Professional Documents
Culture Documents
Research Article
ISSN 1755-4535
Multi-objective control of multi-functional Received on 6th May 2015
Revised on 28th August 2015
grid-connected inverter for renewable energy Accepted on 2nd September 2015
doi: 10.1049/iet-pel.2015.0317
integration and power quality service www.ietdl.org
Abstract: Multi-functional grid-connected inverters (MFGCIs) not only interface renewable energy sources into the utility,
but also provide ancillary power quality enhancement service. Therefore, extra installment of power quality conditioners
can be partially avoided in a micro-grid including MFGCIs. Because the capacity of an MFGCI employed for power quality
compensation is limited, how to balance the multiple functions and optimally utilise the limited capacity becomes a
challenging for MFGCI application, and this is studied in details in this paper. First, to set up a benchmark for balancing
the multiple functions of the MFGCI, a comprehensive power quality evaluation (CPQE) index is presented based on
the catastrophe decision theory to quantify the power quality of a micro-grid. Then, for the strategic utilisation of the
limited capacity, a multi-objective optimal compensation model is proposed in which the objectives are to optimise the
CPQE index and minimise the occupied capacity of an MFGCI for power quality compensation. Finally, the solutions of
the model are derived on the basis of Pareto approach. As a result, the MFGCI can flexibly customise the power quality
of the micro-grid according to its available capacity margin and the users’ requirement. Finally, the experimental
results performed on a 10 kVA MFGCI prototype have confirmed the validity of the proposed model.
where uoabc and uabc stand for the output voltage of the MFGTI and
ya = −x1 /6, yb = 3 x2 /8 (8)
PCC, respectively; Rd represents the equivalent parasitic resistor of
filter inductor, insulated gate nipolar transistors (IGBTs), and lines.
According to (1), the transfer function from the grid-tied current to To guarantee x1, x2, and y all within [0, 1] for decision application, x1
the output voltage of MFGTI can be written as in ya multiplies –6, and x2 in yb are reduced 8 times, as indicated in
[33]. Hence the normalised bifurcation equations can be rewritten as
G1 = I(s)/Uo (s) = 1/(Ls + Rd ) (2)
√ √
y′a = x1 , y′b = 3 x2 (9)
Similarly, the transfer function from the grid-tied current to the
voltage of PCC can be expressed as Since the model in (6) is special and just two factors can be
contained, a further consideration is needed for other analytical
G2 = I(s)/U (s) = −1/(Ls + Rd ) (3) indexes includes more than two factors. According to the
catastrophe theory, when the factors of CPQE index to be
Taking any phase for instance, the block model of the MFGTI can be considered are more than two, some other models such as
derived, with the aid of (2) and (3), as indicated in Fig. 1b, where swallowtail catastrophe, butterfly catastrophe, etc. can be
Kpwm = Udc/2 is the equivalent gain of the MFGTI. employed, as indicated in [33].
To confirm the accuracy current-tracking for RESs integration and To eliminate the influence of variables with different units, the
power quality compensation, a multiple proportional-resonant (PR) factors x1 and x2 need to be normalised. It should be noted that,
is employed, which can be expressed as normalised factors have ‘smaller-and-better’ feature. The
normalised x1 and x2 can be written as
2Krh vch s
GPR (s) = Kp + (4)
s 2 + 2 v s + v2 x p1 = (x1 − x1min )/(x1max − x1min ) = y1 x1 + l1
ch h (10)
h=1,3,5,7
x p2 = (x2max − x2 )/(x2max − x2min ) = y2 x2 + l2
where ω1(ωc1) and ωh(ωch) are angular frequency (cut-off angular
frequency) of fundamental and hth harmonic components; Kp and where
Krh (h = 1, 3, 5, 7) are proportional and integral gains of the PR
controller, respectively. The open-loop and close-loop Bode y1 = 1/(x1max − x1min ) l1 = −x1min /(x1max − x1min )
, (11)
diagrams of the MFGCI with PR controller are given in Figs. 1c y2 = −1/(x2max − x2min ) l2 = x2max /(x2max − x2min )
and d, where ωc1 = ωch = 1 rad/s, Kp = 2.5 and Krh = 20. It can be
seen that appropriate zero-error current-tracking feature can be where x1max and x1min can be chosen as the THD of current through
achieved at the resonant angular frequencies, thanks to the high PCC igabc before the MFGCI compensating and zero, respectively; in
gain and zero phase-shift performance of the PR controller at such contrast, x2max and x2min can be chosen as unity and the TPF of igabc
frequencies, where the controlled plant model is G = GPRG1. before the MFGCI compensating, respectively. Because, the THD
and TPF of igabc are ideally controlled as zero and unity, after the
MFGCI carries out perfect power quality compensation. It should
3 Multi-objective optimal compensation of the be noted that the reference values for x1max and x2max are base
MFGCI values for the normalising process, and they also can be chosen as
other values. For instance, maximum THD x1max also can be 100%,
3.1 Model of the multi-objective optimal compensation and the maximum power quality PF x2max can be 1. In practice, all
the optimal compensation algorithm can be embedded in the
To quantify the cared power quality of the micro-grid, a CPQE micro-grid central controller (MGCC) as discussed in Section 6.
model is established based on the catastrophe decision theory as The multi-objective optimal compensation control strategy for
shown in Fig. 2 [33]. MFGCI application will be established step-by-step as follows.
It should be noted that the contribution of TPF to F1 is the cube root where Q0 and Q are the reactive power flow through PCC before and
of xp2 and it is much less important than factor xp1, since the after power quality compensation, and P is the active power through
contribution of TPF is less than the same THD in F1, as shown in PCC. j0 and j are TPF angles before and after reactive current
Fig. 2. Generally, the reactive results from the loads associated compensation of the MFGCI, while x20 and x2 are the
with electric or magnetic field, and all components of the power corresponding TPFs.
system such as generators, conductors, transformers, and According to the power analysis of reactive compensation shown
switchgear would be increased in size (and cost) to carry the extra in Fig. 3, the capacity of the MFGCI used for reactive compensation
reactive current [34]. On the other hand, the harmonic results from can be expressed as
the non-linear factors of machine, transformer, electronic elements,
etc., and harmonics have a number of undesirable effects on power
DQ = Q0 − Q = P( tan w0 − tan w) (19)
system components and loads. It may cause nuisance tripping of
sensitive electronic loads and overheating of transformers, and
even can degrade meter accuracy, etc. [35]. As a result, a series of The TPF at the PCC before and after compensation satisfies as
standard requirements on harmonic and reactive has been
promulgated to confirm the safety and economic operation of the tan w = tan(cos−1 x2 ) = Q/P = (1 − a2 )Q0 /P (20)
utility. Usually, it is typically required that the TPF is more than
0.9 to 0.95 and the THD is less than 5% [36]. For example, if the where α2 = (Q0–Q)/Q0 is the compensated percentage of the reactive
TPF of a power system is 0.9, the acceptable reactive current current that should be optimally designed. Similar to the results of
should be Iq = I1sin(cos–10.9) = 0.45I1. On the contrary, for a same coefficient α1 mentioned before, the coefficient α2 is also limited
power system, if the THD is 5%, the allowable harmonic current between 0 and 1. There is α2 = 0, if there is no reactive
is Ih = 0.05I1. That is to say, a power system can accept much compensation of the MFGCI. If the MFGCI carries out full
more reactive current compared with harmonic current, hence the reactive compensation, it is α2 = 1. Thus, apparent capacity of the
THD is paid much more attention to quantify the CPQE index. MFGCI for reactive compensation can be written as
The second objective for the MFGCI is the apparent capacity
utilised for power quality enhancement, which can be written as S2 = DQ = P( tan w0 − tan w) = a2 Q0 (21)
F2 = S12 + S22 (13)
3.2 Reference current generation algorithm where ipdq and iqdq are the active and reactive parts of the
fundamental component of equivalent load current iLabc = iabc +
Aforementioned CPQE-based multi-objective optimal compensation igabc in rotating dq frame [26]. udq and idq , corresponding to
model considers the harmonic and reactive components separately, fundamental components of voltage and current in natural abc
and it requires compensating these components distinguishably. frame, are constants in rotating dq frame and can be filtered out by
Therefore, the algorithm of reference current generation should LPFs.
independently detect harmonic and reactive components of the Therefore, the harmonic components of the equivalent load
equivalent load current at PCC. In such condition, an algorithm current in synchronous rotating dq frame ihdq can be derived as
can meet the requirements of the optimal model is presented as
shown in Fig. 4, where Tabc/dq is the Park transformation and Tdq/ ihdq = idq − i pdq − iqdq (29)
abc is its inverse transformation. Besides, an extra block is
employed to detect the zero-sequence current i0 and eliminate its To calculate the compensation part of the reference current icdq of the
influence. To obtain the fundamental positive-sequence MFGCI, the detected harmonic and reactive components ihdq and iqdq
components of voltage and current, low pass filters (LPFs) are also should multiply the compensation coefficients α1 and α2 according
employed as shown in Fig. 4. It should be noted that, thanks to to the multi-objective optimal model mentioned before, which can
the ‘T’-type structure of MFGCI in Figs. 1a and 4, the reference be expressed as
current generation algorithm is independent of the active and
passive elements of local loads, because the total load current can
be obtained by the summation of utility current igabc and MFGCI icdq = a1 ihdq + a2 iqdq (30)
Fig. 7 Pareto solutions of the multi-objective optimal compensation model of the MFGCI
a Pareto fronts
b Optimal compensation coefficients
coefficients for harmonic compensation are 0.6359 and 0.5, the multiple MFGCIs in the micro-grid. With the guide of MGCC,
respectively. Hence employed harmonic compensation in optimal each MFGCI just partly compensates the power quality issues in
compensation II has small difference from the one in half the micro-grid according to the multi-objective optimal model in
compensation. Furthermore, the harmonic distribution is shown in the tertiary control. Thus, each MFGCI can autonomously work as
Fig. 8d. The TPF of the PCC current is 0.9953 and 0.9841, as an individual module without needing the information of other
well as the THD is 8.632% and 6.729% in half compensation and MFGCIs. In the secondary control, the power generation of
optimal compensation II, respectively. In addition, for half MFGCIs is exchanged and modified by the MGCC to ensure the
compensation and optimal compensation II, actual CPQE indexes safe and economic operation of the micro-grid as suggested in
are 0.3582 and 0.3456, as well as the utilised capacity for power [37, 38]. In the local primary controller of the MFGCI, the output
quality compensation is 0.8921 kVA and 0.6459 kVA, current tracking controller is embedded, and the MFGCI can
respectively. Compared with the optimal compensation strategy I, generate the desired active, reactive, and harmonic current to the
strategy II pays much attention to the harmonic and reactive micro-grid to fulfill the functions to interface DERs into utility and
current compensation. As a result, there is more capacity is utilised enhance the power quality of the micro-grid as expected as
and the CPQE index is much smaller than the one of strategy I. In demonstrated in [37, 38].
summary, more capacity usage of the MFGCI, better power
quality at PCC can be guaranteed. If the available capacity for
power quality enhancement is limited, the MFGCI can flexibly and 7 Conclusions
optimally work based on the multi-objective compensation in
Fig. 7 to customise the power quality to a certain level. In micro-grids, provision of power quality enhancement service by
MFGCIs is important. As the apparent power capacity of an
MFGCI is limited, it is necessary to utilise the limited capacity
6 Discussions effectively and enhance the power quality of the micro-grid as
much as possible. A multi-objective optimal compensation
To integrate the multiple MFGCIs in a micro-grid, a configuration is strategy for MFGCI application is proposed in this paper.
displayed in Fig. 9 to show how the proposed method can Besides, a CPQE approach based on catastrophe decision theory
cooperatively handle the power quality issues in a feeder with is applied to evaluate the micro-grid power quality in a
communication. It can be seen from Fig. 9 that the proposed comprehensive and quantitative way. In such a way, an MFGCI
multi-objective optimal compensation model can be solved by the can flexibly and optimally enhance the power quality of the
CPQE-based multi-objective optimal model in the MGCC, and the micro-grid without extra power quality conditioners to a certain
MGCC also download the optimal compensation coefficients α1 extent. Experimental results from a feeder of a micro-grid have
and α2 to MFGCIs’ local controllers. In general, the MGCC can confirmed the validity and feasibility of the proposed optimal
be a hierarchical controller as indicated in [37, 38], and it also compensation strategy.
should be noted that the optimal compensation of MFGCIs may be
embedded in the tertiary control to dispatch the MFGCIs to
provide optimal power quality services. 8 Acknowledgments
In the micro-grid, the MGCC detects the utility voltage uPCC and
current iPCC at PCC, and then the THD and TPF of the current can be Thanks for the financial supports from National Key Basic Research
computed. In addition, the initial information of the optimal model Program of China (‘973 Program’) (under grant no.
can be given. Therefore, the aforementioned multi-objective 2012CB215200), National High Technology Research and
optimal model in Section 3 can be solved in the tertiary control, Development Program (‘863 Program’) (under grant no.
and the obtained Pareto fronts solution is numerically saved in the 2011AA050204), National Natural Science Foundation of China
MGCC. To meet the requirement of the system operator or the (under grant no. 51377184), and Fundamental Research Funds for
owner of the micro-grid on the CPQE index F1, the desired the Central Universities (under grant nos.
compensation coefficients, α1 and α2, can be obtained and sent to 106112015CDJXY150005, CDJZR12150074).