in Tehran, Iran. He


received his BS degree in 2006 in electrical engineering from K.N.Toosi
University of Technology Tehran/Iran. He is currently pursuing MS
degree in this University and he is working in MAPNA Co. His research
interests are in the areas of energy management, renewable energy,
deregulated power system and distributed generation.

Seyed-Masoud Moghaddas-Tafreshi was


born in Tehran, Iran. He obtained his PhD
degree in 1995 in electrical engineering from
Technical University of Vienna, Austria. Dr.
Moghaddas-Tafreshi is now an Assistant
Professor in the Power Engineering
Department in Electrical Faculty of K.N.Toosi
University of Technology in Tehran/Iran. His
research interests are in the areas of Load and
Energy Management, Renewable Energy,
Power System Operation & Planning of the deregulated power system.
He has numerous postgraduate and undergraduate students in these
fields. He is author of one book with title: Electrical Energy Generation
Resources in the 21st Century and co-author/author of more than 50
published papers.

You might also like