You are on page 1of 6

Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

Dynamic Overcurrent Saturation of Distributed


Sources in a DC Microgrid System
Debasish Dhua, A. B. Shyam, Sandeep Anand and Soumya Ranjan Sahoo
Department of Electrical Engineering, Indian Institute of Technology Kanpur, India
debasish.dhua@gmail.com, shyam.ab33@gmail.com, me.sandeepanand@gmail.com, srsahoo@iitk.ac.in

Abstract—Due to the advancement of microgrids, the electrical


distribution network is transforming rapidly. The close proximity
of distributed generations, loads and widespread use of power
electronic converters have brought enormous flexibility and
controllability to the distribution system. The transient studies of
dc microgrid need more attention to ensure stable and reliable
operation of the converters. One of the important phenomena is
overloading of a source converter during a large load transient.
The source current may increase due to the large increase in load
connected to its proximity. Conventionally this issue is addressed
by using a static current limit, which protects the converters.
However, this leads to significant disturbances in the system,
especially when this converter recovers from saturation due to
other converters sharing the load. A suitable dynamic over-
current saturation technique, proposed in this paper is advanta- Fig. 1. A generic representation of dc microgrid.
geous over the existing constant current saturation method. This
paper shows a detailed procedure of determining the dynamic
current limit and its implementation in the controller in line with
the device manufacturers’ guidelines for the enhanced operating methods are recommended in [5] which mostly require under-
condition in dc microgrid. standing the response of the system and suitable tuning of the
Index Terms—anti-winding, boost converter, current limit, dc controller. Although the concept of saturating the uncontrolled
microgrid, dynamic saturation, overcurrent control system output is not new, the choice of saturation margin is
not distinctly elaborated in the existing literature.
I. I NTRODUCTION
To tackle the problem related to large load disturbances
With the growing interest in Distributed Generation (DG) across a converter, multiple control strategies have been in-
using of the microgrid, the power electronic converter topolo- troduced in [6]–[8] which include load current feed-forward
gies and their control techniques are rapidly evolving. Re- compensation and hysteric current mode control. The limita-
cently, the dc distribution system is gaining popularity because tion of these analog control methods mostly lies in large com-
of its simplicity, less number of conversion stages, the absence ponent size, parameter modification, and complex calculation
of reactive power and harmonics, enhanced use of modern procedure. In [9], [10] two separate digital control techniques
dc loads. [1]. A typical dc microgrid which is shown in Fig. have been discussed for steady-state and transient responses
1 consists of multiple DGs, energy storage units and loads which improve the dynamic response of the system at the
connected to the same dc bus. It possesses the flexibility to be cost of higher settling time. To protect the dc/dc converters
operated in grid-connected mode or islanded mode. In case of in transients under the complex dynamics of modern loads,
grid-connected mode, the grid interacts with the dc bus through nonlinear control [11], [12] and robust control [13]–[15]
an interfacing bidirectional converter. In dc microgrids, the techniques are discussed. However, to estimate the controller
DGs inject power into the distribution network through power operation most of the developed control techniques require
electronic converters having multiple control layers: primary, the knowledge of load which changes during operation and
secondary and tertiary. The control of dc/dc power converters that may enhance the non-linearities and number of system
under different operating conditions is receiving significant states. Therefore, a component independent and simple control
attention. [2]–[4]. algorithm is essential to address the transient issues related to
A common practice to deal with the controlled parameter large load disturbances across a converter. The desired control
when it exceeds a preset value is to use a constant saturation. technique should also possess the ability to limit the current
The immediate effect of introducing the saturation under independent of the load fluctuations.
large-signal disturbances is integrator windup which causes Under large load disturbances closer to one of the sources,
additional issues into the system like the sluggish response, its output current may exceed the preset current limit which
oscillations in the system. Some of the accepted anti-winding may damage the power electronic devices (PEDs). Therefore,
978-1-5386-6159-8/18/$31.00 © 2018 IEEE it is mandatory to specify the current limit considering the
Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

Fig. 2. Diagram of a boost converters

recommended safe operating region of the device. Hence,


dynamic current saturation is the most optimum solution to
support the system while protecting the PED of the individual
converters. Dynamic saturation methods have been incorpo-
rated in [16]–[18]; which have only considered the stability
aspect of the system, instead of directly considering the device
current carrying capabilities. The primary objective of this pa- Fig. 3. Normalized data and mean plot derived from the MOSFET datasheets
per is to introduce the dynamic saturation limit maintaining the
thermal capacity limit of the power semiconductor devices and
bringing up the coordination between the parallel connected and s
dc microgrid source converters. PD
ID = (3)
This paper is organized as follows: section II details with the RDS(on) (max)
thermal characteristics and Safe-Operating-Area (SOA) of the s
power semiconductor device. Section III depicts the proposed PDP
dynamic saturation scheme. Section IV elaborates the anti- IDP = (4)
RDS(on) (max)
winding method adopted in this paper. Section V includes
the simulation results and subsequent discussions. Section VI where, ID is the maximum dc current, IDP is the maximum
concludes the paper. pulsed currents, PD is the maximum continuous dissipated
power, PDP are the maximum pulsed dissipated power in
II. T HERMAL C HARACTERISTICS OF MOSFET AND SOA the MOSFET. Tch(max) is the maximum channel tempera-
S ELECTION ture, TC is the case temperature, Rth(ch−c) is the steady-
state thermal resistance, rth(ch−c) (t) is the transient thermal
In a dc/dc converter, for example, boost converter as shown
resistance which is a function of the pulse width, RDS(on) is
in Fig. 2, the output capacitor voltage drops abruptly when
the maximum drain-source on-state resistance at the maximum
there is a sudden increase in load demand. To maintain the
channel temperature. Values of all the above parameters are
capacitor voltage, the inductor current increases so as to
directly available in the data-sheet except rth(ch−c) (t) which is
draw more power from the input side. Therefore, the power
a function of time and is affected by the thermal capacitance of
electronic device is made to carry a larger current under
the device. Therefore, the calculation of the IDP vs tw which
transient high load demands.
can be implemented for the thermal protection of the device
The thermal characteristics of MOSFET under over-current in overcurrent condition involves the following steps,
condition can be correctly estimated from the manufacturers’
• Step I: Determine rth(ch−c) (t) from rth(ch−c) /Rth(ch−c)
datasheet [19]. Under ideal heat dissipation environment, the
current passing through the MOSFET can be broadly classified vs tw plot from datasheet using the available information
into two categories, continuous dc current which is considered of Rth(ch−c) . Since in this case, the MOSFET is consid-
for a duration of more than 10 seconds and pulsed current ered for extended exposures to overcurrent condition, tw
which is regarded for the device ON duration of 1 second or is taken for single pulse with pulse duration starting from
less. The maximum possible current values can be determined 10 µsec to 10 s.
• Step II: Calculate IDP vs tw using (2) and (4).
from the maximum dissipated power as given below so that
• Step III: Normalize the individual plots obtained from
the device channel temperature does not exceed its thermal
limit. different datasheets with corresponding MOSFET rated
currents.
Tch (max) − Tc After plotting the individual normalized data for IDP vs
PD = (1) tw obtained from different data-sheets [20]–[26] of different
R(ch−c)
manufacturers, the mean plot is derived by calculating the
Tch (max) − Tc mean of the data points corresponding to each specified pulse
PDP = (2) duration, as shown in Fig. 3. By using the basic curve fitting
r(ch−c) (t)
Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

Fig. 5. Closed-loop control scheme with dynamic saturation block shown in


red

Fig. 4. A typical MOSFET SOA plot with different limits indicated

technique for cubic polynomials, an equation for the mean plot


is generated which is given by,
It = 0.019x3 + 0.21x2 − 0.18x + 1.2 (5)
x = log10 (tw ) (6)
According to the initial part, for tw < 1 ms of the plot,
the calculated pulse current is significantly large but the pre-
scribed Safe Operating Area (SOA) of a MOSFET datasheet
introduces some restrictions for pulse currents of different
pulse widths. The restrictions generated by different limits are
indicated in Fig. 4. In this study, the maximum current Limit
information is considered from it. Fig. 6. The flowchart of dynamic current saturation algorithm
Therefore, the normalized current value is restricted to 4
times the maximum package rated current which is equal to
4Ir for the pulse duration up to 1 ms. For the pulse duration The dynamic control algorithm is developed based on the
of 1 ms to 1s, the current must follow (5)-(6) and gradually technique shown in the flowchart in Fig. 6, where ’Y’ and ’N’
reduces as the pulse duration increases so as to maintain the are marked to indicate the decision YES and NO, respectively
thermal limit. After 1s the MOSFET carrying current must during signal flow. The entire overcurrent saturation method
not exceed the maximum package limit rated current which is is classified into three categories based on the duration t. For
equal to Ir . the duration of 1 ms, it is kept with the limit of 4Ir . The
over-current duration 10 ms < t < 1 s is of prime interest,
III. P ROPOSED A LGORITHM FOR DYNAMIC S ATURATION during which the multiplication factor it is estimated based on
From the Section II, it is clear that the MOSFET must be (5) which reduces with time. After 1 s duration of continuous
undergone through a limited amount of current for a limited overcurrent, the saturation block strictly keeps it within the
duration and the MOSFET can withstand higher value of rated value. For every measurement of the saturation block
overcurrent for the shorter duration without hitting its thermal input current ii if the ii > ir comes as true, the overcurrent
limits. To implement this concept in a boost converter it is duration computation starts with ’count t’ whereas, if it comes
necessary to dynamically saturate the input inductor current as false the counter resets and the saturation block output
even if higher current is demanded by the output capacitor to current io becomes equal to ii . It is clear from the flowchart,
maintain its voltage at the preset reference value. The closed- once the current hits the saturation limit the converter operates
loop controller for the dc/dc converter is shown in Fig. 5, with in current control mode instead of taking care of the output
its dynamic saturation block for inductor current reference voltage.
shown in red. Input and output of the saturation block are
indicated as ii and io respectively. IV. S ELECTION OF A NTI - WINDING M ETHOD
The purpose of the dynamic saturation block is to determine Since the proposed method deals with the protection of
the output current io for the specific input current ii by mon- the MOSFET under over-current condition by saturating the
itoring the duration t for which the over-current is demanded current dynamically, this avoids saturation up to some level
and by finally comparing it with the device rated current ir . of load change. However, in case of a higher load change,
Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

Fig. 7. A typical dc microgrid system.

winding up effect may be visible due to the saturation limit.


The integral windup effect in PI controller is observed when
the control signal saturates the actuator. A further rise of the
control signal causes the integrator to accumulate error. Later,
when the control signal gets reduced, the error has to wind
down for a long time which leads to large overshoot and
larger settling time [27]. Primarily there are four different anti-
windup methods [28]–[30], i) conditional integrator, ii) limited
Integrator, iii) tracking anti-windup, iv) modified tracking
method. In this case, the conditional integrator method is
adopted as the anti-winding technique by limiting the integra-
tion procedure above a certain value which is slightly higher
than the desired dynamic saturation level.

V. S IMULATION R ESULTS AND D ISCUSSION


The simulated dc microgrid system to test the proposed
scheme is shown in Fig. 7. It consists of four boost converters
C1 , C2 , C3 and C4 each having a local load connected to it. Fig. 8. The simulation results with no saturation for inductor current
Power sharing among the converters is achieved by emulating reference. (a) converter output currents (b) converter output voltages, (c)
converter 4 saturation block input and output signals ii and io , (d) converter
droop resistance in each of it, d. Out of the four loads, L1 , L2 4 voltage control loop error signal ev with simulation time
and L3 are resistive loads whereas, L4 consists of resistive
load and controlled current sink in parallel. The current sink
adds stress to the converter C4 under overload condition. The objective of this simulation is to observe the coordi-
The data used in simulation for different converters and nation provided by the parallel connected source converters
distribution lines are listed in Table I and Table II. when one of them is overloaded due to the large change in
nearby load. Further, it would help to explore this phenomenon
TABLE I in presence of dynamic current saturation technique developed
DC MICROGRID SYSTEM PARAMETERS in previous sections.
During the simulation, initially, the total average load is
Parameter Value
Converter nominal voltage 400 V maintained to be 50% of rated load. At 0.1 second a large step
Converter power rating 32 kW change in load is introduced across C4 through the controlled
Cable resistance r1 , r2 , r3 0.25 Ω current sink which demands close to 1.5 times the rated
Cable inductance l1 , l2 , l3 0.27 mH
current. At 0.9 second the controlled current sink is removed
and the total load demand returns to the initial condition.
The advantage of the proposed dynamic current saturation
TABLE II
B OOST CONVERTER PARAMETERS
technique over the existing constant saturation is elaborated
through the following cases.
Parameter Value
Input voltage 200 V A. Case - I : Without Saturation
Rated output voltage 400 V
Rated power 32 kW In this case, there is no saturation applied to limit the
Switching frequency 10 kHz inductor current, the response of the 4-converters dc microgrid
Input inductance 2 mH system is shown in Fig. 8. In absence of the saturation limit,
Output capacitance 1000 µF
Current loop Kp , Ki 0.11, 565. 4 the inductor current of C4 reaches up to 120 A due to the
Voltage loop Kp , Ki 0.03, 48.5 large change in load at 0.1 second as depicted in Fig. 8(a). The
Droop gain 0.25 Ω output current of other converters are at their rated values or
Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

Fig. 9. The simulation results with constant saturation for inductor current Fig. 10. The simulation results with dynamic saturation for inductor current
reference. (a) converter output currents (b) converter output voltages, (c) reference. (a) converter output currents (b) converter output voltages, (c)
converter 4 saturation block input and output signals ii and io , (d) converter converter 4 saturation block input and output signals ii and io , (d) converter
4 voltage control loop error signal ev with simulation time 4 voltage control loop error signal ev with simulation time

in this case because the converter output current is forcefully


below since C4 supports the voltage completely by enhancing
restricted by introducing the constant saturation. It is seen from
its output current. The further the distance from the stressed
Fig. 9(b), C4 output voltage is at 355 V. The presence of
converter C4 , the lower the impact on the converter voltage
the constant saturation is clearly seen in Fig. 9(c). As there
and current waveforms as seen from Fig. 8(a) and Fig. 8(b).
is a constant difference maintained in between ii and io , a
C4 voltage is seen as 370 V whereas, the other converter
constant error value is noticed in Fig. 9(d) but is restricted by
voltages are close to the rated value with corresponding
the adopted anti-winding technique. In this case, the PED of
voltage drop due to individual droop resistance, higher the
each converter will be protected at the cost of poor voltage
output current higher is the voltage drop due to the droop
regulation by limiting the inductor current within a specified
resistance. The saturation block input and output currents ii
value.
and io overlap each other in this case as seen in Fig. 8(c) and
make error signal zero all throughout the simulation except C. Case - III : With Dynamic Saturation
at the switching transients which is shown in Fig. 8(d). It is In this case, a dynamic saturation is applied to the inductor
evident that the MOSFET in C4 will be severely stressed in current reference which reduces it, as the overcurrent duration
this case and the thermal blow out is evident well ahead of increases. From Fig. 10(a) it is clearly seen that C4 output
the time when the large load is removed. current gradually reduces from 120 A to 80 A, whereas, C3
output current increases to support C4 output voltage due to its
B. Case - II : With Constant Saturation proximity. From Fig. 10(b), the output voltages can be noticed,
In presence of a constant saturation, the inductor current where C4 voltage level is improved to 365 V than in Case II.
reference of each converter is limited to 150 A. Therefore, C4 The saturation block output current io is reducing with time
output current stays at the rated value of 80 A even when the when the input current ii is kept constant at slightly higher
large change in load is applied similar to the previous case, value by using the anti-winding technique as shown in Fig.
as shown in Fig. 9(a). The voltage regulation is even worse 10(c). Subsequently, the error in Fig. 10(d) also grows but stay
Proceedings of the National Power Systems Conference (NPSC) - 2018, December 14-16, NIT Tiruchirappalli, India

within an acceptable limit as the difference between ii and io [7] C. Song and J. L. Nilles, “Accuracy analysis of hysteretic current-
increases with time. In presence of the dynamic saturation, the mode voltage regulator,” in Applied Power Electronics Conference
and Exposition, 2005. APEC 2005. Twentieth Annual IEEE, IEEE,
device is protected from the thermal blow out in addition to vol. 1, 2005, pp. 276–280.
the improved transient voltage undershoot. [8] R. Miftakhutdinov, “Optimal design of interleaved synchronous buck
converter at high slew-rate load current transients,” in Power Elec-
VI. C ONCLUSION tronics Specialists Conference, 2001. PESC. 2001 IEEE 32nd Annual,
IEEE, vol. 3, 2001, pp. 1714–1718.
In this paper, the limitation of the existing constant current [9] M. Milanovic, M. Truntic, and P. Slibar, “Fpga implementation
saturation technique is introduced in boost converter in dc mi- of digital controller for dc-dc buck converter,” in System-on-Chip
crogrid system. The significance of the power semiconductor for Real-Time Applications, 2005. Proceedings. Fifth International
Workshop on, IEEE, 2005, pp. 439–443.
device manufacturers’ guidelines to tackle the thermal stress [10] A. Barrado, J. Quintero, A. Lazaro, C. Fernandez, P. Zumel, and
caused by the device overcurrent is emphasized. A detailed ap- E. Olias, “Linear-non-linear control applied in multiphase vrm,” in
proach has been developed here to include this thermal margin Power Electronics Specialists Conference, 2005. PESC’05. IEEE 36th,
IEEE, 2005, pp. 904–909.
into the converter control scheme followed by the algorithm [11] S.-K. Kim and K.-B. Lee, “Robust feedback-linearizing output voltage
implemented. It is evident that the MOSFET may get damaged regulator for dc/dc boost converter,” IEEE Transactions on Industrial
due to the flow of current more than its rated values for a Electronics, vol. 62, no. 11, pp. 7127–7135, 2015.
[12] G. C. Konstantopoulos, Q.-C. Zhong, B. Ren, and M. Krstic,
significant duration. To protect it from the thermal blow out, an “Bounded integral control of input-to-state practically stable nonlinear
algorithm is developed which enables it to carry high current systems to guarantee closed-loop stability,” IEEE Transactions on
up to short time duration (1 ms) and thereafter, it gradually Automatic Control, vol. 61, no. 12, pp. 4196–4202, 2016.
[13] H. Rodriguez, R. Ortega, G. Escobar, and N. Barabanov, “A robustly
reduces until the current reaches its rated output value within stable output feedback saturated controller for the boost dc-to-dc
the 1-second interval. Moreover, in a dc microgrid having converter,” Systems & control letters, vol. 40, no. 1, pp. 1–8, 2000.
multiple converters, the transient output voltage undershoot is [14] T. A. Theunisse, J. Chai, R. G. Sanfelice, and W. M. H. Heemels,
“Robust global stabilization of the dc-dc boost converter via hybrid
also improved due to the additional power delivering capacity control,” IEEE Transactions on Circuits and Systems I: Regular
in presence of dynamic saturation in comparison to that of Papers, vol. 62, no. 4, pp. 1052–1061, 2015.
constant overcurrent saturation technique. Hence, this paper [15] S. Buso, “Design of a robust voltage controller for a buck-boost
converter using/spl mu/-synthesis,” IEEE Transactions on Control
realizes the protection of the power semiconductor device, at Systems Technology, vol. 7, no. 2, pp. 222–229, 1999.
the same time utilizing it to the fullest, with a simple and [16] G. C. Konstantopoulos and Q.-C. Zhong, “Current-limiting dc/dc
effective control method. Although this technique is applied power converters,” IEEE Trans. on Control Systems Technology, 2018.
[17] Q.-C. Zhong and G. C. Konstantopoulos, “Current-limiting droop
to boost converter in this case, it can also be applied to other control of grid-connected inverters,” IEEE Transactions on Industrial
dc/dc converters. Electronics, vol. 64, no. 7, pp. 5963–5973, 2017.
[18] G. C. Konstantopoulos and Q.-C. Zhong, “Nonlinear control of dc/dc
ACKNOWLEDGMENT power converters with inherent current and power limitation,” in Con-
trol and Automation (MED), 2016 24th Mediterranean Conference on,
This work is supported by the project Indo-UK Center for IEEE, 2016, pp. 949–954.
Education and Research in Clean Energy by Department of [19] TOSHIBA, “Power mosfet maximum ratings toshiba;
toshiba.semicon-storage.com/info/docget.jsp?did=13414,” Visited
Science & Technology (DST), Govt. of India and U.S.-India Date: 02 May 2018,
collAborative for smart diStribution System wIth Storage (UI- [20] 30-v n-channel nexfet power mosfets, CSD17579Q5A, Initial release,
ASSIST) funded by Indo-US Science and Technology Forum Texas Instruments, Mar. 2015.
[21] 40-v n-channel nexfet power mosfet, CSD18502KCS, Revision B,
(IUSSTF). Texas Instruments, Aug. 2012.
[22] 100-v n-channel nexfet power mosfet, CSD19532Q5B, Revision B,
Texas Instruments, Dec. 2013.
R EFERENCES [23] Toshiba field effect transistor silicon n channel mos type, 2SK2611,
[1] L. Meng, Q. Shafiee, G. F. Trecate, H. Karimi, D. Fulwani, X. Lu, TOSHIBA, Jan. 2010.
and J. M. Guerrero, “Review on control of dc microgrids and multiple [24] Toshiba field effect transistor silicon n channel mos type, 2SK3878,
microgrid clusters,” IEEE Journal of Emerging and Selected Topics TOSHIBA, Nov. 2006.
in Power Electronics, vol. 5, no. 3, pp. 928–948, 2017. [25] Toshiba field effect transistor silicon n channel mos type, TK6A80E,
[2] E. Rodriguez-Diaz, J. C. Vasquez, and J. M. Guerrero, “Intelligent Rev 3.0, TOSHIBA, Mar. 2014.
dc homes in future sustainable energy systems: When efficiency and [26] “Infineon optimos power mosfet datasheet explanation,” Infinion Ap-
intelligence work together,” IEEE Consumer Electronics Magazine, plication note,
vol. 5, no. 1, pp. 74–80, 2016. [27] C. Bohn and D. Atherton, “An analysis package comparing pid anti-
[3] H. Kakigano, Y. Miura, and T. Ise, “Low-voltage bipolar-type dc windup strategies,” IEEE Control Systems, vol. 15, no. 2, pp. 34–40,
microgrid for super high quality distribution,” IEEE transactions on 1995.
power electronics, vol. 25, no. 12, pp. 3066–3075, 2010. [28] L. Rundqwist, “Anti-reset windup for pid controllers,” IFAC Proceed-
[4] P. Garcia, L. M. Fernandez, C. A. Garcia, and F. Jurado, “Energy man- ings Volumes, vol. 23, no. 8, pp. 453–458, 1990.
agement system of fuel-cell-battery hybrid tramway,” IEEE Transac- [29] N. J. KRIKELIS, “State feedback integral control with intelligent
tions on Industrial Electronics, vol. 57, no. 12, pp. 4013–4023, 2010. integrators,” International Journal of Control, vol. 32, no. 3, pp. 465–
[5] R. Ottersten and J. Svensson, “Vector current controlled voltage 473, 1980.
source converter-deadbeat control and saturation strategies,” IEEE [30] A. Glattfelder and W. Schaufelberger, “Start-up performance of
Transactions on Power Electronics, vol. 17, no. 2, pp. 279–285, 2002. different proportional-integral-anti-wind-up regulators,” International
[6] S. Kanemaru, T. Hamada, T. Nabeshima, T. Sato, and T. Nakano, Journal of Control, vol. 44, no. 2, pp. 493–505, 1986.
“Analysis and optimum design of a buck-type dc-to-dc converter
employing load current feedforward,” in Power Electronics Specialists
Conference, 1998. PESC 98 Record. 29th Annual IEEE, IEEE, vol. 1,
1998, pp. 309–314.

You might also like