You are on page 1of 6

Quantum-based PSO applied to Hour-Ahead

Scheduling in the Context of Smart Grid Management


João Soares, Marco Silva, Zita Vale P. B. de Moura Oliveira
GECAD - Knowledge Engineering and Decision-Support INESC TEC – INESC Technology and Science
Research Centre UTAD University
Polytechnic of Porto (ISEP/IPP) 5001-801 Vila Real, Portugal
4200-072, Porto, Portugal oliveira@utad.pt
{joaps, marsi, zav}@isep.ipp.pt

Abstract—This paper presents a Quantum Particle Swarm N ST Total number of storage units
Optimization (QPSO) applied to hour-ahead scheduling in N SP Total number of external suppliers
Smart Grid (SG). The unforeseen events not considered in day-
ahead scheduling, must be overcome when approaching intraday NV Total number of electric vehicles
operation. This implies new constraints in hour-ahead PCh ( ST ) Active power charge of storage ST (W)
formulation. The developed methodology aims to complement
PCh (V ) Active power charge of electric vehicle V (W)
the day-ahead scheduling tools already available on the
literature. The unforeseen events in the hour-ahead can include PChLimit ( ST ) Maximum active power charge of storage unit ST (W)
change of forecasted load demand, market prices and
PChLimit (V ) Maximum active power charge of electric vehicle V (W)
availability of renewable generation. The QPSO solves the
problem in adequate execution time under the hour-ahead time PCut ( L ) Active power of curtailment program of load L (W)
scale. This is demonstrated using a scenario with high PDch ( ST ) Active power discharge of storage ST (W)
penetration of distributed generation and gridable vehicles.
Furthermore, a comparison with GAMS is presented. PDch (V ) Active power discharge of electric vehicle V (W)
PDchLimit ( ST ) Maximum active power discharge of storage unit ST (W)
Index Terms—Electric Vehicles, Energy Resource Management,
Hour-Ahead Scheduling, Quantum PSO, Smart Grid. PDchLimit (V ) Max. active power discharge of electric vehicle V (W)
PDchMin ( ST ) Minimum active power discharge of storage unit ST (W)
I. NOMENCLATURE PDchMin (V ) Min. active power discharge of electric vehicle V (W)
f Objective function (m.u.) PDG ( DG ) Active power generation of DG unit (W)
cCut ( L ) Demand response curtailment cost of load L (m.u./Wh) PDGMax ( DG ) Maximum active power generation of DG unit (W)
cDch ( ST ) Discharge cost of storage ST (m.u./Wh) PDGMin ( DG ) Minimum active power generation of DG unit (W)
cDch (V ) Discharge cost of electric vehicle V (m.u./Wh) PGCP ( DG ) Generation curtailment power by DG unit (W)
cDG ( DG ) Generation cost of DG unit (m.u./Wh) PLoad ( L ) Active power demand of load L (W)
cGCP ( DG ) Generation curtailment power cost of DG unit (m.u./Wh) PMaxCut ( L ) Maximum DR curtailment of load L (W)
cNSD ( L ) Non-supplied demand cost of load L (m.u./Wh) PMaxRed ( L ) Maximum DR reduction of load L (W)
cRed ( L ) Demand response reduction cost of load L (m.u./Wh) PMinCut ( L ) Minimum DR curtailment of load L (W)
cSP ( SP ) Energy price of the external supplier SP (m.u./Wh) PMinRed ( L ) Minimum DR reduction of load L (W)
MPCh ( ST ) Selling price for the charge of storage ST (m.u./Wh) PNSD ( L ) Active non-supplied demand for load L (W)
MPCh (V ) Selling price for the charge of vehicle V (m.u./Wh) PRed ( L ) Active power of reduction program of load L (W)
MPLoad ( L ) Selling price of load L (m.u./Wh) PSell Active power selling energy of VPP (W)
MPSell Selling price of selling energy to the market (m.u./Wh) PSP ( SP ) Active power generation of the external supplier SP (W)
N DG Total number of distributed generators PSPMax ( SP ) Maximum active power of the external supplier SP (W)
NL Total number of loads PSPMin ( SP ) Minimum active power of the external supplier SP (W)
This work is supported by FEDER Funds through COMPETE program and
by National Funds through FCT under the projects FCOMP-01-0124-
FEDER: PEst-OE/EEI/UI0760/2014, PTDC/SEN-ENR/122174/2010, and
SFRH/BD/87809/2012, and by the GID-MicroRede, project nº 34086, co-
funded by COMPETE under FEDER via QREN Programme, and by the
SASGER-MeC, project nº NORTE-07-0162-FEDER-000101, co-funded by
COMPETE under FEDER Programme.
II. INTRODUCTION
The short-term Energy Resources Management (ERM) is
one of the main challenges in modern energy systems
considering the high penetration of Distribution Energy
Resources (DER). Distribution system operators, and
aggregators such as Virtual Power Players (VPP) will need to
adopt new strategies, in order to reduce the operation costs, to
face the requirements of a new distribution network operation
paradigm [1-3]. In order to achieve adequate and realistic
results, the ERM must consider all the resources, including
Distributed Generation (DG), Demand Response (DR), and
storage resources including Electric Vehicles (EV). The
number of variables and the non-linearity of the problem
typically considered in ERM approaches, along with the need
to obtain a rapid response, requires finding alternatives to
mathematical approaches and conventional techniques. The
available solvers for non-convex and non-linear problems
sometimes fail to find a feasible solution and require a lot of
computational resources. The alternative to mathematical
approaches lies in the use of advanced optimization
techniques, such as Artificial Intelligence (AI) techniques,
namely metaheuristics [4]. The use of metaheuristics allows Figure 1. ERM platform
the VPP not only to ensure the secure distribution network
operation but also to quickly find an optimal generation In this work it is assumed that a VPP is able to control and
scheduling for its resources. Nevertheless, mathematical and operate different DERs in a specific area of the grid [9]. The
metaheuristics approaches can coexist in the same external supplier simulates contracts with another VPPs or the
environment. The VPP can decide to use both mathematical electricity market. The presented framework requires a
and metaheuristics approaches in their decision tools in order database, with the information of all the units of the DERs
to take timely and proper actions with the best available involved and updated forecasts, allowing to support the ERM
solutions. Several works have been reported in the literature and the participation in the electric markets. The input data
[2, 5-7]. The ERM is addressed in [2, 7] in several horizons presents several information about the resources involved for
using traditional techniques. The contribution of this paper lies each simulation period (e.g., DG, storage units, load, EVs,
in the development and implementation of the Quantum-based distribution network and external supplier). The input data of
Particle Swarm Optimization (QPSO) to solve the hour-ahead the methodology includes the characteristics of each resource,
scheduling in the ERM platform considering unforeseen such as the generation power limits, the generation costs, the
events not predicted in day-ahead scheduling. These events type of contract and the units location in the distribution
can include change of forecasted load demand, market prices network. The first phase ERM objective is to perform the day-
and availability of renewable generation. QPSO has proven in ahead resources scheduling, i.e., the analysis of energy
[8] to outperform traditional PSO in most benchmark resources for the next day, considering the information
functions. contained in the input data and relevant day-ahead forecast for
each resource involved [2]. The scheduling is performed for a
After this introductory section, this paper is followed by a 24 hours period. The first ERM phase has been studied in [6,
section II that introduces the methodology of the hour-ahead 10]. This work focuses in the second ERM phase which so far
ERM. Then, section III details some aspects of the QPSO is less studied. The second phase corresponds to the hour-
approach and in section IV, some results are already shown. ahead scheduling, which takes into account updated hour-
Finally, Section VI presents the main work conclusions. ahead forecasts and the results of the day-ahead scheduling.
The results of the day-ahead scheduling are used as constraints
III. SHORT-TERM ERM of the hour-ahead problem (see Fig. 2). The second phase
A. Hour-ahead scheduling in ERM context results are used in the third phase as problem constraints and
with accurate forecasts for the next few minutes. The ERM
The proposed methodology aims to optimize the DERs third phase is out of the scope of this paper but it is studied in
management that are available in a smart grid managed by a [2, 7].
VPP, and considers all the available resources: DG, EVs,
storage units, and DR. The goal is to satisfy demand aiming at B. Hour-ahead problem formulation
obtaining the maximum profit. The problem formulation The hour-ahead problem formulation is classified as a
considers the equipment technical characteristics, their
Mixed-Integer Non-Linear Programming (MINLP) because it
operation costs, and the envisaged demand response actions.
includes the AC electrical network constraints. The VPP aims
The ERM can be divided in three phases (day-ahead, hour- to maximize his own profits that are equal to the income
ahead and real-time scheduling). Fig. 1 depicts the several minus the operation costs, as shown in (1). The day-ahead
phases of the ERM platform.
DER scheduling formulation can be seen in [2, 7]. This paper
is focused in the hour-ahead formulation, detailed next.

Maximize f = In − C (1)

The VPP can receive its income (In), namely the revenue ⎛ NDG N DG
⎞ ⎛ NL NL

⎜ ∑ PDGF2 ( DG ) − ∑ PDG1( DG ) ⎟ ≥ ⎜ ∑ PLoadF2 ( L ) − ∑ PLoad1( L ) ⎟
from supplying the demand power to consumers. The second ⎝ DG =1 DG =1 ⎠ ⎝ L =1 L =1 ⎠

one derives from selling energy to the electricity market. The


third one is the revenue from the charging process of storage
units. Finally, the last one also comes from the charging of Indexes :
1 Day-ahead variables

EVs. 2 Hour-ahead variables


F 2 Updated hour-ahead forecast

Generators : Generators :
• PDGmin ( DG ) ≤ PDG2 ( DG ) ≤ PDG1 ( DG ) • PDG1( DG )≤ PDG2 ( DG ) ≤ PDGmax ( DG )

⎡ NL
⎤ External suppliers : External suppliers :

⎢ ∑ MPLoad ( L ) × PLoad ( L ) + MPSell × PSell + ⎥ • PSPmin ( SP ) ≤ PSP2 ( SP ) ≤ PSP1 ( SP ) • PSP1( SP )≤ PSP2 ( SP ) ≤ PSPmax ( SP )

In = ⎢ N ⎥
L =1 Loads : Loads :

⎢ ST NV ⎥
(2) • PR e dmin ( L )≤ PRe d2 ( L ) ≤ PRe d1 ( L ) • PR e d1 ( L ) ≤ PR ed2 ( L ) ≤ PR e dmax ( L )

⎢ ∑ MPCh ( ST ) × PCh ( ST ) + ∑ MPCh (V ) × PCh (V ) ⎥ • PCutmin ( L )≤ PCut2 ( L ) ≤ PCut1 ( L ) • PCut1 ( L ) ≤ PCut2 ( L ) ≤ PCutmax ( L )

⎣ ST =1 V =1 ⎦ Storage system : Storage system :


• PCh1 ( ST ) ≤ PCh2 ( ST ) ≤ PChmax ( ST ) • PCh2 ( ST ) = PCh1 ( ST )
• PDchmin ( ST ) ≤ PDch2 ( ST ) ≤ PDch1 ( ST ) • PDch1 ( ST ) ≤ PDch2 ( ST ) ≤ PDchmax ( ST )

Function C (3) evaluates the operation cost of the Electric vehicles : Electric vehicles :
• PCh1 ( EV ) ≤ PCh2 ( EV ) ≤ PChmax ( EV ) • PCh2 ( EV ) = PCh1 ( EV )
resources managed by the VPP. It considers the cost with • PDchmin ( EV ) ≤ PDch2 ( EV ) ≤ PDch1 ( EV ) • PDch1 ( EV ) ≤ PDch2 ( EV ) ≤ PDchmax ( EV )
distributed generation, external suppliers, discharge of storage
and EVs, demand response programs, non-supplied demand
and penalization with DG units’ generation curtailment. Figure 2. Decision process for the hour-ahead ERM

C. Quantum PSO
⎡ N DG N SP

⎢ ∑ cDG ( DG ) × PDG ( DG ) + ∑ cSP ( SP ) × PSP ( SP ) + ⎥ The traditional PSO relies on the convergence to the global
⎢ DG =1 SP =1 ⎥ best (usually the best solution of the swarm) particle,
⎢ N ST NV ⎥ independently of the position of other particles. This particular
⎢ ∑ cDch ( ST ) × PDch ( ST ) + ∑ cDch (V ) × PDch (V ) + ⎥ feature is the major shortcoming of the method when there are
C = ⎢ STN =1 V =1 ⎥
(3) few particles far away from the global best position. The
⎢ L NL ⎥
⎢ ∑ cCut ( L ) × PCut ( L ) + ∑ cRed ( L ) × PRed ( L ) + ⎥ concept of lagged particles corresponds to the personal bests
⎢ L =1 L =1 ⎥ of a few particles, which are far away from the global best
⎢ NL N DG ⎥
⎢ ∑ cNSD ( L ) × PNSD ( L ) + ∑ cGCP ( DG ) × PGCP ( DG ) ⎥ position and the rest of the particles. In the QPSO method the
⎢⎣ L =1 DG =1 ⎦⎥ lagged particles are not abandoned by the swarm. The lagged
particles affect the mean best position and therefore the lagged
Fig. 2 depicts the ERM hour-ahead model decision particles are shifted towards the rest of the swarm. The mean
process. The set of hour-ahead constraints changes if there is best position is intrinsically related with the movement
surplus of generated energy or shortage of generated energy equation of QPSO (5). The particles’ distribution affects the
taking into account the new forecasted demand (done on an convergence rate; however, QPSO can provide stronger global
hourly basis). In the case of overproduction or surplus of search ability than traditional PSO [8]. QPSO was chosen due
generated energy the hour-ahead model limits the generators to its rich features and adapted to the ERM problem under
study.
variables upper bounds, the DR variables, and the storage and
EVs discharge variables to the obtained values in the day- Equation (4) presents the local focus equation which is a
ahead results. The lower bounds remains unchanged. On random point located within the hyperplane constructed
another hand, storage and EVs charge values can increase. In between Pi,n and Gn in the search space. The local focus is then
this case, the lower bounds are the day-ahead results. In the used in equation (5). The contraction–expansion (CE)
overconsumption case the hour-ahead model changes the coefficient α, which is vital to the dynamical behavior of an
constraints regarding the following rules: the lower bounds individual particle and the convergence of the algorithm. The
for generators variables, DR variables, and storage and EVs algorithm decides to add or subtract the second term of
discharge variables are set to the values obtained in the day- equation (5) using a probability of 50%. More details about
ahead results, while the upper bounds remain unchanged, i.e. QPSO aspects are extensively studied in [8].
the maximum allowed valued regarding the respective
technology. The storage and EVs charge variables are fixed to pij, n +1 = φij, n Pi ,jn + (1 − φij, n )Gnj (4)
the previous values of the day-ahead results. The given rules
are evaluated every hour during the intraday operation and ⎛ 1 ⎞
verified with the previous day-ahead results. X i j, n +1 = pij, n ± α X i ,jn − Cnj ln ⎜ j ⎟ (5)
⎜u ⎟
⎝ i , n +1 ⎠
IV. CASE STUDY TABLE I. GRID RESOURCES DATA

The proposed methodology was tested using a case study Availability Prices
implemented in a distribution network with 33 buses Energy resources (kW) (m.u./kWh)
considering a future scenario regarding the penetration of DG min – max (min-max)
units and EVs, as shown in Fig. 3 [11]. The VPP was able to Biomass 0 – 350 0.09
CHP 0 – 1150 0.06
manage a large wind farm (bus 1, see Fig. 3), 66 DG units, 10
Fuel Cell 0 – 140 0.15
external suppliers (main grid supply in bus 33), 1800 EVs, 15 Small Hydro 0 – 70 0.07
storage units with 324 kWh capacity each, and several Photovoltaic 0 – 870 0.20
consumers with DR programs (15% load reduce program). Waste-to-energy 0 – 10 0.10
Wind 182 – 860 0.15
Large Wind 1172 – 1801 0.07
External Suppliers 6200 0.06 – 0.15
Charge 0 – 900 0.14
Storage
Discharge 0 – 900 0.19
Electric Charge 0 – 5870 0.14
Vehicle Discharge 0 – 5720 0.19
Demand Red 0 – 1174 0.150 – 0.160
Response Cut 0 – 579 0.16
Load 3015 – 5303 0.14
Market 0 – 1000 0.08

a) small wind and PV power production profile (p.u.)

Figure 3. Network configuration used in the case study

Table I depicts the prices and the available capacity of the


resources used in the case study. The fixed loads price was set
to 0.14 m.u./kWh. The load demand varied between 3015 and
5303 kW depending on the period. The discharge price for
storage systems was set to 0.19 m.u./kWh while the charge
price was set to 0.14 m.u./kWh (same as fixed load). The
considered external suppliers (main grid) capacity was 6200
kW (permanent maximum capacity). The most abundant b) power consumption profile (p.u.)
dispersed energy resource considered was wind energy,
followed by CHP and PV. Fuel cell, biomass and small hydro
considered but with small share.
Fig. 4 presents the used forecast profiles for the day-ahead
and hour-ahead in what concerns the large wind production,
the load demand (without EVs), and the small wind and PV
production. While the day-ahead forecast aims to feed the
inputs of day-ahead model, the hour-ahead forecast is done on
an hourly basis to feed the inputs of the hour-ahead model.
The generated differences between day-ahead and hour-ahead
aim to simulate a realistic behavior. Price variations were not
simulated in this case study to simplify the results analysis in c) large wind power production profile (p.u.)
the following section. The values in the figure are depicted in Figure 4. Forecast in day-ahead and hour-ahead
p.u. because a base of 1 MVA was used in the simulations.
Fig. 5 presents the EVs trip demand of the 1800 EVs V. RESULTS
scenario. The EVs scenarios were developed using a tool The present simulations has been tested on a PC with an
reported in [12]. The total expected consumption in the day- Intel Core i7-3520M 2.90 GHz processor, with 4 Cores, 8 GB
ahead was 13770 kWh corresponding to an average of 7.7 of random-access-memory and Windows 8 operating system.
kWh per vehicle. In the hour-ahead we assumed that these
trips did not changed for the sake of simplicity. Fig. 6 presents the day-ahead results using QPSO
approach. The total operation cost for the day-ahead results
was 11636 m.u. and the expected total income was 18849 m.u.
while the total profit was 7213 m.u. The DG use of is higher
during the day due to the higher load demand and the
availability of photovoltaic generation. There was no active
scheduling of storage and vehicles discharges. The total
generation from DG was 92360 p.u. and 45030 p.u. from
external suppliers. The total load was 102970 p.u., the storage
charge was 2480 p.u. and the vehicle charge was 29520 p.u.
while DR was not scheduled in this case.

Figure 5. EVs trip demand

Table II depicts the overconsumptions and


overproductions simulated in the case study. This table shows
that for each period if an overconsumption or overproduction
was verified in the scenario according to the rules defined in
section III (see Fig. 2). It can be seen in Table II that during
the first 16 periods of the day an overconsumption occurred
and an overproduction afterwards. QPSO was implemented
for the day-ahead scheduling and hour-ahead scheduling. The
a) power production profile (p.u.)
meta-heuristic was configured with 10 particles randomly
generated at start and a CE coefficient α decreasing from 0.4
to 0.3 during the evolution stage.

TABLE II. OVERCONSUMPTION AND OVERPRODUCTION

Hour 1 2 3 4 5 6 7 8 9 10 11 12

+/-* + + + + + + + + + + + +
Hour 13 14 15 16 17 18 19 20 21 22 23 24
+/-* + + + + - - - - - - - -
*(+) is overconsumption and (-) is overproduction (-)
b) power consumption profile (p.u.)
Table III shows the QPSO parameters used in the case Figure 6. Results of day-ahead scheduling
study. The stopping criteria was defined with a maximum of
2000 iterations and a minimum of 500 iterations. A threshold Fig. 7 presents the hour-ahead results using QPSO
value was programmed to allow the QPSO to stop earlier if approach. The total operation cost for the hour-ahead results
the fitness function did not evolved after 400 iterations. The was 11587 m.u. and the expected total income was 18728 m.u.
maximum and minimum variables position were defined while the total profit was 7141 m.u. (1% lower than expected
equally to the upper and lower bound, respectively. In the in the day-ahead). The decrease of profits is possible in the
hour-ahead case the rules mentioned earlier were applied to intraday operation due to natural variations and unforeseen
these limits. events. The use of DG was higher during the day due to the
higher load demand and the availability of photovoltaic
TABLE III. PARAMETERS OF QUANTUM-BASED PSO (QPSO) generation. There was no active scheduling of storage and
vehicles discharges during the entire day. This fact already
Parameters QPSO
occurred in the day-ahead results. The reason was mainly
# Individuals 10
related to the defined discharge price (see Table I). The total
Initial Solution Random
# Iterations 2000 generation from was 90930 p.u. from DG and 45310 p.u. from
Stopping Max. 2000 iterations Min. 500 iterations And fitness external suppliers. The total load was 102000 p.u., the storage
Criteria threshold (1-9) during last 400 iterations charge was 2240 p.u. and the vehicle charge was 29520 p.u.
Max. Positions Equal to the upper bound of variables while DR was not scheduled in this case. The DG generation
Min. Positions Equal to the lower bound of variables reduce 1.5% while the external suppliers increased 0.7%. The
Alpha ( α ) Linear decreasing 0.4 to 0.3 storage charge in the hour-ahead reduced 10% when compared
to the day-ahead results.
the mathematical approach MINLP were included. The
results have shown that QPSO outperforms the mathematical
approach in terms of execution time while presenting similar
solution quality in the day-ahead. In the hour-ahead, the
results have demonstrated very similar performance and
solution quality. The tools such as the one presented in this
paper, clearly indicate the importance of their use in the
context of smart grid operation to support the VPPs operation
and ultimately reducing costs. Further research is encouraged
to improve the hour-ahead model constraints considering
c) power production profile (p.u.)
variation of EVs demand in the hour-ahead model and
stochastic scenarios in the day-ahead model.
VII. REFERENCES
[1] H. Morais, P. Kadar, P. Faria, Z. A. Vale, and H. M. Khodr, "Optimal
scheduling of a renewable micro-grid in an isolated load area using
mixed-integer linear programming," Renewable Energy, vol. 35, pp.
151-156, Jan 2010.
[2] M. Silva, H. Morais, Z. Vale, and P. Faria, "Short-term Scheduling
Considering Five-minute and Hour-ahead Energy Resource
Management," 2012 IEEE Power and Energy Society General Meeting,
2012.
[3] A. Arif, F. Javed, and N. Arshad, "Integrating renewables economic
d) power consumption profile (p.u.) dispatch with demand side management in micro-grids: a genetic
algorithm-based approach," Energy Efficiency, vol. 7, pp. 271-284, Apr
Figure 7. Results of hour-ahead scheduling 2014.
[4] Z. Vale, G. Venayagamoorthy, J. Ferreira, and H. Morais,
Table IV and Table V presents the obtained results in day- "Computational Intelligence Applications for Future Power Systems," in
ahead and hour-ahead, accordingly. A deterministic technique Computational Intelligence for Engineering Systems. vol. 46, A.
results are included in both tables for comparison purposes. Madureira, J. Ferreira, and Z. Vale, Eds., ed: Springer Netherlands,
2011, pp. 176-193.
The MINLP implemented in GAMS [13] took more than 14 [5] M. Motevasel and A. R. Seifi, "Expert energy management of a micro-
hours to solve the day-ahead problem while QPSO took just a grid considering wind energy uncertainty," Energy Conversion and
few minutes with a very close profit to MINLP. In the hour- Management, vol. 83, pp. 58-72, Jul 2014.
ahead, MINLP and QPSO presented very similar results with [6] J. Soares, T. Sousa, H. Morais, Z. Vale, B. Canizes, A. L. D. Silva, et al.,
fast execution time (less than 1 minute). The problem reduces "Application-Specific Modified Particle Swarm Optimization for energy
resource scheduling considering vehicle-to-grid," Applied Soft
from multi-period in day-ahead to one period in hour-ahead
Computing, vol. 13, pp. 4264-4280, 2013.
which decreases the computational time. [7] M. Silva, H. Morais, T. Sousa, and Z. Vale, "Energy resources
management in three distinct time horizons considering a large variation
in wind power," in European Wind Energy Conference & Exhibition
TABLE IV. COMPARISON OF OBTAINED RESULTS (DAY-AHEAD)
2013, 2013.
Operation [8] J. Sun, C. H. Lai, and X. J. Wu, Particle Swarm Optimisation: Classical
Income Profits Execution and Quantum Perspectives: Taylor & Francis, 2011.
Method Costs #Iterations
(m.u.) (m.u.) time (s) [9] P. B. Andersen, B. Poulsen, M. Decker, C. Traeholt, and J. Ostergaard,
(m.u.)
"Evaluation of a Generic Virtual Power Plant Framework Using Service
QPSO 18942 11729 7213 736 893
Oriented Architecture," 2008 Ieee 2nd International Power and Energy
MINLP 18885 11636 7259 (14 hours) -
Conference: Pecon, Vols 1-3, pp. 1212-1217, 2008.
[10] T. Sousa, H. Morais, J. Soares, and Z. Vale, "Day-ahead resource
TABLE V. COMPARISON OF OBTAINED RESULTS (HOUR-AHEAD) scheduling in smart grids considering Vehicle-to-Grid and network
constraints," Applied Energy, vol. 96, pp. 183-193, Aug 2012.
Operation [11] P. Faria, Z. A. Vale, and J. Ferreira, "DemSi - A Demand Response
Income Profits
Method Costs #Iterations Simulator in the context of intensive use of Distributed Generation,"
(m.u.) (m.u.)
(m.u.) IEEE International Conference on Systems, Man and Cybernetics (SMC
QPSO 18728 11587 7141 504 2010), 2010.
MINLP 18728 11587 7141 - [12] J. Soares, B. Canizes, C. Lobo, Z. Vale, and H. Morais, "Electric Vehicle
Scenario Simulator Tool for Smart Grid Operators," Energies, vol. 5, pp.
VI. CONCLUSIONS 1881-1899, Jun 2012.
The hour-ahead model in the context of ERM was [13] L. K. Tartibu, B. Sun, and M. A. E. Kaunda, "Multi-objective
optimization of the stack of a thermoacoustic engine using GAMS,"
developed and implemented using a Quantum-behaved PSO Applied Soft Computing, vol. 28, pp. 30-43, Mar 2015.
(QPSO) meta-heuristic based approach. The day-ahead model
is also implemented with QPSO. The results of day-ahead
jointly with the updated forecasts are used as an input of the
hour-ahead model. For comparison purposes, the results of

You might also like