IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO.

2, JUNE 2014 507

A Novel Approach for Ramp-Rate Control of Solar

PV Using Energy Storage to Mitigate Output
Fluctuations Caused by Cloud Passing
M. J. E. Alam, Student Member, IEEE, K. M. Muttaqi, Senior Member, IEEE, and D. Sutanto, Senior Member, IEEE

Abstract—The variability of solar irradiance with a high ramp- such a high ramp-rate on the distribution system, the PV pen-
rate, caused by cloud passing, can create fluctuation in the PV out- etration was limited to 50% of the PV plant capacity [4]. The
put. In a weak distribution grid with a high PV penetration, this can one-second resolution irradiance data collected from Oahu Is-
create significant voltage fluctuations. Energy storage devices are
used to smooth out the fluctuation using traditional moving average land by NREL [5] show that the irradiance level can fluctuate
control. However, moving average does not control the ramp-rate with a ramp-rate of more than 50% between two consecutive
directly; rather the ramp-rate depends on previous values of PV measurements. A PV output ramp-rate of 20% of the rated ca-
output. This paper proposes a strategy where the ramp-rate of pacity/second is recorded in a study [6] conducted on a system
PV panel output is used to control the PV inverter ramp-rate to a in Mesa del Sol, New Mexico. In Australia, the Commonwealth
desired level by deploying energy storage (which can be available
for other purposes, such as storing surplus power, countering volt- Scientific and Industrial Research Organisation (CSIRO) has re-
age rise, etc.). During the ramping event, the desired ramp-rate ported [7] that the PV output ramp-rates can be up to 8% of the
is governed by controlling the energy storage based on an inverse rated capacity/second at various locations. With a high penetra-
relationship with the PV panel output ramp-rate to improve the tion of grid-connected PV resources in weak radial distribution
fluctuation mitigation performance. In contrast to the moving av- systems, such high ramp-rate variations of the PV output can
erage method, the proposed strategy is able to control the desired
ramp-rate independent of the past history of the PV panel output. introduce significant voltage fluctuations [8]. Therefore, ramp-
A dynamic model of the PV-storage integrated system is developed rate control strategies to reduce fluctuations in PV outputs and
to verify the proposed strategy in the presence of physical device the resulting voltage fluctuations in PV integrated distribution
time lags. The proposed strategy is verified using simulation results systems are necessary in order to increase the PV penetration
based on an Australian distribution system. A laboratory exper- level in the networks.
iment is also conducted to validate the concept of the proposed
control strategy. Different kinds of energy storage technologies have been pro-
posed for smoothing of PV output, such as, electric double-layer
Index Terms—Energy storage, fluctuation, ramp-rate control, capacitor (EDLC) [9]–[11], superconducting magnetic energy
solar PV.
storage [12], fuel cell [13], and battery energy storage [14], [15].
I. INTRODUCTION A moving average-based ramp-rate control is proposed in [9]
for smoothing PV output fluctuations using EDLC. The moving
OLAR photovoltaic power generation can be highly vari-
S able in nature due to the irregularities in the sun irradiance
level caused by cloud passing. Although the investigation on
average method was also used in [14] for controlling battery
energy storage to reduce PV fluctuations. An exponential mov-
ing average has been used in [16] that gives more weight on
PV output variability has been continuing for a long time as the recent values of the fluctuating PV output. A modified Euler
evidenced in [1] and [2], it is still being addressed as an issue type moving average model is proposed in [11] for prediction
of contemporary interest [3], especially with a high penetration of moving average values for a hybrid system with PV, fuel
level of PV resources. Recent operating experiences from La cell, EDLC, and battery. Controlling an energy storage device
Ola Island PV plant [4] has revealed that the ramp-rate of the using traditional moving average method can reduce PV output
PV output can be as high as 63% of the rated capacity/minute, fluctuation, but it may not necessarily control the PV output to a
whereas it was intended to limit the ramp-rate up to 30% of desired ramp-rate. Also, a moving average with a long averaging
the rated capacity/minute [4]. To avoid any negative impact of window would require a storage device to operate to match the
difference between the actual PV output and the moving aver-
age, even if the actual PV output is not significantly fluctuating.
Manuscript received September 3, 2013; revised December 14, 2013; ac- The battery and charging facility available in an electric-vehicle
cepted January 27, 2014. Date of publication March 11, 2014; date of current
version May 15, 2014. This work was supported by the Australian Research has been used in [15] to mitigate the short-term intermittency
Council (ARC) and Essential Energy Linkage under Grant LP100100618. Pa- using a high-pass filter for fluctuation mitigation. In this method,
per No. TEC-00517-2013. the ramp-rate of PV inverter is limited using the characteristics
The authors are with the Australian Power Quality and Reliability Center,
School of Electrical Computer and Telecommunications Engineering, Univer- of the filter, such as the corner frequency. A dynamic filtering
sity of Wollongong, NSW 2522, Australia (e-mail: [email protected]; controller and dynamic rate limiter approach is used in [17] for
[email protected]; [email protected]). smoothing of PV and wind power generation fluctuations.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. In [18], the power delivered/absorbed by energy storage
Digital Object Identifier 10.1109/TEC.2014.2304951 for ramp-rate control is determined using a battery state of

0885-8969 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
508 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

charge (SoC) versus allowable ramp-rate droop characteristic

for generic renewable resources, where the allowable ramp-rate
is related to the SoC, but not directly related to the renewable
resource ramp-rate. An application of this strategy for a utility
scale PV plant is demonstrated in [19]. Fluctuation mitigation
using moving average control and a first-order low-pass filter is
presented in [20] where the SoC of the energy storage device
is maintained within a range by modifying the energy storage
power output using a gain parameter, which requires tuning.
The impact of different storage time constants has also been
investigated using time constants up to 30 s. Abdollahy et al. Fig. 1. Actual PV output and a 20-min moving average.
in [6] have applied both ramp-rate control and moving average
method for intermittency mitigation of the aggregate output of a moving average method for smoothing PV output fluctuations
distribution system with several distributed resources including in [14] and [20]. Although this is a traditionally accepted way
an energy storage system. There is no unique strategy for PV to reduce the fluctuations in PV output, it requires the energy
output smoothing, as it depends on many factors including cost storage device to inject/absorb power even if the PV output is
and response time of the compensating resource and system op- not fluctuating significantly. Also, such moving average strat-
eration policy [6]. Also, the delay in data transmission can cause egy does not control the ramp-rate directly. These points are
less smooth output or may even add more fluctuation to the ag- discussed later.
gregate output [6]. However, if the fluctuation is mitigated at the It is well known that moving average has a memory effect.
PV connection point with local information, the transmission of Depending on the length of the averaging window, a moving
data would not be necessary. average would highly depend on the past history and therefore,
The contribution of this paper is to propose a new method for may be significantly different than the present value of the fluctu-
mitigation of PV inverter output fluctuation by ramp-rate control ating variable. For PV output smoothing, the integrated storage
using energy storage. The energy storage system may be avail- device will have to operate to make-up for the aforementioned
able for other purposes, such as storing excess energy for use difference between the actual PV output and the moving average.
when the sun is not available, mitigating voltage rise, etc., [21]. Numerically, a moving average with a w-length window con-
The novelty of the proposed approach is an algorithm to control tains only (1/w)% of the present value of the fluctuating variable.
the ramp-rate of the PV output fluctuation using the appropri- A realistic PV output from Oahu Island, Hawaii, on the 18th of
ate strategies during the ramping and post-ramping event. Dur- March 2010, available from NREL [5] and its 20-min moving
ing a ramp-down or ramp-up event, the energy storage system average is shown in Fig. 1. Although the PV output does not
is controlled based on an inverse characteristic of the desired fluctuate significantly during the time periods in the red circles,
ramp-rate of PV inverter with respect to the ramp-rate of the PV the moving average values during these times still deviate sig-
panel output to improve the fluctuation mitigation performance. nificantly from the actual PV outputs due to the memory effect
Once the ramping event is over, the energy storage system will of the moving average method. To make-up for this deviation,
be controlled using a SoC droop characteristic to consider the energy storage device needs to operate more than necessary.
available capacity in selecting the desired ramp-rate. Unlike the Further, with a given window length w, the moving aver-
traditional moving average method, the proposed strategy does age method may not provide a desired ramp-rate; rather the
not have a memory effect and therefore the storage devices can ramp-rate will depend on the past history of the fluctuating PV
be controlled according to the requirement at each instant of output. The ramp-rate of the moving average of a variable PV
time without being influenced by the PV panel output at the output at the kth time instant can be defined using the following
previous instants or without affecting the compensating power expression:
at a future time instant. A dynamic model of the PV-storage
integrated system is developed and integrated with the distribu- M oRR(k)
tion network model to test the validity of the proposed strategy w −1  −1
1 PDC (k − i) − w i=0 PDC (k − 1 − i)
in a dynamic environment with physical device time-lags. Volt- = × i=0
age fluctuation mitigation achieved by applying the proposed w t (k) − t (k − 1)
control strategy is also investigated. The proposed control strat- 1 PDC (k) − PDC (k − w)
egy is verified using a daylong quasi-steady-state and dynamic = × (1)
w t (k) − t (k − 1)
simulation based on an Australian distribution system and also
using a basic experimental setup. where MoRR(k) is the ramp-rate of the moving average of PV
panel dc output, w is the length of the moving average window
in terms of the number of data points, PDC is the PV panel
II. LIMITATIONS OF TRADITIONAL MOVING AVERAGE dc output and t(k) is the time corresponding to the kth time
METHOD IN RAMP-RATE CONTROL
instant. Observing (1), it is clear that the ramp-rate obtained by
Traditionally, moving average is used to smooth out a fluc- following a moving average of the PV output depends on the PV
tuating variable. Energy storage device has been controlled by output at the kth and (k − w)th instant. Although, a significantly
ALAM et al.: NOVEL APPROACH FOR RAMP-RATE CONTROL OF SOLAR PV 509

Observing (4) it is apparent that when the ramp-rate of PDC

is zero, considering the dead-band function mentioned in mod-
ification (a), the ramp-rate of PCOM P will be governed by the
desired ramp-rate of PINV . However, when the ramp-rate of PDC
is zero, then if PCOM P has already attained such a level that re-
moval of PCOM P would not create an unacceptable fluctuation
Fig. 2. Conceptual schematic of the proposed ramp-rate control strategy.
in PINV , then PCOM P does not need to be changed anymore. To
accommodate this, a switching function S is incorporated with
low ramp-rate can be obtained by following a moving average (4), as given in following expression:
⎧   
with a long window, it is not directly related to the actual desired 
⎨ 0, if |PCOM P | < dPINV  and f
dPDC
=0
level.
S= dt des dt
⎩
III. PROPOSED RAMP-RATE CONTROL STRATEGY 1, otherwise.
(6)
A. Concept of the Proposed Ramp-Rate Control Strategy The modified control equation can be described using the
following expression:
In the proposed ramp-rate control strategy, a compensation      
power will be added with the PV panel dc power to control the dPCOM P 1 dPINV  dPDC
=S× × −η ×f .
dt des
INV
ramp-rate of the PV inverter output within a desired level at any dt ηINV dt
instant of time. The basic concept of ramp-rate control strategy (7)
is explained in Fig. 2. The inverter efficiency ηINV will depend on the power output
Based on Fig. 2, the PV inverter output PINV can be obtained of the inverter. The efficiency curve for a given inverter over a
from the PV panel dc power PDC and the compensation power range of operating power can be obtained from the manufacturer
PCOM P using the following expression: or developed through testing. As the desired PINV at any given
instant is known, ηINV at the desired PINV can be taken from the
PINV = ηINV × (PDC + PCOM P ). (2) efficiency curve and can be used in (7) to control the ramp-rate
In Fig. 2 and (2), ηINV is the inverter efficiency. Taking time of PCOM P at the necessary level.
derivative of (2) at both sides, the ramp-rate of PINV can now Considering a discrete time representation, the ramp-rate of
be related with the ramp-rate of PCOM P and PDC , as PDC at the kth instant can be determined using the following
  expression:
dPINV dPDC dPCOM P
= ηINV × + (3) dPDC [PDC (k) − PDC (k − 1)]
dt dt dt (k) = (8)
dt t (k) − t (k − 1)
where dPdtI N V is the PV inverter output ramp-rate, dPdtD C is the where t(k) is the time at the kth instant.
PV panel dc power output ramp-rate, and dP CdtO M P is the com- The value of the switching function is also incorporated with
pensation power ramp-rate. PCOM P to force it to zero when the conditions of S = 0 are sat-
To control the ramp-rate of the PV inverter output within a isfied, as given in (6). PCOM P at the kth instant can be obtained
desired level, the ramp-rate of PCOM P can be controlled using using
the expression derived from (3), as follows: 
   dPCOM P
PCOM P (k) = S × PCOM P (k − 1) + (k)
dPCOM P 1 dPINV  dPDC dt
= × − η × (4)
dt des 
INV
dt ηINV dt
 × {t (k) − t (k − 1)} . (9)
where dPdtI N V des is the desired ramp-rate of the PV inverter,
which will be negative during a ramp-down event and positive To illustrate the proposed ramp-rate control strategy, a ficti-
during a ramp-up event. tious PV output fluctuation shown in Fig. 3(a) is used, where the
The expression given in (4) is the basic control equation that PV panel output decreases at a ramp-rate of PVRR1 kW/s from
will be followed to control the ramp-rate of PV inverter output. time t1 to t2 , remains constant at the level reached at time t2 up
However, the following modifications are made to accommodate to time t3 , and again increases from time t3 to time t4 at a ramp-
practical situations. rate of PVRR2 kW/s. After reaching time t4 , the PV output re-
To allow for the slow variations of PDC with the progress mains constant. To control the ramp-rate of PINV , (7) and (9) are
of the day, a dead-band function “f ” is incorporated with (4) used to produce the appropriate amount of PCOM P . The desired
that will force the ramp-rate of PDC to be zero if it is within an ramp-rate of PINV is defined as the maximum allowable ramp-
acceptable limit, defined using the desired ramp-rate of PINV , rate (MARR) in W/s. To apply the proposed strategy to any size
as follows:
⎧    of PV plant, the ramp-rate units can also be defined in percent of
 dPDC  dPINV 
  ⎪ ⎪
⎨ 0, if η × 
 dt 
 ≤ the rated PV capacity/s. Before time t1 , PDC remains at a stable
dt des
INV
dPDC state. Therefore, the ramp-rate of PCOM P is zero according to
f =
dt ⎪
⎪ (7), and remains zero. From time t1 to t2 , PDC sharply decreases
⎩ dPDC , otherwise.
with a ramp-rate of PVRR1 . During this time the ramp-rate of
dt
(5) PCOM P is determined using PVRR1 and MARR. At time t2 ,
510 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

deviation will remain in effect for all future time instants even
if it is not necessary because of acceptable variations in PDC .
In contrast, according to the proposed ramp-rate control strat-
egy, the amount of PCOM P to control the ramp-rate at a given
kth instant can be governed by the desired ramp-rate of PINV for
the kth instant only and it does not impact the future ramp-rates.
According to (9), PCOM P at the kth time instant is
   
dPINV 
PCOM P (k) = S × PCOM P (k−1)+φ (k) .
dt des
(12)
In (12), φ(·) denotes a function that determines the ramp-rate
of PCOM P for obtaining the desired ramp-rate of PINV at the
given ramp-rate of PDC . The value of φ(·) is obtained using
(7). As the proposed strategy does not have a memory effect,
PCOM P can be changed at each instant dynamically based on
the requirement. This advantage is used in this paper to improve
Fig. 3. An illustration of the proposed ramp-rate control strategy. (a) A hy- the fluctuation mitigation performance.
pothetical PV output fluctuation. (b) PV inverter output with appropriately
controlled P C O M P to limit the P IN V ramp-rate within a desired level.
B. Improvement of the Fluctuation Mitigation Performance
PVRR1 becomes zero, but PCOM P is not zero yet. Therefore, For a given PV output fluctuation, the duration of the ramp-
the ramp-rate of PCOM P from time t2 to t3 is determined using ing event is inversely proportional to the ramp-rate of the PV
MARR only. From time t3 to t4 , PDC sharply increases with output, because the higher the ramp-rate, the faster the PV out-
a ramp-rate of PVRR2 . During this time, PCOM P is controlled put changes from one stable state to another stable state. As
with a ramp-rate determined using PVRR2 and MARR. After the time duration for a high-rate ramping event is very short,
time t4 , PVRR2 becomes zero, however, PCOM P is not zero yet. the amount of energy needs to be used for ramp-rate control is
Therefore, from time t4 PCOM P is controlled using MARR only. small. Therefore, during the ramping event, the MARR can be
At time t5 , PCOM P becomes zero and the ramp-rate of PDC is reduced in an inverse proportion to the PV panel output ramp-
also zero. Therefore, the ramp-rate of PCOM P is zero after t5 rate (PVRR) to improve the fluctuation mitigation performance
which keeps PCOM P at zero. without discharging a significant amount of storage capacity.
For moving average control, PCOM P at a given kth instant Such an inverse characteristic is given in (13) and shown in
depends on the previous values of PDC within the averaging Fig. 4(a)
window w as ⎧
⎨ |PVRR (k)| , if |PVRR (k)| < RRlim
w −1 MARRρ (k) = γ
PDC (k − i) ⎩ , if |PVRR (k)| ≥ RRlim .
PCOM P (k) = i=0 − PDC (k) . (10) |PVRR (k)|
w
(13)
Therefore, if moving average is applied to control the ramp- In (13), MARRρ is the inverse characteristic-based desired
rate of the fluctuation shown in Fig. 3(a), then PCOM P at t > ramp-rate, to be applied during the ramping event and γ is
t1 will actually depend on the PDC values before t1 . It will not a factor to control the degree of inverse characteristic; RRlim
directly depend on the required PCOM P to control the ramp-rate is a threshold of ramp-rate of PDC beyond which the inverse
at a desired level. According to (10), PCOM P is only control- characteristic is applied. RRlim should be set to a value higher
lable using the length of the averaging window w. The higher than the ramp-rate of slow variations of PDC to allow PINV to
the number w is chosen, the more is the smoothing effect. How- vary with the progress of the day. Once the ramping event is over,
ever, if a high w is chosen for improving fluctuation mitigation the PCOM P can be brought to zero with an MARR determined
performance at the kth instant, it not only impacts the PCOM P using a SoC droop characteristic as given by (14) and shown in
at the kth instant, but also at all the future time instants. For Fig. 4(b)
example, if w is increased by Δw to increase the smoothing
effect at the kth instant, then at any future (k+j)th instant, the MARRσ (k) =
deviation of PCOM P due to the increase of w (i.e.,Δw) will be ⎧
⎪
⎪ MARRm in , if |ΔSoC(k)| < SoCLB
 ⎪
⎪
 w +Δ w −1 P (k + j − i) ⎪
⎪ DBM ARR
 DC ⎨ × [|ΔSoC(k)| − SoCLB ] ,
ΔPCOM P (k + j) =  i=0 MARRm in +
 w + Δw DBSoC (14)
⎪
⎪
w −1  ⎪
⎪ if SoCLB ≤ |ΔSoC(k)| ≤ SoCUB
 ⎪
⎪
i=0 PDC (k + j − i)  ⎩
− . (11) MARRm ax , if |ΔSoC(k)| > SoCUB .
w 
In (14), MARRσ is a SoC droop-based desired ramp-rate to be
The deviation ΔPCOM P at the (k+j)th instant is solely due applied once the ramping event is over; DBM ARR and DBSoC
to the increase of w at the kth instant. For a given Δw, this are the dead-bands of MARR and SoC, respectively; SoCLB
ALAM et al.: NOVEL APPROACH FOR RAMP-RATE CONTROL OF SOLAR PV 511

Fig. 5. Energy utilization in the proposed ramp-rate control strategy. (a) Dur-
ing negative ramp-rate control. (b) During positive ramp-rate control.

PVRR, followed by a low PV event E2 (time t2 to t3 ). During

the event E1 , MARRρ is applicable which is determined by γ. If
γ = 1, MARRρ is in direct inverse relation with the ramp-rate of
PDC . If γ < 1, the degree of the inverse characteristic increases,
and therefore, PINV is improved compared to γ = 1. However,
this will be performed at the expense of additional energy dis-
charged from the storage device as shown in Fig. 4(d) using the
shaded region. If γ > 1, the degree of the inverse characteristic
decreases, and therefore, PINV is reduced compared to γ = 1,
but at the same time less amount of energy is discharged from
the storage device, as shown in Fig. 4(d). Although for high
ramp-rate events, the variation of storage energy discharge is
Fig. 4. Ramping event dependent selection of MARR. (a) During the ramping not significant, it provides the user a control over the fluctuation
event. (b) After the ramping event. (c) Application of MARRρ and MARRσ . mitigation performance of the proposed strategy.
(d) The effect of γ and MARR versus SoC droop on ramp-rate control.
During the event E2 , PCOM P slowly decreases at MARRσ
determined from the SoC droop characteristic in (14). For a
given fluctuation, MARRm in is selected as MARRρ so that PINV
and SoCUB are the lower and upper bands of SoC, respectively,
can be reduced at the same ramp-rate as it reduces during event
and |ΔSoC(k)| is the absolute value of the deviation between
E1 , which provides a smooth transition in ramp-rate control
SoC(k–1) and a user defined reference SoCref . Using the ab-
from event E1 to E2 . However, if |ΔSoC| is higher than SoCLB ,
solute deviation makes the droop characteristic applicable for
then MARRσ is increased and therefore, PINV reduces at a
change of SoC in either direction.
higher ramp-rate as shown in Fig. 4(d). MARRm ax in (14) needs
Fig. 4(c) identifies the ramping events in a real PV output fluc-
to be set to a value that allows for the slow variations of PDC ,
tuation and shows the event driven application of MARRρ and
as mentioned for selecting RRlim in (13).
MARRσ to control the ramp-rates. SoCLB defines the amount of
The discussions presented above provide only basic guide-
energy available for controlling the ramp-rate at MARRm in be-
lines for selection of the control parameters for the proposed
fore entering the droop characteristic. Therefore, SoCLB should
strategy. In general, these parameters are user defined and can
be set to a value that represents a sufficient amount of energy
vary depending on the scenario.
to control the ramp-rate at MARRm in (during the worst-case
fluctuation) for a certain amount of time that allows for any
back up power to come online or any sensitive load to be shut C. Energy Utilization for the Control of Ramp-Rate
down safely. SoCUB should be set up to a value that represents The amount of energy to be utilized by an energy storage
a substantial portion of the energy needed to control the ramp- device for controlling the ramp-rate will mainly depend on the
rate (during the worst-case fluctuation) before exceeding the PV panel output power PDC just prior to start of the fluctuation,
saturation limit of the droop characteristic (MARRm ax ). and the maximum allowable ramp-rate MARR. A relation of
The effect of the parameter γ in (13) and the parameters the energy usage for ramp-rate control to the rated output of a
MARRm in and MARRm ax in (14) on the fluctuation mitigation PV panel is established and discussed later using Fig. 5.
performance is shown in Fig. 4(d) using a fictitious PV fluctua- To consider a worst-case fluctuation, a PV output power hav-
tion containing a ramping event E1 (time t1 to t2 ) at a ramp-rate ing a sudden fall from the rated value of PDC-rated to zero at
512 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

Fig. 7. Flowchart of the proposed ramp-rate control strategy.

Fig. 6. Variations of the energy utilization for ramp-rate control with different
rated PV output at different MARR.

time t1 is shown in Fig. 5(a). In this case, it is assumed that PDC

does not return to its previous value due to a likely phenomenon
of a sustained cloudy period, after time t1 . According to the pro-
posed strategy, the PV inverter output will be slowly reduced at
the maximum allowable ramp-rate MARR and will become zero
at time t2 , instead of suddenly dropping to zero. The value of
MARR will be determined depending on the situation as shown
Fig. 8. Dynamic model of the PV-storage integrated system to consider the
in Fig. 4. The energy to be discharged by the storage device in effect of physical device time-lags.
controlling the ramp-rate is shown in Fig. 5(a) using a shaded
area. A similar type of worst-case scenario for positive ramp-rate
is shown in Fig. 5(b), where the PV output suddenly increases rate is replaced with MARRρ . If PVRR(k) is found lower than
from zero to PDC-rated at time t1 . In this case, the positive ramp- MARRρ , then a test is performed to find if removal of PCOM P
rate of the PINV is controlled by charging the energy storage would create an unacceptable fluctuation in PINV . If “Yes,”
with certain amount of energy, as indicated using a shaped area then S is set to 1 and the ramp-rate of PCOM P is determined
in Fig. 5(b). The energy utilized for ramp-rate control, ERRC using (7), but in this case the desired ramp-rate is replaced with
(kWh) can be derived using the following expression: MARRσ , determined from (14). If “No,” then the ramp-rate of
N t o t PCOM P is determined using (7) with S = 0. The ramp-rate of
(PDC-rated − i × MARR)
ERRC = i=1 (15a) PCOM P (k) is applied to determine PCOM P (k) using (9). With
3600 the value of PCOM P (k) obtained, the present level of SoC is
where Ntot is the total number of time-steps between time t1 checked. If it is found within the lower (SoCLL ) and upper
(seconds) and t2 (seconds), which can be determined using (SoCUL ) limits of SoC, obtained using the maximum depth of
PDC-rated discharge (DoDm ax ), the maximum state of charge (SoCm ax ),
Ntot = . (15b) and buffer energy EBuff , then a charge/discharge operation is
MARR
performed depending on the sign of PCOM P (k). According to
In (15a), 3600 is used to bring the unit to kilowatthour, as the
the proposed strategy, a +ve sign means discharge, and −ve
calculations are being performed in per second unit. A plot of the
sign means charge. Once the discharging/charging decision for
ERRC for rated PV capacities up to 5 kW is derived using (15)
the time instant is made and the storage is put into operation,
for different values of MARR and is shown in Fig. 6. Increase
the control system proceeds with a transition into the next time
of energy utilization for ramp-rate control with increase in PV
instant.
capacity and decrease in MARR is observed in Fig. 6.
Although this paper has mainly concentrated on the applica-
Instead of using PDC-rated , PDC (k) can be used in (15) to
tion of battery energy storage, the fundamental idea of ramp-rate
determine the energy required to control the ramp-rate from
control proposed in this paper would be applicable for other type
the present value of PDC . The amount of energy determined
of energy storage (such as ultra-capacitor).
in this way can be identified as an energy buffer (EBuff ) to
approximately check the minimum amount of capacity need to
be present in the storage to control the ramp-rate according to E. Dynamic Model of PV-Storage Integrated System
the proposed strategy.
The development of a dynamic model is necessary to under-
stand the performance of the proposed strategy in the context of
D. Flowchart of the Proposed Control Strategy practical system components that are subject to physical device
The flowchart of the proposed ramp-rate control strategy for time lags. Detailed switching models of the power electronic
a given kth time instant is shown in Fig. 7. The ramp-rate of converters are used in power-electronics-based investigations
PDC at the kth instant PVRR(k) is calculated using (8), and where subcycle phenomena are of interest. However, for the
based on this MARRρ is calculated using (13). If PVRR(k) is scope of the investigations of this paper, such fast dynamics are
equal to or higher than MARRρ , S is set to 1 and the ramp-rate of not of interest. Therefore, a simplistic dynamic model of the
PCOM P is determined using (7) and in this case the desired ramp- PV-storage integrated system is developed, as shown in Fig. 8,
ALAM et al.: NOVEL APPROACH FOR RAMP-RATE CONTROL OF SOLAR PV 513

that represents the time-lag related to the associated devices and

components.
The dynamic model consists of four states which are the PI
controller output (y1 ), the storage power output (y2 ), storage
power output measurement signal (y3 ), and the PV inverter
output (y4 ). The state equations are as follows:
dy1 1
= −KP × × (y2 − y3 ) + KI
dt Tm eas
× (PCOM P-ref − y3 ) (16a) Fig. 9. LV test feeder with the PV-storage integrated systems.
dy2 1
= × (y1 − y2 ) (16b)
dt Tsto
dy3 1
= × (y2 − y3 ) (16c)
dt Tm eas
dy4 1
= × (ηinv [PDC + y2 ] − y4 ) (16d)
dt Tinv
where PCOM P-ref is the reference to the PI controller; Tinv is the
inverter time constant; Tsto is the storage time constant; Tm eas
is the storage output measurement circuit time delay; KP and
KI are proportional and integral gain of the PI controller.
It is necessary that the energy storage device responds to the
charging/discharging commands as fast as possible to control
the ramp-rate satisfactorily. Studies have shown that battery
energy storage devices, such as NaS, Li-ion, and valve-regulated
lead-acid (VLRA) batteries, have response time in the order of
milliseconds [22], [23]. Therefore, it is expected that with an
appropriate selection of the energy storage device, the ramp-
rate can be controlled according to the requirements. Using
the developed dynamic model, the impact of different storage
response times on ramp-rate control can be investigated.
Fig. 10. PVRR during slow variation with the progress of the day. (a) Energy
The PI controller is tuned to have a low rise time so that used to control the worst case fluctuation and MARRσ characteristic.
the fast response time of the storage device can be effectively
used for fluctuation mitigation. The initial controller parameters
are obtained using the open-loop frequency response of the
simulation. Energy storage devices are modeled incorporating
linearized plant model. The tuning task is performed with the
the nonlinear voltage versus SoC characteristics and the effect
aid of Simulink automatic PID tuner [24] that performs the initial
of internal resistance, as described in [21]. The LV feeder under
tuning based on a reference tracking algorithm while keeping a
study is shown in Fig. 9. The results shown correspond to the
tradeoff between performance and robustness. Once the initial
PV output fluctuations in household no. 28 (HH 28) connected
parameters are obtained, further tuning is performed to reduce
to phase a, bus 10 of the LV feeder.
the rise time.
The PV output data correspond to the irradiance profile of
This paper has mainly discussed the concept of the proposed
the 18th of March, 2010, captured by NREL in the Oahu Island,
strategy and its system level application. To carry out a com-
Hawaii [4]. For simplicity, constant inverter efficiency (89.5%)
ponent level detailed design, practical issues, such as battery
is used in the simulations. To select the ramp-rate threshold limit
power losses, heat generated, and dc-link voltage fluctuations
to apply the inverse characteristic in (13), RRlim is selected by
caused by high current pulses during high-ramp rate operations,
observing the ramp-rate during slow variation of PDC from
need to be considered.
7:00 h to 9:00 h. During this time frame, more than 99% of
the ramp-rates were less than 3 W/s, as shown in Fig. 10(a).
IV. APPLICATION EXAMPLE OF THE PROPOSED
Therefore, to allow for this variation, RRlim is set to a higher
RAMP-RATE CONTROL STRATEGY value of 5 W/s. MARRm ax in (14) is also set to 5 W/s for the
To analyze the performance of the proposed strategy in a dis- same purpose. Simulation trials have been run using a range of
tribution network environment, a low voltage (LV) distribution 0.1 ≤ γ ≤ 50. For γ ≤ 1, fluctuation mitigation performance
feeder extracted from the New South Wales distribution system does not improve much, whereas with γ > 1, deterioration of the
is considered. All the households in the LV feeder are assumed fluctuation mitigation performance appears with γ ≥ 20. The
to have rooftop PV units with integrated battery storage devices. value of γ has been selected as 5 (with Watt to kW conversion:
A typical 12 kWh valve regulated lead acid battery is used for 0.005).
514 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

Fig. 11. Mitigation of PV output fluctuation. (a) Using moving average.

(b) Using the proposed ramp-rate control strategy.

The storage devices are allowed to operate between a typ-

ical range of 40% to 100% SoC. To allow for charging and
discharging for ramp-rate control, SoCref is set to 70%, which
is at the mid-point of the allowable range of SoC. The SoC
bands, SoCLB and SoCUB , are selected considering the worst-
case fluctuation of PDC from the rated value (4 kW for the PV
systems in the network under study) to zero at the worst-case
ramp-rate of PDC-rated /s and after the fall, PDC does not re-
cover due to a continuous cloudy period. The PDC profile with
the storage support is shown in Fig. 10(b), where the ramp- Fig. 12. Mitigation of PV output fluctuation. (a) P D C and P IN V profiles.
rate control is performed according to (14), as shown using the (b) P C O M P with moving average control. (c) P C O M P with the proposed
MARRσ profile. SoCLB is set to 2.5% which corresponds to strategy.
the amount of energy that can control the ramp-rate at the value
of MARRm in for about 5 min which is sufficient for bringing
emergency power online or disconnecting sensitive loads safely. that the moving average method needs to continue the storage
SoCUB is set to 5% which accounts for about 67% of the total operation due to the past history of PDC , even if it is not nec-
energy used for bringing PDC to zero at a controlled ramp- essary as the ramp-rate is within desired limit. This is observed
rate. The parameters used for the simulation of the proposed from PCOM P profiles in Fig. 12(b) for the moving average meth-
ramp-rate control strategy are given in Table I in the Appendix. ods and 12(c) for the proposed approach. In Fig. 12(c), zero
Initially, the results of the quasi-steady state simulations using values of PCOM P are highlighted using dotted circular shapes.
1-s resolution data are presented to investigate the validity of Examining the PCOM P profiles it is found that according to the
the strategy over a daylong variation. Later, dynamic simulation proposed strategy, the storage device operates for about 50%
results in milliseconds time-frame are presented. less time in the 10 h period (from 8:00 to 18:00 h) in com-
PV panel dc power, PDC (multiplied by the inverter effi- parison to the moving average method. This can contribute to
ciency), the storage compensation power PCOM P , and the in- increase the lifetime of a battery storage device.
verter output PINV are shown in Fig. 11(a) and (b) for the The percentage ratio of PINV to PDC based on the profiles
moving average control and the proposed strategy, respectively. in Fig. 12(a) is compared between the proposed approach and
For moving average control, a 20-min moving average [9] of moving average method in Fig. 13. Due to the dependency of
PDC is used. previous PDC values, PINV obtained from the moving average
In general, both of the methods (moving average and the method is different (lower) than PDC , even during the periods
proposed approach) can mitigate the fluctuations in PV output, when PDC exhibits very low fluctuation. In contrast, the pro-
as observed from Fig. 11. However, detailed investigations re- posed approach tracks PDC in those periods and therefore, the
veal the specific advantages of the proposed approach. Fig. 12(a) PINV to PDC ratio is higher using the proposed approach, as
shows PDC and PINV profiles in a shorter window, from 10:00 h compared to the moving average method.
to 13:00 h. According to the proposed strategy, PINV tracks PDC The ramp-rate profiles of PDC and PINV from 10:00 h to
when the ramp-rate of PDC does not exceed the desired ramp- 13:00 h are shown in Fig. 14(a); the ramp-rates of PINV can
rate at that time. Therefore, the proposed strategy does not need be controlled within a specified limit by operating the storage
to operate the storage at all the times. In contrast, (10) suggests device according to the proposed strategy. The ramp-rates of
ALAM et al.: NOVEL APPROACH FOR RAMP-RATE CONTROL OF SOLAR PV 515

Fig. 13. P D C to P IN V ratio.

Fig. 15. Usefulness of the proposed inverse characteristic in the improvement
of fluctuation mitigation.

Fig. 16. Mitigation of PCC voltage fluctuation using the proposed strategy.
(a) PCC voltage profile. (b) P C O M P profile.
Fig. 14. Ramp-rate control. (a) The ramp-rate profiles of P D C and P IN V .
(b) Comparison of the P IN V ramp-rates obtained using the moving average
control and the proposed strategy. MARR value of 5 W/s. The proposed inverse MARR and SoC
droop-based characteristics can mitigate the fluctuation better
PINV with 20-min moving average control of storage device than a constant MARR value. For reference, the fluctuation mit-
are compared with those obtained using the proposed strategy igation using the 20-min moving average is also included in
in Fig. 14(b) for the time period when the maximum negative Fig. 15.
and positive ramp-rates in PDC appeared. For reference, the The PCC voltage profile at household no. 28 (HH 28) of the
ramp-rate of PDC is also included in Fig. 14(b). During the neg- test feeder from 10:00 h to 13:00 h is shown in Fig. 16(a). For
ative fluctuation of PDC with a ramp-rate of −1.767 kW/s, the reference, PCOM P profile is also shown in Fig. 16(b). When
proposed strategy controls the PINV ramp-rate at −0.0028 W/s sharp decrease in PCC voltage appears as a result of sudden
which comes from the inverse characteristic in (13). With the decrease in PV output, energy storage device is discharged to
20-min moving average, the ramp-rate of PINV is −1.485 W/s, control the high negative ramp-rates. This action mitigates the
which is not controlled at this value; rather it is produced as a associated voltage dips. Again, when sharp rise in PCC voltage
result of the previous PDC samples over the last 20 min period. appears due to sudden increase in PV output, energy storage de-
During the positive ramping event at 1.725 kW/s, the ramp-rate vice is charged to control the positive ramp-rates and this action
with the proposed strategy is 0.0029 W/s which is controlled mitigates the sharp voltage rise as well. Without the proposed
using (13), whereas, with the moving average it is 0.0183 W/s, ramp-rate control strategy, the voltage ramp-rate at 10 h : 16 min
which is not actually controlled to this value. Although the mov- : 54 s (when the largest negative ramp-rate in the PV output ap-
ing average method provides significant fluctuation mitigation peared) is −9.3 V/s and at 10 h : 16 min : 58 s (when the largest
at the expense of operating the storage device all the time, the positive ramp-rate in PV output appeared) is 9.1 V/s. Control-
proposed strategy provides better mitigation during the time of ling the ramp-rate of PV output using the proposed strategy, the
fluctuation with a high ramp-rate (such as the one shown in PCC voltage ramp-rate is reduced to less than 0.1 mV/s.
Fig. 14). The validation of the proposed strategy in a dynamic environ-
The usefulness of the proposed MARR characteristics in (13) ment is tested using the proposed PV-storage integrated system
and (14) can also be observed in Fig. 15 where the PV output model using 1 ms time step and the results are presented in
fluctuation mitigation using the proposed inverse MARR char- Fig. 17. The largest ramp-rate observed in the PV output data
acteristic from 12:00 h to 13:00 h is compared with a constant used for the daylong simulation is used for this purpose. The
516 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

Fig. 17. Performance analysis using the proposed dynamic model. (a) Nega-
tive ramp-rate control. (b) Positive ramp-rate control.

Fig. 19. Experimental results (a) Data points for experimental verification.
(b) Captured current waveforms for points P 1 , P 2 , and P 3 .

Fig. 18. The waveforms are captured using an Agilent Infiniivi-

sion (Model: DSO7034 A) series oscilloscope.
A segment of the real PV generation profile used to obtain the
simulation results is used for the experiment where the highest
Fig. 18. Experimental setup for verification of the proposed strategy. negative and positive ramp-rate of PDC is included. A scale
factor is used to scale down the simulation data to comply with
the ratings of the equipment. The data points obtained from
dynamic simulation is performed with the data shown in Table II the simulation are shown in Fig. 19(a) that shows that from
in the Appendix. the second to the third data point, the PV output decreases at a
The PV output decreases at a ramp-rate of −1.767 kW/s at ramp-rate of −55%/s of the rated output, then remains at low
10 h : 16 min : 54 s. Fig. 17(a) shows that following the negative values of less than 15% of the rated output from the third to sixth
ramp-rate of PDC , PCOM P increases within a few milliseconds data point. Then, it increases at a ramp-rate of 53%/s of the rated
time to increase the PINV at the level that maintains the desired output from the sixth to the seventh data point. With the proposed
ramp-rate MARRρ specified at the time. After 4 s time, the PDC control approach, PCOM P is controlled using (13) from the
again increases with a ramp-rate of 1.725 kW/s at 10 h : 16 min second to the third data points and from the sixth to seventh data
: 58 s. Fig. 17(b) shows that following the positive ramp-rate of points (ramping events). From the third to the sixth data point
PDC , PCOM P decreases to the level necessary to maintain the (post-ramping event), PCOM P is controlled using (14). As a
desired ramp-rate of PINV at that time. The impact of the storage result of controlling PCOM P according to the proposed strategy,
time constant Tsto is also shown in Fig. 17; PINV with Tsto of the ramp-rate of PINV is maintained at such a low value over
5 ms settles down faster and with less swing as compared to the the period of fluctuation (first to seventh data point) that PINV
PINV with Tsto of 20 ms. Therefore, a storage device with a fast remains nearly constant, as shown in Fig. 19(a). The captured
response time is necessary for a satisfactory mitigation of the waveforms of the dc source current IDC , the battery current
fluctuation. ICOM P , and the inverter current IINV at the points (P1 ), (P2 ),
and (P3 ) identified in Fig. 19(a), are shown in Fig. 19(b). The
time differences among the data points are represented using
V. EXPERIMENTAL RESULTS
dashed lines. The decrease of IDC from P1 to P2 caused by cloud
An experimental setup with a controllable dc power source passing is compensated by the increase of ICOM P and therefore,
simulating the PV panel dc power, an Ultratech 12-V 7Ah IINV remains nearly constant from P1 to P2 . The increase of
sealed lead-acid battery, a charge/discharge controller (Model: IDC from P2 to P3 , when the cloud passing is over, makes
CM3024Z), a PowerBright 400 W, 12 V dc to 220 V ac inverter ICOM P to decrease. Therefore, IINV remains nearly constant
(Model: ERP400-12), and a Chroma programmable ac/dc elec- from P2 to P3 . The results from this experiment indicate that
tronic load (Model: 63804) is used to verify the main concept the proposed strategy would be able to control the ramp-rate of
of the proposed approach. The experimental setup is shown in PINV by appropriately controlling the PCOM P support.
ALAM et al.: NOVEL APPROACH FOR RAMP-RATE CONTROL OF SOLAR PV 517

TABLE I sachusetts photovoltaic experiment,” IEEE Trans. Energy Convers., vol. 4,

SIMULATION PARAMETERS OF THE PROPOSED CONTROL STRATEGY no. 2, pp. 184–190, Jun. 1989.
[3] C. Trueblood, S. Coley, T. Key, L. Rogers, A. Ellis, C. Hansen, and
E. Philpot, “PV measures up for fleet duty: Data from a tennessee plant
are used to illustrate metrics that characterize plant performance,” IEEE
Power Energy Mag., vol. 11, no. 2, pp. 33–44, Mar. 2013.
[4] J. Johnson, B. Schenkman, A. Ellis, J. Quiroz, and C. Lenox. (2012). San-
dia National Laboratories. Initial Operating Experience of the La Ola 1.2-
MW Photovoltaic System [Online]. Available: http://energy.sandia.gov/
wp/wp-content/gallery/uploads/SAND2011–8848_Lanai_DOE_Report_
TABLE II FINAL.pdf
PARAMETERS USED FOR DYNAMIC SIMULATION [5] NREL. 1-Second Global Horizontal Irradiance, Oahu, Hawaii [Online].
Available: http://www.nrel.gov/midc/oahu_archive/
[6] S. Abdollahy, A. Mammoli, F. Cheng, A. Ellis, and J. Johnson, “Dis-
tributed compensation of a large intermittent energy resource in a dis-
tribution feeder,” in Proc. IEEE PES Innovative Smart Grid Technol.,
Washington, DC, USA, Feb. 24–27, 2013, pp. 1–6.
[7] S. Sayeef, S. Heslop, D. Cornforth, T. Moore, S. Percy, J. K. Ward,
A. Berry, and D. Rowe. CSIRO. (Jun. 2012). Solar Intermittency:
VI. CONCLUSION Australia’s Clean Energy Challenge. Characterising the Effect of High
This paper has proposed a new approach for PV inverter ramp- Penetration Solar Intermittency on Australian Electricity Networks.
[Online]. Available: https://publications.csiro.au/rpr/download?pid=
rate control using an integrated energy storage device. Unlike csiro:EP121914&dsid=DS1
the traditional moving average method, which is biased by the [8] H. Sugihara, K. Yokoyama, O. Saeki, K. Tsuji, and T. Funaki, “Economic
previous values and does not directly control the ramp-rate at and efficient voltage management using customer-owned energy storage
systems in a distribution network with high penetration of photovoltaic
a desired level, the proposed method can limit the ramp-rate systems,” IEEE Trans. Power Syst., vol. 28, no. 1, pp. 102–111, Feb.
within a desired level. An inverse characteristic of the desired 2013.
ramp-rate with the PV panel output ramp-rate is proposed to im- [9] N. Kakimoto, H. Satoh, S. Takayama, and K. Nakamura, “Ramp-Rate
control of photovoltaic generator with electric double-layer capacitor,”
prove the fluctuation mitigation performance during the ramping IEEE Trans. Energy Convers., vol. 24, no. 2, pp. 465–473, Jun. 2009.
event. Once the ramping event is over, the desired ramp-rate is [10] T. Kinjo, T. Senjyu, N. Urasaki, and H. Fujita, “Output levelling of renew-
switched to a droop characteristic with SoC so that the available able energy by electric double-layer capacitor applied for energy storage
system,” IEEE Trans. Energy Convers., vol. 21, no. 1, pp. 221–227, Mar.
capacity is taken into account while controlling the ramp-rate. 2006.
Simulation results show that the mitigation of the PV output fluc- [11] T. Monai, I. Takano, H. Nishikawa, and Y. Sawada, “A collaborative
tuation achieved by the proposed ramp-rate control strategy is operation method between new energy-type dispersed power supply and
EDLC,” IEEE Trans. Energy Convers., vol. 19, no. 3, pp. 590–598, Sep.
comparable to those obtained using the moving average method, 2004.
while not using the energy storage all the time. Also, during very [12] K. S. Tam, P. Kumar, and M. Foreman, “Enhancing the utilization of pho-
high ramp-rate events, the proposed strategy can provide tighter tovoltaic power generation by superconductive magnetic energy storage,”
IEEE Trans. Energy Convers., vol. 4, no. 3, pp. 314–321, Sep. 1989.
control of the ramp-rate due to the inverse characteristic. The [13] S. Rahman and K. S. Tam, “A feasibility study of photovoltaic-fuel cell
mitigation of voltage fluctuation in weak networks caused by hybrid energy system,” IEEE Trans. Energy Convers., vol. 3, no. 1, pp. 50–
PV output fluctuation using the proposed method is also demon- 55, Mar. 1988.
[14] T. D. Hund, S. Gonzalez, and K. Barrett, “Grid-tied PV system energy
strated. The validity of the proposed strategy has been verified smoothing,” in Proc. Photovoltaic Spec. Conf., Honolulu, HI, USA, Jun.
in a dynamic environment using a dynamic model of the PV- 20–25, 2010, pp. 2762–2766.
storage integrated system developed in this paper. The results [15] J. Traube, L. Fenglong, D. Maksimovic, J. Mossoba, M. Kromer, P. Faill,
S. Katz, B. Borowy, S. Nichols, and L. Casey, “Mitigation of solar irradi-
from a laboratory experiment using a PV inverter, a battery ance intermittency in photovoltaic power systems with integrated electric-
storage system, and an electronic load demonstrates that the vehicle charging functionality,” IEEE Trans. Power Electron., vol. 28,
proposed approach is able to control the PV inverter ramp-rate no. 6, pp. 3058–3067, Jun. 2013.
[16] S. G. Tesfahunegn, Ø. Ulleberg, P. J. Vie, and T. M. Undeland, “PV fluc-
by providing appropriate support from the storage device. tuation balancing using hydrogen storage—A smoothing method for inte-
gration of PV generation into the utility grid,” Energy Procedia, vol. 12,
APPENDIX pp. 1015–1022, 2011.
[17] L. Xiangjun, H. Dong, and L. Xiaokang, “Battery energy storage station
See Tables I and II. (BESS)-Based smoothing control of photovoltaic (PV) and wind power
generation fluctuations,” IEEE Trans. Sustainable Energy, vol. 4, no. 2,
pp. 464–473, Apr. 2013.
ACKNOWLEDGMENT [18] Managing renewable power generation, by C. J. Coe, A. N. Hurst,
M. I. Hardin, M. C. Such, and R. T. Jennings. (Nov. 2011). U.S. Patent
Authors gratefully acknowledge the support of Mr. Sean El- 20110273129. [Online]. Available: https://docs.google.com/ viewer?url =
phick from Australian Power Quality and Reliability Centre in www.google.com/patents/US20110273129.pdf
conducting the laboratory experiments. [19] C. A. Hill, M. C. Such, C. Dongmei, J. Gonzalez, and W. M. Grady,
“Battery energy storage for enabling integration of distributed solar power
generation,” IEEE Trans. Smart Grid, vol. 3, no. 2, pp. 850–857, Jun. 2012.
REFERENCES [20] A. Ellis, D. Schoenwald, J. Hawkins, S. Willard, and B. Arellano, “PV
output smoothing with energy storage,” in Proc. Photovoltaic Spec. Conf.,
[1] W. Jewell and R. Ramakumar, “The effects of moving clouds on elec- Austin, TX, USA, Jun. 3–8, 2012, pp. 1523–1528.
tric utilities with dispersed photovoltaic generation,” IEEE Trans. Energy [21] M. J. E. Alam, K. M. Muttaqi, and D. Sutanto, “Mitigation of rooftop solar
Convers., vol. EC-2, no. 4, pp. 570–576, Dec. 1987. PV impacts and evening peak support by managing available capacity of
[2] E. C. Kern, E. M. Gulachenski, and G. A. Kern, “Cloud effects on dis- distributed energy storage systems,” IEEE Trans. Power Systems, vol. 28,
tributed photovoltaic generation: Slow transients at the Gardner, Mas- no. 4, pp. 3874–3884, Nov. 2013.
518 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 29, NO. 2, JUNE 2014

[22] H. Chen, T. N. Cong, W. Yang, C. Tan, Y. Li, and Y. Ding, “Progress in D. Sutanto (SM’89) received the B.Eng. (Hons.) and
electrical energy storage system: A critical review,” Progress Natural Sci., Ph.D. degrees from the University of Western Aus-
vol. 19, pp. 291–312, 2009. tralia, Crawley, Australia.
[23] B. Espinar and D. Mayer. International Energy Agency. (Jul. 2011). The He is currently a Professor of power engineering
Role of Energy Storage for Mini-Grid Stabilization, [Online]. Available: at the University of Wollongong, Wollongong, Aus-
http://www.iea-pvps-task11.org/HTMLobj-187/Act_24_Final.pdf tralia. His research interests include power system
[24] Mathworks. (Oct. 2013). PID Tuning Algorithm for Linear Plant Model, planning, analysis and harmonics, FACTS, and bat-
[Online]. Available: http://www.mathworks.com.au/help/control/ref/ pid- tery energy storage systems. He is currently the IEEE
tune.html IAS Area Chair for Region 10 (Asia Pacific).

M. J. E. Alam (S’10) received the B.Sc. and M.Sc.

degrees in electrical and electronic engineering from
Bangladesh University of Engineering and Technol-
ogy, Dhaka, Bangladesh, in 2005 and 2009, respec-
tively, with focus on electrical energy and power sys-
tems. He is currently working toward the Ph.D. degree
at the University of Wollongong, New South Wales,
Australia.
His research interests include modeling and anal-
ysis of power systems considering the impacts of
distributed and renewable energy resources. Prior to
starting PhD studies, he has been involved with the power industry in Bangladesh
for 4.5 years, where he worked in the field of power generation, transmission
and distribution.

K. M. Muttaqi (M’01–SM’05) received the B.Sc.

degree in electrical and electronic engineering from
Bangladesh University of Engineering and Tech-
nology (BUET), Dhaka, Bangladesh, in 1993, the
M.Eng.Sc. degree in electrical engineering from the
University of Malaya, Malaysia, in 1996, and the
Ph.D. degree in electrical engineering from Multi-
media University, Malaysia, in 2001.
Currently, he is an Associate Professor at the
School of Electrical, Computer, and Telecommunica-
tions Engineering, and member of Australian Power
Quality and Reliability (APQRC) at the University of Wollongong, Australia.
He was associated with the University of Tasmania, Australia as a Research
Fellow/Lecturer/Senior Lecturer from 2002 to 2007, and with the Queensland
University of Technology, Australia as a Research Fellow from 2000 to 2002.
Previously, he also worked for Multimedia University as a Lecturer for three
years. He has more than 17 years of academic experience and has authored or
coauthored over 165 papers in international journals and conference proceed-
ings. His research interests include distributed generation, renewable energy,
electrical vehicles, smart-grid, power system planning and control.
Dr. Muttaqi is an Associate Editor of the IEEE TRANSACTIONS ON INDUSTRY
APPLICATIONS.