Sie sind auf Seite 1von 6

Optimal generation scheduling strategy in a microgrid

Lucian Toma, Ion Tristiu, Constantin Bulac, Andreea-Georgiana Neagoe-Stefana


Department of Electrical Power Systems,
University POLITEHNICA of Bucharest
Romania
Email: lucian.toma@upb.ro

AbstractThis paper presents a strategy for optimal


generation scheduling in a microgrid. Several small generation
units of different type are considered, that is a wind system, a
photovoltaic system and a gas engine unit. In order to maximize
the use of renewable energy sources (RES) a storage battery is
also integrated. Two generation scheduling algorithms are
elaborated, either aiming at minimizing the use of fuel based unit
or at maximizing the lifetime of the battery. The simulations have
been performed in Matlab.
Keywordsmicrogrid, battery management, renewable energy
sources

I. INTRODUCTION
With the advent of new technologies running on renewable
energy sources (e.g. wind and solar) and small size but highly
efficient fossil fueled power plants (i.e. natural gas and
petroleum), the number of distributed generators in the low
voltage and medium voltage networks has significantly
increased in the last decade and will continue to increase in the
near future. Generation scheduling of small size units, similar
to large generators, requires market strategy in terms of
generation availability and generation costs. Under the actual
power market conditions, a market strategy is strongly related
to an accurate generation forecast and generation/load bidding.
It is well known that the renewable energy sources show a
stochastic behavior, thus there is always a risk for balancing
the commercial agreements in which such units are part of.
Flexible generation solutions, e.g. gas fired units or flexible
hydro units, are thus required to appropriately balance the
mismatches in generation caused by renewables.
The generation units are differently considered for
balancing responsibility in terms of the installed power. In
Romania, for instance, the owner of a power plant of installed
power totaling over 10 MW can sell the energy either on the
day ahead market (DAM) or by bilateral contracts and must
enter into balancing agreements, while units under 10 MW are
allowed to produce the available power without penalizations
for unbalancing. Besides, units under 10 MW can sell energy
to the balancing responsible parties (BRP) only, through a
bilateral agreement. Thereby, the greater the unbalance a
distributed generator (DG) produces the smaller the energy
price it can get. Vice versa, the smaller the unbalance the
greater the energy price can be negotiated. Therefore, if
capable, the owner of a DG or group of DGs is interested to

balance the energy bids, minimizing the unbalances, in order to


get higher price.
There is an important support for development of
renewable energy sources (RES) at international level. These
RES units are replacing, from year to year, the classical units
thereby generating a bigger risk in ensuring proper power
reserves either for frequency control or for congestion
management. Solutions should be found to optimally operate
the small size, but flexible, units in order to limit the problems
created by the RES.
II. THE MICROGRID CONCEPT
The concept of microgrid was first introduced in the
technical literature in [1] and [2] as a solution for the reliable
integration of distributed energy resources (DERs), including
Energy Storage Systems (ESSs) and controllable loads [3].
The microgrids are electrical networks of various size
delimited from the main grid by the power transformer or the
electrical interface. A microgrid can operate either
synchronized to the main grid or insulated from it. When
operating in islanding mode, a local control system to provide
generation-load balancing is required.
Figure 1 suggests an example of a microgrid. The loads
can be monitored using Advanced Metering Infrasctrure
(AMI), and the generation units can be remotely controlled via
SCADA and remote terminal units (RTU).

Fig. 1. Generic representation of a microgrid.

Various mathematical models are provided in the


literature. A battery energy storage system (BESS) was
considered in [4] as a main way of balancing the powers
within and isolated microgrid. A multiagent system was
designed to govern each entity to participate in the balancing
process based on predefined rules. The modeling and control
of a battery management system used in a microgrid for both
grid-connected and autonomous modes is investigated in [5].
The overall management system is demonstrated to be
effective by six case studies at different microgrid operation
modes. The authors of [6] propose a solution for aggregation
of distributed generators in order to reduce the imbalance risk
in the market, by means of an existing methodology based on
stochastic programming. The authors of [7] and [8] propose a
bidding strategy on the electricity market, as a nonequilibrium model based on the deterministic price-based unit
commitment which takes the supply-demand balancing
constraint and security constraints of VPP itself into account.
The storage battery systems gains more and more space in
the power system applications. However, the technology
cannot be considered mature as they are limited to a certain
number of years of operation of a certain charging/discharging
cycles [9].
III. MATHEMATICAL MODEL
Three types of generation units are considered in the
microgrid, that is a photovoltaic power plant, a wind power
plant, and a gas engine unit. Additionally, a battery is used to
store the renewable energy when there is a surplus from the
two units.
The mathematical model is formulated to run in discrete
steps of 1 minute, and the simulation is performed for a 24
hour period, which means 1440 minutes.
The main objective of mathematical model is to minimize
the power generation cost with the gas engine
60x 24

MIN

GE (t )

t =1

subject to load-generation balance


Pload (t ) = Ppv (t ) + Pw (t ) + PGE (t ) + Pbat (t ) + Psys (t )
and other operational constraints that are presented as follows:

A. Characteristics of the wind and solar power plants.


The main purpose of the renewable energy sources (RES)
is to replace the classical power plants to reduce the emissions
and protect the environment. The RES are characterized by
large initial investments. Then, in operation only maintenance
costs are incurred. Therefore, since no fuel costs are involved,
the highest priority is assigned for operation to such
generation units. For this reason, in our application, the RES
are simply represented by generation curves Ppv(t) and Pw(t),
as real time values.
There are many techniques for forecasting the generation
from wind and solar power plants. For national power

systems, the share of generation from RES can be significant


and the forecast errors may take acceptable values. However,
in a microgrid, since the RES are located geographically near,
the error can be very high.

B. Characteristics of the gas engine


Small gas engine driven generators, of installed power
from hundreds of kW to few MW, are currently in operation in
distribution networks for combined heat and power. Similar to
an internal combustion vehicle, the gas engine is very flexible
and is capable of changing the mechanical load very quickly
so that the synchronous generator can change the set-point
within few seconds.
A SCADA system is updating the data acquired from the
network on a 2-3 seconds rate basis, while a primary
frequency control system is set to intervene for frequency
deviations after 30 seconds. For this reason, in our application,
the classical generator is assumed to change the generated
power instantly.
The classical generator is defined by its installed power
PGE,inst, in MW, and the instantaneous power generation
PGE(t), in MW.
In our application, we consider that the gas engine
generator produces energy when RES and battery cannot
supply the load, that is Pload (t ) > Ppv (t ) + Pw (t ) + Pbat (t ) .
C. Characteristics of the battery
A battery is characterized by the total installed energy
Ebat,inst, in MWh, and the maximum instantaneous power
Pbat,max, in MW.
The battery charges when there is a surplus of energy from
the renewable energy units only, that is when
Ppv (t ) + Pw (t ) > Pload (t ) .
The battery charging/discharging management is ensured
using two generation scheduling algorithms. The storage
battery can therefore be charged and discharged in cycles to
store energy from the RES and use it efficiently based on the
scheduling algorithm.
In order to increase the lifetime of the battery, a minimum
and a maximum state of charge, SOCmin and SOCmax, are
considered.
Negative values of Pbat show that the battery is charging,
and positive values show that the battery is producing energy.

D. Grid connection
An interfacing with the main grid is considered. When
there is a surplus of energy from the renewable energy sources
and the battery is fully charged, the surplus of energy is
injected into the main power grid. When
The power exchanged with the main grid is denoted by
Psurplus. Positive values represent import from the grid,
whereas negative values represent export to the grid.
E. Management Algorithm
Two generation scheduling algorithms are developed and
implemented for simulation. The first algorithm aims to reduce
the number of charging/discharging cycles. This can be
achieved if the battery system will not change the operation

mode until it is fully charged or fully discharged. The second


algorithm aims to maximize the use of the renewable energy
sources and thus to minimize the generation from the gas
engine unit. This algorithm will allow the battery to charging
any time there is a surplus of generation from RES or
discharging to cover the load.
The two algorithms are shown as follows.
ALGORITM 1
Step 1. Set the initial values for generation and load powers
Step 2. Calculate generated power at time instant t
IF Cbat=1 [the battery is in discharging mode]
IF
PL>(PW+PPV) and PL(PW+PPV)<Pbat,max
Pbat=PL-(PW+PPV)
PGE=0
ELSEIF PL>(PW+PPV) and PL(PW+PPV)>Pbat,max
Pbat=Pbat,max
PGE=PL(PW+PPV+Pbat)
IF PGE>PGE,inst
PGE=PGE,inst
ENDIF
ELSE
Pbat=0
PGE=0
ENDIF
ELSEIF Cbat=0 [the battery is in charging mode]
IF
PW+PPV>PL and PW+PPV-PL<Pbat,max
Pbat=-((PW+PPV)-PL)
PGE=0
ELSEIF PW+PPV>PL and (PW+PPV)-PL>=Pbat,max
Pbat=-Pbat,max
PGE=0
ELSE
Pbat=-Pbat,max
PGE=0
ENDIF
Pbat=0
PGE= PL-(PW+PPV)
IF PGE>Pgas,inst
PGE=Pgas,inst
ENDIF
ENDIF

Step 3. Calculate the power surplus


Step 4. Update the instantaneous battery energy
Ebat(t)=Ebat(t-1)-Pbat/60

and calculate the state of charge


SoC(t)=Ebat(t)/Ebat,inst

ALGORITHM 2
Step 1. Set the initial values for generation and load powers

Step 2. Calculate generated power at time instant t


IF PL>(PW+PPV)
IF SoCmin < SoC < SoCmax
IF PW+PPV+Pbat,max < PL
Pbat=Pbat,max
PGE=PL-(PW+PPV+Pbat,max)
IF PGE>Pgas,inst
PGE=Pgas,inst
ENDIF
ELSE
Pbat=PL-(PW+PPV)
PGE=0
ENDIF
ELSE
Pbat=0
PGE=PL-(PW+PPV)
IF PGE>Pgas,inst
PGE=Pgas,inst
ENDIF
ENDIF
ELSEIF PL<(PW+PPV)
IF SoC < SoCmax
Pbat=PL-(PW+PPV)
IF Pbat < -Pbat,max
Pbat=-Pbat,max
ENDIF
ELSE
Pbat=0
ENDIF
PGE=0
ENDIF

Step 3. Calculate the power surplus


Step 4. Update the instantaneous battery energy
IV. CASE STUDY
Let us assume a microgrid consisting of a wind generation
system, a photovoltaic generation system, a gas engine
generator a battery system and the local load. Three cases will
be studied to determine the economic efficiency regarding the
use of additional fuel to run the gas engine.
The main characteristics of the microgrid are:
- gas engine installed power, PGE,inst = 1.4 MW;
- battery installed power Pbat,max = 0.6 MW;
- battery installed energy, Ebat,inst = 1 MWh; the total
energy stored in the battery is greater than the
instantaneous power that can be produced or absorbed
by the battery;
- battery minimum state o charge, SOCmin = 25%;
- battery maximum state of charge, SOCmax = 75%;
- the load is represented in black;
- the wind generation is represented in green, the PV
generation is represented in cyan;

A. Case 1
The first case assumes employing the first algorithm, in
which the battery is not allowed to discharged when operates
in charging mode, and is not allowed to charge when operates
in discharging mode. This means that the battery is used to
produce energy after a full charging cycle.
Figure Fig. 2 shows the load and generation curves for
Case 1. The algorithm applied has resulted in unused power
(Surplus) from the renewable sources. This is due to that fact
that the battery is not allowed to charge when in discharging
mode although there is a surplus of power from renewables.
Since the battery is not fully charged and thus is not
available, for generation, the gas engine is used. This results in
a total energy generation from the gas engine unit of 5.65
MWh, while the unused energy (Surplus) is 0.523 MWh.

0.9

0.8

Battery - state of charge [-]

the battery is represented in magenta; positive values


shows power generation (the battery is discharging),
and negative values shows power consumption (the
battery is charging)
All simulated cases assumes that the initial state of charge
of the battery is SoC = 0.56 (56%), and the battery operates in
discharging mode with Cbat = 1.

0.7

0.6

0.5

0.4

0.3

0.2

500

1000

1500

Time [minutes]

Fig. 3. State of charge of battery for Case 1.

1
0.9
0.8
0.7

Ppv
Pw
Pgas
Pbat
Pload
Surplus

1.5

Charging mode

2.5

Generation-Load profile [MW]

0.6
0.5
0.4
0.3
0.2

0.1
0

0.5

500

1000

1500

Time [minutes]

Fig. 4. Charging mode of the battery for Case 1.


0

-0.5

500

1000

1500

Time [minutes]

Fig. 2. Generation scheduling in the microgrid for Case 1.

Figure Fig. 3 shows the battery state of charge. Since the


battery is set to operate initially in the discharging mode, it
continues to produce energy until it reaches the minimum
accepted SoC, then starts charging when the energy generation
from renewables is greater than the load. The full charging of
the batter takes too long for two reasons: there generation
from renewables is mainly smaller than the load, and the
maximum capacity of the battery might be too large.
The change of the operating mode is illustrated in Figure
Fig. 4. When operating in the charging mode, a Cbat index is
set to 1, while when operating in discharging mode the index
is set to 0. When using Algorithm 1, the number of
charging/discharging cycles is reduced very much. This may
increase the lifetime of the battery but may lead to
unoptimized use of the gas engine unit.

B. Case 2
In order to overcome the inconvenience met in the Case 1,
we assume a smaller size battery, with the installed power of
0.3 MW. All the other parameters and characteristics were
maintained the same. Reduction of the battery size was
decided in order to allow achieving more charging/discharging
cycles.
The simulation with the new battery size has resulted in a
new generation scheduling, for the battery and the gas engine
unit (Fig. 6). The total energy produced by the gas engine in
the Case 2 was 5.37 MWh, and the energy not used from the
renewable sources was 0.437 kWh. Comparing these value
with the ones obtained in Case 1 we may say that when
Algorithm 1 is employed, smaller size batteries should be
used. However, the results depend on both the load profile and
the wind and PV generation profiles. In order to maximize the
use of renewable energy, more than one battery should be
used.

2.5
Ppv
Pw
Pgas
Pbat
Pload
Surplus

1.5

0.9
0.8
0.7
Charging mode

Generation-Load profile [MW]

0.5

0.6
0.5
0.4
0.3
0.2

0.1
0

-0.5

500

1000

500

1500

1000

1500

Time [minutes]

Time [minutes]

Fig. 7. Charging mode of the battery for Case 2.

Fig. 5. Generation scheduling in the microgrid for Case 2.

Figure Fig. 6 shows the battery state of charge for Case 2.


It is obvious that a smaller size battery exhibits more
charging/discharging cycles, mainly in the first half of the day
when the load is smaller than the power generation from
renewable sources. If several small batteries are used, one
battery can be scheduled to produce energy during the peak
load and charger during night periods or during high
availability of renewable sources. However, more batteries
requires larger investment costs. Beyond any short-term
optimization, a long term analysis to choose the number and
size of the batteries may be required.
0.9

0.7
2.5
Ppv
Pw
Pgas
Pbat
Pload
Surplus

0.6
2

0.5

0.4

0.3

0.2

500

1000

1500

Time [minutes]

Fig. 6. State of charge of battery for Case 2.

Figure Fig. 7 shows the charging mode changes for Case 2.


In the second half of the day the charging mode do not change
because the power generation from renewable sources is
smaller than the load.

Generation-Load profile [MW]

Battery - state of charge [-]

0.8

C. Case 3
The third case considers the same data used for Case 1.
The second scheduling algorithm is employed for this case,
which assumes that the battery is charged any time there is a
surplus of generation from renewable sources, and is
discharged any time the renewable sources cannot cover the
load. This algorithm allows thus the battery to change its
charging mode.
Figure Fig. 8 illustrates the generation scheduling for the
24 hours window. The figure shows that there is no waste
energy since the surplus line is always zero. On the other
hand, the battery is used mainly in the half part of the day
because it was fully charged during the night while the load
was also at low values. The gas engine is used during the peak
load because the PV system is not available, and the wind
system is not capable of covering the entire load.

1.5

0.5

-0.5

500

1000

1500

Time [minutes]

Fig. 8. Generation scheduling in the microgrid for Case 3.

Figure Fig. 9 shows the state, SoC, of charge of the


battery. Since renewable generation is available in the half

part of the day, the battery changes from charging mode to


discharging mode to maximize the use of renewable energy.
On the second half of the day the battery remains discharged
because the load is higher than the renewable generation.
0.9

Battery - state of charge [-]

0.8

0.7

0.6

0.5

0.4

0.3

0.2

500

1000

1500

Time [minutes]

Fig. 9. State of charge of battery for Case 3.

The changes between charging and discharging modes for


the battery is also observed in Figure Fig. 10.

control implemented at the microgrid level allows a local


generation-load balancing thus reducing the negative effects
of the intermittency shown by RES. In order to fully benefit
from the availability of the RES generation, storage batteries
can be used.
Two algorithms have been used for simulations. The first
algorithm assumes that if the battery operates in charging
mode it is not allowed to produce energy, whereas when
operating in discharging mode it is not allowed to absorb
energy from the network. The second algorithm has no
restriction on the operating mode of the battery. The battery
can be switched to charging or discharging mode any time is
necessary in order to fully use the energy from the renewable
sources.
The simulations have shown that the second algorithm
achieves minimum generation from the gas engine unit and
thus less fuel. On the other hand, the first algorithm involves a
smaller number of charging/discharging cycles. It is important
to note that, when employing the first algorithm a smaller size
battery should be used.
The simulations have been performed using a code
implemented in Matlab, using also the same input data. It is
expected that if different load profile or RES generation
profile are used, more conclusions may be drawn regarding
the size and the strategy for battery use may be required.

REFERENCES

0.9

[1]

0.8

Charging mode

0.7

[2]

0.6
0.5

[3]

0.4

[4]

0.3
0.2

[5]

0.1
0

500

1000

1500

Time [minutes]

[6]

Fig. 10. Charging mode of the battery for Case 3.

This case required the use of the gas engine mostly during
the peak load period, and the total generation provided was
4.98 kWh. Compared with the first two cases, this case
ensures the smallest cost for additional fuel. However, the
battery exhibit a large number of charging/discharging cycles.
V. CONCLUSIONS
The microgrid concept is a way of integrating the
renewable energy resources into the electrical networks. The

[7]

[8]

[9]

N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, Microgrids,


IEEE Power Energy Mag., vol. 5, no. 4, pp. 7894, Jul./Aug. 2007.
B. Lasseter, Microgrids [distributed power generation], in Proc. IEEE
Power Engineer. Soc. Winter Meet., Jan. 2001, vol. 1, pp. 146149.
C.A. Canizares, et al., Trends in Microgrid Control, in IEEE
Transactions on Smart Grid, vol. 5, no. 4, pp. 1905-1919, July 2014.
C.-H. Yoo, I.-Y. Chung, H.J. Lee, and S.-S. Hong, Intelligent Control
of Battery Energy Storage for Multi-Agent Based Microgrid Energy
Management, Energies, ISSN 1996-1073, 2013, vol. 6, pp. 4956-4979,
2013.
Z. Miao, L. Xu, V. R. Disfani, L. Fan, An SOC-Based Battery
Management System for Microgrids, IEEE Transactions on Smart Grid,
vol. 5, no. 2, March 2014.
G. Foggia, B. Delinchant, N. Hadjsaid, F. Wurtz, Optimization by
stochastic programming for the aggregation of a commercial virtual
power plant, XI-th International Workshop Optimization and Inverse
Problems in Electromagnetism, Sofia, Bulgaria, 14-18 September 2010.
E. Mashhour, S.M. Maghaddas-Tafreshi, Bidding strategy of Virtual
Power Plant for participating in energy and spinning reserve markets?
Part I: Problem Formulation, IEEE Transactions on Power Systems,
Vol. 26, No. 2, pp. 949-956, May 2011.
E. Mashhour, S.M. Maghaddas-Tafreshi, Bidding strategy of Virtual
Power Plant for participating in energy and spinning reserve markets?
Part I: Numerical Analysis, IEEE Transactions on Power Systems, Vol.
26, No. 2, pp. 957-964, May 2011.
V. A. Boicea, Energy Storage Technologies: The Past and the Present,
in Proceedings of the IEEE, vol. 102, no. 11, pp. 1777-1794, November
2014.

Das könnte Ihnen auch gefallen