Professional Documents
Culture Documents
A Fully Distributed Hierarchical Control Framework For Coordinated Operation of DERs in Active Distribution Power Networks
A Fully Distributed Hierarchical Control Framework For Coordinated Operation of DERs in Active Distribution Power Networks
6, NOVEMBER 2019
0885-8950 © 2018 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.
XIA et al.: FULLY DISTRIBUTED HIERARCHICAL CONTROL FRAMEWORK FOR COORDINATED OPERATION OF DERS IN ADPN 5185
using an additional innovation term to ensure that the power of ADPN. In this way, all the three layers are designed to
mismatch gradually approaches to zero. In [11], the consensus operate without a central coordinator and therefore a fully dis-
algorithm was hybrid with the alternating direction method of tributed hierarchical control framework is well established.
multipliers (ADMM) for solving the optimal power flow prob- The main contributions of this paper are as follows.
lem with demand response in a distributed manner. The above 1) To accommodate the feature of highly penetrating and
literatures mainly focused on applying distributed algorithms widely scattering DERs, a fully distributed hierarchical
for off-line optimal operation of DERs in tertiary generation framework without requiring a centralized coordinator, as
regulation layer, while the off-line solutions are not suitable comparisons to the centralized hierarchical framework in
to be implemented for the real-time coordination of DERs in [12], [13] or decentralized one in [14], [15], is innova-
distribution networks. tively designed for ADPN economical operation by using
A complement to the off-line optimization in a single layer, ADMM and DFCA for the tertiary and secondary layer
some scholars have put forward the hierarchical control structure as well as the local droop controller for the primary layer,
to address different requirements with different control hierar- respectively.
chy. For example, authors of [12] proposed a hierarchical control Compared with ordinary consensus based ADMM for
scheme to improve the system overall efficiency while consid- DC-OPF problem [11], here the DFCA is incorporated
ering the State-of-Charge balance at the same time, and the sec- into ADMM for formulating an efficient distributed op-
ondary and tertiary control were implemented in a micro-grid timization algorithm with less complexity and good data
central controller. In [13], a potential-function was used for the privacy, which only needs very few iterations to calculate
secondary and tertiary control of a micro-grid, and an all-round the power imbalances in tertiary layer compared with the
central controller was needed to communicate with all DERs. ordinary consensus algorithm as demonstrated by Fig. 5.
Besides of these centralized hierarchical control schemes for Different from the conventional centralized generation
power system operation, some center-free hierarchical control correction approach implemented with a control center
strategies were also put forward. For instance, a decentralized to collect fuel cost coefficients of all DERs for calcu-
hierarchical framework was proposed in [14] and [15] to opti- lating power participation factors [20], the DFCA is uti-
mize the charging process of PEVs for frequency regulation in lized in this paper to form a distributed generation cor-
a competitive electricity market. In [16], [17], new distributed rection strategy and conduct the economical generation
control strategies were established for both the frequency and correction without any central controller in the secondary
voltage regulation of micro grids, in which only the localized layer.
information and nearest-neighbor communication were needed. 2) The proposed fully distributed framework could optimally
Dr. M. A. Mahmud of [18] designed a nonlinear distributed con- control the active power output of multiple DERs with the
troller capable of maintaining both active and reactive power minimal cost under the fluctuating load condition, while
balance to prevent overloading. Further, the work in [19] is not ensuring satisfactory system dynamic frequency perfor-
only able to restore the system frequency and network voltage in mance. Compared with the distributed control scheme in
a distributed manner but also ensure its reactive power sharing. [4], [16]–[19] to conduct the generation corrections ac-
However, the above works mainly perform the power quality cording to its rated power, which neglected the economy
regulation and solve the power sharing problem proportionally of the generation corrections in the secondary layer, the
according to DERs’ power ratings, while the economical op- proposed distributed approach utilizes the DFCA to real-
eration of the whole system is not taken into account and thus ize the economical generation corrections in the secondary
optimal power sharing cannot be achieved. layer and thereby the corporation of tertiary and secondary
Motivated by the above works, this paper proposed a fully layer can obtain the real time optimal solution of DERs
distributed hierarchical control framework for optimally coor- outputs, which has been validated by the benchmark re-
dinating DERs operation in ADPN by three layers. In the tertiary sults of traditional centralized methods. (please refer to
generation regulation layer, a fully distributed dispatch strategy the Table I in Case Study).
incorporating the ADMM into discrete finite-time consensus 3) The proposed hierarchical framework is robust to accom-
algorithm (DFCA) was properly designed to optimize the ac- modate the complexities and uncertainties of DERs, such
tive power base point of DERs in a center-free manner. In the as the random plug-in/out behaviors of DERs.
secondary generation regulation layer, a distributed fair gener- 4) Comprehensive simulations on a distribution system with
ation correction strategy was proposed by using the DFCA to multiple DERs have demonstrated that the proposed fully
counterbalance the load deviations in a shorter timescale on top distributed hierarchical framework is effective to opti-
of the tertiary generation regulation, which was also without mally coordinate the power output of DERs and can well
relying on a central coordinator. Collectively, the tertiary and regulate the system frequency.
secondary layers provided the primary control layer with ac- The rest of this paper is organized as follows. The preliminary
tive power tracking reference points. In the primary layer, by of DERs’ economically coordinated operation and the discrete
using locally measured frequency deviations, the power droop finite-time algorithm are introduced in Section II, and then a
controller was designed for the fast acting devices including fully distributed three-layer hierarchical control framework is
the DERs, battery energy storages (ESs) and the controllable designed in Section III. Afterwards three case studies are con-
load to maintain satisfactory frequency dynamic performance ducted in Section IV and V to validate the effectiveness of the
5186 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 6, NOVEMBER 2019
proposed hierarchical scheme, and the conclusion is drawn in time actual load demand would be denoted by the off-line load
the last section. prediction Dpr ej plus the on-line load deviation Ddev j as (4).
the DER distributed off-line economic dispatch for predicted + λk Ptj − Pti + Pti − Dpr ej
load demands. The secondary layer is designed to have faster j =1 j
dynamics response than that of the tertiary layer and is used to
ρ
n
(k ) (k )
m
conduct the generation corrections for load deviations in dis- + Ptj − Pti + Pti − Dpr ej (8)
tributed manner. The bottom layer is the primary control which 2 j =1 j
locally provides the frequency regulation and ensures reference n (k ) m
tracking with fast responses. Denote j =1 Ptj − j Dpr ej = e(k ) , then (8) can be sim-
In the tertiary layer, a fully distributed dispatch strategy based plified as
on DFCA is properly designed to optimally allocate the active (k )
(k +1) ρ[Pti − e(k ) ] − λ(k ) − bi
power base point of DERs in a center-free manner. For the sec- Pti = (i = 1, . . . , n) (9)
ondary generation regulation, a distributed fair generation cor- ρ + 2ai
rection strategy is proposed by using DFCA based on the power (k +1)
with constraint Pim in < Pti < Pim ax for ith DER output
sharing ratio to timely counterbalance the load deviations in a Pti . The dual variable is updated by
shorter timescale on top of the tertiary generation regulation. ⎛ ⎞
Collectively, the tertiary and secondary layers provide the pri- n
(k +1)
m
mary control layer with active power tracking reference points. λ(k +1) = λ(k ) +ρ ⎝ Pti − Dpr ej ⎠ = λ(k ) + ρe(k +1)
In the primary layer, a local frequency droop controller is pre- i=1 j
sented for DERs, battery energy storage and controllable load (10)
to maintain satisfactory system frequency dynamics. In the fol-
Equation (9) and (10) are updated iteratively for the optimal
lowing, we will describe the three-layer generation regulation
solution until the following criterions are both satisfied [28].
scheme in detail. ⎛ ⎞
n
(k +1)
m
||r(k +1) ||2 = ρ ⎝ Pti − Dpr ej ⎠ < εpr i
i=1 j
A. Distributed Tertiary Generation Regulation
n (k +1) (k ) 2
In the tertiary generation regulation layer, we designed a fully ||s(k +1) ||2 = (Pti − Pti ) < εdu al (11)
i=1
distributed optimization approach based on ADMM and DFCA
to conduct the off-line generation dispatch for the predicted load where εpr i and εdu al are tolerance parameters pre-specified in
Dpr ej , and this optimal generation of DER in tertiary layer is terms of expected accuracy.
(k +1)
indicated as Pti in Fig. 1. If Gi and Dj are substituted by Pti From formula (9), the updating of Pti relies on ith DER’s
and Dpr ej in (1), (2) for merely considering load predictions, local parameters ai , bi , and the system total power imbalance
DERs off-line generation dispatch problem could be converted e(k ) . As e(k ) is the global information of the whole system and
5188 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 6, NOVEMBER 2019
Assume both the first and second order derivatives of fuel cost
Algorithm 1: DFCA for locally estimating e(k ) . function with respect to DER output are available (i.e., both Fi
1) Suppose there are totally n buses connected with loads and Fi exist for fuel cost function Fi in (1)). It is well-known
and DERs, and the power imbalance for node i is initialized that the incremental costs or the first order derivatives Fi (i =
as yi (0) = Pti (0) − Dpr ei , where Pti (0) and Dpr ei stand for 1, . . . n) of all DERs are equal to the Lagrange multiplier λ
the DER initial output and load prediction at bus i. (Note: when DERs reach the optimal solution. Denote the incremental
for a bus only with load then Pti (0) = 0, and similarly cost of DER i for the base point Pti in tertiary layer as λ0 =
Dpr ei = 0 for a node only with DER). Initialize the Fi (P
weighting factors ωij by (5) based on the pre-configured ti ). Since the system total load is changed by a total amount
of Ddev j , the marginal cost and optimal output of DER i
communication network; should move to λ0 + Δλ and Pti + Psi accordingly with the
2) Repeat to update yi (k ) by DFCA using (5) in K steps, relationship λ0 + Δλ = Fi (Pti + Psi ), then we have
where yi (k ) stands for the average power imbalance
estimated at bus i; Δλ = Fi (Pti + Psi ) − Fi (Pti ) = Fi (Pti )
3) The total unbalanced power e(k ) could be calculated × Psi for i = 1, . . . n (12)
as n × yi (k ) for bus i, where n is the total number of buses.
To satisfy the power generation and load balance, we have
needs to be estimated by collecting all DERs’ outputs and all
n
m
Psi = Ddev j (13)
loads in distribution networks with an information hub, it is just
i=1 j =1
a decentralized approach instead of a fully distributed one when
directly using ADMM to solve the problem. A fully distributed The solution of (12) and (13) is
approach should be capable of calculating the global imbalanced ⎛ ⎞
m
power e(k ) only based on the local information without any cen- Psi = pfi × ⎝ Ddev j ⎠ (14)
tral coordinator. For this purpose, the previous introduced DFCA j =1
in Section II.B could be introduced to estimate e(k ) by sharing
limited information with neighbor agents as Algorithm 1. where pfi = F1 / nj=1 F1 is the power participation factor for
i j
The proposed distributed dispatch approach based on ADMM DER i. Formula (14) indicates that the economical generation
and DFCA mainly includes the following steps for the tertiary correction Psi of DER i for the total load deviation Ddev j is
generation regulation. inversely proportional to the second derivative Fi of fuel cost.
(0)
1) Initialize the output of DER i as Pti and the dual variable For conventional generation correction, a central controller is
(0)
λ , prepare the weighting factors ωii and ωij by (5) needed to collect fuel cost coefficients of all DERs for calculat-
according to the communication network; ing the summation of Fj (j = 1, . . . n) and the power participa-
2) Estimate the total power imbalance e(k ) of distributed tion factors pfi [20]. In order to design a distributed generation,
system networks by Algorithm 1; formula (14) is rewritten in the form of (15) for DER i and j.
(k +1)
3) Update ith DER output Pti by (9) and the dual variable
m
λ (k +1)
by (10); Ddev j = Psi /pfi = Psj /pfj or
(k +1) j =1
4) If the stop condition (11) is satisfied, output Pti as the
final optimal solution; otherwise go to step 2). 1 1
yi = Psi = Psj = yj (15)
Fi Fj
B. Distributed Secondary Generation Regulation where yi and yj are the defined power sharing ratio in terms
In the tertiary generation regulation, we only find the DER of DER generation correction to DER power participation fac-
optimal output (called base point hereafter) for the predicted tor. These power sharing ratios yi and yj are also the tailored
load demand Dpr ej . However, the actual load demand Dj in real consensus variables for DER i and j which could be used in the
time is usually different from the predicted load as expressed discrete finite-time consensus algorithm as shown in Fig. 1.
in (4), and our ultimate goal is to economically coordinate the With (15), the secondary regulator compares the local power
DERs outputs for the real-time actual load demand. sharing ratio yi of DER i with its neighbors’ power sharing ra-
As the tertiary layer has already optimized DER operation tio yj on a communication graph and accordingly ensures all
base point Pti for the predicted load Dpr ej , if we could further DERs’ power sharing ratio converging to the consensus value
conduct the economical generation correction (indicated as Psi in finite-time, which could be guaranteed by Lemma of DFCA
here) for the real time load deviation Ddev j on top of the base introduced in Section II.B. Thereby the equality (15) is satisfied
point, then the integrated generation in tertiary and secondary in finite time, and an economical generation correction strat-
layer as Pr ef i = Pti + Psi would be the DER optimal solution egy is established based on the power sharing ratio to timely
for the real-time actual load demands. In other words, as the counterbalance load deviations Ddev j on top of the tertiary
actual load deviates from the load prediction, we need to inves- generation regulation layer.
tigate how much each DER should be adjusted (i.e., ‘participate’ To provide the input signal for secondary generation regu-
in the load change) from the base point in order that the actual lation in Fig. 1, the frequency deviation Δf, calculated as the
load can be served by DERs in the most economical manner. frequency reference f ∗ minus the distribution network actual
XIA et al.: FULLY DISTRIBUTED HIERARCHICAL CONTROL FRAMEWORK FOR COORDINATED OPERATION OF DERS IN ADPN 5189
⎧
⎪
⎪ −PE S i m ax (KE S i Δf < −PE S i m ax ) & (SOCt < SOCm ax )
⎪
⎪
⎪
⎪
⎪ KE S i (f ∗ −f ) = KE S i Δf
⎪ (−PE S i m ax < KE S i Δf < PE S i m ax ) & (SOCm in < SOCt < SOCm ax )
⎨
Pr eq ,i,t = PE S i m ax (KE S i Δf > PE S i m ax ) & (SOCt > SOCm in ) (17)
⎪
⎪
⎪
⎪
⎪
⎪ 0 (KE S i Δf < 0) & (SOCt > SOCm ax )
⎪
⎪
⎩
0 (KE S i Δf > 0)], & (SOCt < SOCm in )
V oc = V nom + α(RT /F ) ln(SOC/(Cnom − SOC)) (18)
5190 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 6, NOVEMBER 2019
Voc to SOC; R, F and T are the gas constant, the Faraday constant
and battery temperature, respectively. When SOC is kept within
a range of 10%−95% to preserve battery life, Rseries , Rtrans and
Ctrans can be approximated as constants with typical values ob-
tained from [31]. The dynamics of battery charging could be
described by the block diagram in Fig. 3 [25],where Pr eq ,i,t
is the reference point of ES charging/discharging power de-
termined by the primary frequency regulation controller using
(17).
For the distribution network dynamic, an inertia block dia-
gram 1/(Ms + D) is used to model the effect of the net load
change on frequency deviation by parameter D and the inertia
M of distribution network as shown in Fig. 1.
Remarks:
1) Based on the proposed hierarchical framework, DER gen- Fig. 4. Modified IEEE 33-bus ADPN with 3 WFs and 15 DERs.
erations could be effectively coordinated based on the
DFCA (in the tertiary and secondary generation regula-
tion layer) and the local frequency droop control (in the 15, 17, 19, 21, 23, 25, 27, 29} while the fuel cost coefficients
primary frequency regulation layer), which only needs are cited from Table II of [33] and the upper limits of power
the local information and parameters without any cen- outputs are enlarged 2.5 times to match the system total load
tral information hub, and hence the proposed hierarchical level. As illustrated in Fig. 4, the communication links (indi-
framework is a fully distributed control approach; cated as blue dash line) are mainly configured based on the
2) The proposed fully distributed hierarchical framework can physical network, which can conveniently utilize the power line
optimally coordinate the active power output of multiple communication techniques, plus extra links between bus 8–22,
DERs with the minimal cost by optimizing the base point 18–33, 25–29 to satisfy the n-1 communication reliability. Each
in the tertiary layer plus economically conducting gen- bus in Fig. 4 is deployed with a smart agent, which is capa-
eration correction in the secondary layer, while ensuring ble of communicating with its neighbor agents for sharing the
satisfactory system frequency with fast response in the load and generation information. The tertiary generation reg-
primary layer. ulation layer is updated every 300 s based on the short-term
3) The proposed distributed framework can be fitted in a two- load and wind power prediction, while the secondary genera-
segment power market composed of intra-day market and tion regulation and the first layer control are updated in real time.
real-time spot market. The power generation scheduling of The parameters of ADPN inertia block are D = 1.25 MW/Hz,
tertiary layer could be performed in the intra-day market M = 50 s. The droop coefficient of DERs is mi = 0.1 MW/Hz.
based on the price signals such as the fuel costs of local And the parameters of controllable load are KC L = 1.5 MW/Hz
generators, and the combination of ADMM and DFCA and TC L = 5 s MW/Hz. The droop coefficient of two ESs are
could be used to address the optimal solutions in a center- KE S 1 = 2 MW/Hz for KE S 2 = 1 MW/Hz.
free intra-day market, in which the market participants For the parameter ρ of ADMM, as there are no general rules of
will coordinate to make bids periodically for each timeslot how to select the optimal penalty parameter ρ, in the following
(such as 5–15 minutes) without market decision center and case studies we choose ρ by trial and errors and fix it as 10−5 to
update their generators output accordingly. The secondary guarantee the good convergence and yield the optimal solutions.
and primary frequency regulation layers can be used to
provide balancing services in the real-time spot market,
and the generation correction and droop control strategies A. Case 1: Investigate the Performance of Proposed Scheme
will be used to reallocate the supply-demand unbalance by 33-Bus ADPN With Simplified Load Condition
among generators and ES. While the technical content For zooming in the performance of proposed scheme,
of the proposed distributed scheme is the focus here, the we firstly conducted simulations for the following simplified
application of proposed scheme in a centre-free market scenarios: 1) To check the effectiveness of the tertiary genera-
environment will be thoroughly investigated in the future. tion regulation layer, the predicted total load level of the 33-bus
ADPN is assumed as 3715 kW during 0–300 s, while at 300 s
bus1 to bus10 each increases 20 kW and therefore the predicted
IV. CASE STUDY FOR 33-BUS ADPN WITH WIND FARMS load demand changes to 3715 kW + 200 kW = 3915 kW during
The performance of proposed hierarchical framework for 300 s–2400 s; 2) To validate the performance of the secondary
DERs coordinated operation was firstly tested on a modified generation correction layer, the total load deviation of this
33-bus ADPN [32] with a single Point of Common Coupling 33-bus ADPN is assumed zero during 0 s–700 s and 450 kW
(PCC) as shown in Fig. 4, which consists of 15 DERs, 32 loads, during 700 s–2400 s, and the step moment is indicated by the
2 energy storages, 1 controllable load and 3 wind farms. The blue dot-line B in Fig. 5; 3) DER11 is disconnected at 1300 s in-
15 DERs are deployed in the bus set {1, 3, 5, 7, 9, 11, 13, dicated by the blue dot-line C and reconnected at 1900 s marked
XIA et al.: FULLY DISTRIBUTED HIERARCHICAL CONTROL FRAMEWORK FOR COORDINATED OPERATION OF DERS IN ADPN 5191
TABLE I
SOLUTIONS OF PROPOSED SCHEME AND CENTRALIZED ED AT 900 S
VI. CONCLUSION
In this paper, a fully distributed three-layer control framework
is proposed for the coordinated operation of multiple DERs
in ADPN. Based on the discrete finite-time consensus algo-
rithm, a fully distributed dispatch scheme and a fair generation
correction strategy are designed in the tertiary and secondary
generation regulation layer respectively. While in the primary
Fig. 13. Frequency regulation performance for 132-bus ADPN.
generation regulation layer, a frequency droop controller is pre-
sented to ensure DERs outputs perfectly tracking the power
and 1.5ai for DERs in area D. The predicted load and wind reference of the tertiary and secondary layers, and DERs with
power generation are similarly generated as that for 33-bus energy storage and controllable load could well maintain satis-
ADPN with proportional modifications, which is indicated by factory system dynamic frequency in real time. Tests on the
the black line in Fig. 12(a). The wind power deviations and the modified IEEE 33-bus and 132-bus ADPN with large-scale
ultimately resultant load demand are provided as the pink and DERs have demonstrated the effectiveness of the proposed con-
green line in Fig. 12(a). To observe the plug-in/out behaviors of trol scheme in optimally coordinating DERs outputs, ensuring
DERs, DER1 in subsystem A is assumed to be plugged out at satisfactory system frequency and robustly accommodating un-
9000 s and plugged in at 15000 s. certainties of DERs. Since the proposed fully distributed hier-
Fig. 12(b) to Fig. 12(d) demonstrated the optimized base oper- archical scheme to exclusively regulate system frequency may
ation point in tertiary regulation layer, the generation corrections result in operation constraint violations, such as violating the
in secondary regulation layer and the actual generation outputs voltage or line capacity limits, further ongoing research would
of 60 DERs solved by the proposed hierarchical scheme. As elaborately design an all-round distributed hierarchical scheme
illustrated by these figures, the tertiary layer could effectively (e.g., based on fully distributed optimal power flow model) to
address the optimal base point according to the predicted load simultaneously regulate system frequency and ensure static se-
demand and wind power generation, while the secondary gener- curity of distribution networks.
ation regulation layer could further economically conduct gen-
eration corrections on top of tertiary generation regulation, and REFERENCE
finally the outputs of 60 DERs are coordinated to match the real [1] X. Shen, M. Shahidehpour, S. Zhu, Y. Han, and J. Zheng, “Multi-stage
time load demands of 132-bus ADPN as shown in Fig. 12(d). planning of active distribution networks considering the co-optimization
of operation strategies,” IEEE Trans. Smart Grid, vol. 9, no. 2, pp. 1425–
It is also noted that when DER1 is disconnected during 1433, Mar. 2018.
9000 s–15000 s, the optimal generation base point in tertiary [2] A. Madureira, C. Gouveia, C. Moreira, L. Seca, and J. P. Lopes, “Coordi-
layer, the generation corrections of secondary layer and the ac- nated management of distributed energy resources in electrical distribution
systems,” in Proc. IEEE PES Conf. Innovative Smart Grid Technol., Apr.
tual generation outputs for DER1 are all decreased to zero. And 15–17, 2013, pp. 1–8.
the proposed scheme is still effective to guide other 59 DERs [3] W. Zheng, W. Wu, B. Zhang, H. Sun, and Y. Liu, “A fully distributed re-
to seek the optimal outputs for minimizing the fuel cost and active power optimization and control method for active distribution net-
works,” IEEE Trans. Smart Grid, vol. 7, no. 2, pp. 1021–1033, Mar. 2016.
regulating system frequency well. As DER41 is much cheaper [4] Q. Shafiee, V. Nasirian, J. C. Vasquez, J. M. Guerrero, and A. Davoudi, “A
than other DERs, its generation output is very large and fre- multi-functional fully distributed control framework for AC microgrids,”
quently restricted by the upper generation limit, while DER7, IEEE Trans. Smart Grid, vol. 9, no. 4, pp. 3247–3258, Jul. 2018.
[5] W. T. Elsayed and E. F. El-Saadany, “A fully decentralized approach for
8, 9, 10, 34, 35, 37 and 39 etc. occasionally reach their lower solving the economic dispatch problem,” IEEE Trans. Power Systems,
generation limits during 7000 s–14000 s as shown in Fig. 12(d) vol. 30, no. 4, pp. 2179–2189, Jul. 2015.
because of their expensive marginal fuel costs. Fig. 12(e) also [6] S. Yang, S. Tan, and J. Xu, “Consensus based approach for economic
dispatch problem in a smart grid,” IEEE Trans. Power Syst., vol. 28, no. 4,
demonstrates the ESs power outputs and the controllable load pp. 4416–4426, Nov. 2013.
level accordingly during the procedure of counterbalancing the [7] Z. Zhang and M. Chow, “Convergence analysis of the incremental cost
frequency deviations. consensus algorithm under different communication network topologies
in a smart grid,” IEEE Trans. Power Syst., vol. 27, no. 4, pp. 1761–1768,
Fig. 13 plotted the frequency deviations of 132-bus ADPN for Nov. 2012.
three different schemes: 1) the ED algorithm with DER primary [8] H. Pourbabak, J. Luo, T. Chen, and W. Su, “A novel consensus-based
frequency droop control; 2) the proposed three-layer hierarchi- distributed algorithm for economic dispatch based on local estimation of
power mismatch,” IEEE Trans. Smart Grid, vol. 9, no. 6, pp. 5930–5942,
cal control scheme without the primary frequency droop control Nov. 2018.
of ESs and controllable load; 3) the proposed completed three- [9] G. Hug, S. Kar, and C. Wu, “Consensus innovations approach for dis-
layer hierarchical control scheme. It is evident that the proposed tributed multiagent coordination in a microgrid,” IEEE Trans. Smart Grid,
vol. 6, no. 4, pp. 1893–1903, Jul. 2015.
scheme 3) can effectively regulate the system frequency and the [10] S. Kar and G. Hug, “Distributed robust economic dispatch in power sys-
maximal frequency deviation is near zero as the blue line shown tems: A consensus+innovations approach,” in Proc. IEEE Power Energy
in Fig. 13. However, the scheme 1) or 2) could only regulate the Soc. General Meeting, Jul. 22–26 2012, pp. 1–8.
5196 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 34, NO. 6, NOVEMBER 2019
[11] Y. Wang, L. Wu, and S. Wang, “A fully-decentralized consensus-based [34] X. Luo, S. Xia, and K. W. Chan, “A decentralized charging control strat-
ADMM approach for DC-OPF with demand response,” IEEE Trans. Smart egy for plug-in electric vehicles to mitigate wind farm intermittency and
Grid, vol. 8, no. 6, pp. 2637–2647, Nov. 2017. enhance frequency regulation,” J. Power Sour., vol. 248, pp. 604–614,
[12] L. Meng, T. Dragicevic, J. Vasquez, J. Guerrero, and E. R. Sanseverino, Feb. 2014.
“Hierarchical control with virtual resistance optimization for efficiency [35] F. Milano, “An open source power system analysis toolbox,” IEEE Trans.
enhancement and State-of-Charge balancing in DC microgrids,” in Proc. Power Syst., vol. 20, no. 3, pp. 1199–1206, Aug. 2005.
IEEE 1st Int. Conf. DC Microgrids, Jun. 7–10, 2015, pp. 1–6.
[13] A. Mehrizi-Sani and R. Iravani, “Potential-function based control of a Shiwei Xia (M’12) received the B.Eng. and M.Eng.
microgrid in islanded and grid-connected modes,” IEEE Trans. Power degrees in electrical engineering from the Harbin In-
Syst., vol. 25, no. 4, pp. 1883–1891, Nov. 2010. stitute of Technology, Harbin, China, in 2007 and
[14] J. Tan and L. Wang, “A game-theoretic framework for vehicle-to-grid fre- 2009, respectively, and the Ph.D. degree in power sys-
quency regulation considering smart charging mechanism,” IEEE Trans. tems from The Hong Kong Polytechnic University,
Smart Grid, vol. 8, no. 5, pp. 2358–2369, Sep. 2016. Hung Hom, Hong Kong, in 2015. Then, he worked
[15] J. Tan and L. Wang, “Coordinated optimization of PHEVs for frequency as a Research Associate and subsequently as a Post-
regulation capacity bids using hierarchical game,” in Proc. IEEE PES doctoral Fellow with the Department of Electrical En-
General Meeting, Jul. 2015, pp. 1–5. gineering, The Hong Kong Polytechnic University, in
[16] J. W. Simpson-Porco, Q. Shafiee, F. Dörfler, J. C. Vasquez, J. M. 2016–2018. He is currently with the State Key Lab-
Guerrero, and F. Bullo, “Secondary frequency and voltage control of is- oratory of Alternate Electrical Power System with
landed microgrids via distributed averaging,” IEEE Trans. Ind. Electron., Renewable Energy Sources, School of Electrical and Electronic Engineering,
vol. 62, no. 11, pp. 7025–7038, Nov. 2015. North China Electric Power University, Beijing, China, and also with The Hong
[17] X. Lu, X. Yu, J. Lai, J. M. Guerrero, and H. Zhou, “Distributed secondary Kong Polytechnic University, Hong Kong. His research interests include secu-
voltage and frequency control for islanded microgrids with uncertain com- rity and risk analysis for power systems with renewable energies, distributed
munication links,” IEEE Trans. Ind. Inform., vol. 13, no. 2, pp. 448–460, optimization and control of multiple sustainable energy sources in active distri-
Apr. 2017. bution network, and smart grid.
[18] M. A. Mahmud, M. J. Hossain, H. R. Pota, and N. K. Roy, “Nonlinear
distributed controller design for maintaining power balance in Islanded
microgrids,” in Proc. IEEE PES General Meeting | Conf. Expo., Jul. 27– Siqi Bu (S’11–M’12–SM’17) received the Ph.D. de-
31, 2014, pp. 1–5. gree from the electric power and energy research
[19] Q. Shafiee, J. M. Guerrero, and J. C. Vasquez, “Distributed secondary cluster, The Queen’s University of Belfast, Belfast,
control for islanded microgrids-a novel approach,” IEEE Trans. Power U.K., in 2012, where he continued his Postdoctoral
Electron., vol. 29, no. 2, pp. 1018–1031, Feb. 2014. research work before entering industry. Subsequently,
[20] A. J. Wood and B. F. Wollenberg, Power Generation, Operation, and he joined National Grid UK as a Power System En-
Control. Hoboken, NJ, USA: Wiley, 2012. gineer and then became an experienced UK National
[21] X. Luo, S. W. Xia, K. W. Chan, and X. Lu, “A hierarchical scheme for Transmission System Planner and Operator. He is
utilizing plug-in electric vehicle power to hedge against wind-induced an Assistant Professor with The Hong Kong Poly-
unit ramp cycling operations,” IEEE Trans. Power Syst., vol. 33, no. 1, technic University, Kowloon, Hong Kong, and also
pp. 55–69, Jan. 2018. a Chartered Engineer with UK Engineering Council,
[22] E. Ela and M. O’Malley, “Studying the variability and uncertainty impacts London, U.K. His research interests include power system stability analysis
of variable generation at multiple timescales,” IEEE Trans. Power Syst., and operation control, including wind power generation, PEV, HVDC, FACTS,
vol. 27, no. 3, pp. 1324–1333, Aug. 2012. ESS, and VSG.
[23] A. Y. Kibangou, “Graph Laplacian based matrix design for finite-time He was a recipient of various prizes due to excellent performances and out-
distributed average consensus,” in Proc. Amer. Control Conf., Jun. 27– standing contributions in operational and commissioning projects during the
29, 2012, pp. 1901–1906. employment with National Grid UK. He is also the recipient of Outstanding
[24] T. M. D. Tran and A. Y. Kibangou, “Distributed estimation of graph Reviewer Awards from the IEEE TRANSACTIONS ON SUSTAINABLE ENERGY,
laplacian eigenvalues by the alternating direction of multipliers method,” IEEE TRANSACTIONS ON POWER SYSTEMS, Renewable Energy and Interna-
IFAC Proc. Vol., vol. 47, no. 3, pp. 5526–5531, 2014. tional Journal of Electrical Power and Energy Systems.
[25] A. Y. Kibangou and C. Commault, “Decentralized Laplacian eigenvalues
estimation and collaborative network topology identification,” IFAC Proc. Can Wan (M’15) received the B.Eng. and Ph.D. de-
Vol., vol. 45, no. 26, pp. 7–12, 2012. grees from Zhejiang University, Hangzhou, China, in
[26] E. Ghadimi, A. Teixeira, I. Shames, and M. Johansson, “Optimal parameter 2008, and The Hong Kong Polytechnic University, in
selection for the alternating direction method of multipliers (ADMM): 2015.
Quadratic problems,” IEEE Trans. Autom. Control, vol. 60, no. 3, pp. 644– He is a Research Professor with the Col-
658, Mar. 2015. lege of Electrical Engineering, Zhejiang University,
[27] T. H. Chang, M. Hong, and X. Wang, “Multi-agent distributed optimization Hangzhou, China, under the university Hundred Tal-
via inexact consensus ADMM,” IEEE Trans. Signal Process., vol. 63, ents Program. He was a Postdoc Fellow with the De-
no. 2, pp. 482–497, Jan. 2015. partment of Electrical Engineering, Tsinghua Uni-
[28] S. Boyd, N. Parikh, E. Chu, B. Peleato, and J. Eckstein, “Distributed versity, Beijing, China, and held research positions
optimization and statistical learning via the alternating direction method with the Technical University of Denmark, The Hong
of multipliers,” Found. Trends Commun. Inf. Theory, vol. 1, no. 3, pp. 1– Kong Polytechnic University, and the City University of Hong Kong. He was
122, 2011. a Visiting Scholar with the Center for Electric Power and Energy, Technical
[29] A. Khalil and Z. Rajab, “Load frequency control system with smart me- University of Denmark, and Argonne National Laboratory, Lemont, IL, USA.
ter and controllable loads,” in Proc. 8th Int. Renewable Energy Congr., His research interests include forecasting, renewable energy, active distribution
Amman, Jordan, Mar. 21–23, 2017, pp. 1–5. network, integrated energy systems, and machine learning.
[30] C. Jiang, B. Li, and J. Shen, “Controllable load management approaches
in smart grids,” Energies, vol. 2015, no. 8, pp. 11187–11202, 2015.
Xi Lu received the B.Eng. degree in electrical engi-
[31] T. Kim and W. Qiao, “A hybrid battery model capable of capturing dy-
neering from North China Electric Power University,
namic circuit characteristics and nonlinear capacity effects,” IEEE Trans.
Beijing, China, in 2015. He is currently working to-
Energy Convers., vol. 26, no. 4, pp. 1172–1180, Dec. 2011. ward the Ph.D. degree in electrical engineering with
[32] B. Venkatesh, S. Chandramohan, N. Kayalvizhi, and R. P. Kumudini Devi,
the Hong Kong Polytechnic University, Hong Kong.
“Optimal reconfiguration of radial distribuion system using artificial intel-
His research interests include application of robust
ligence methods,” in Proc. IEEE Toronto Int. Conf. Sci. Technol. Humanity,
optimization and distributionally robust optimization
Sep. 26/27, 2009, pp. 660–665.
in power system operation.
[33] S. J. Ahn, S. R. Nam, J. H. Choi, and S. I. Moon, “Power scheduling of
distributed generators for economic and stable operation of a microgrid,”
IEEE Trans. Smart Grid, vol. 4, no. 1, pp. 398–405, Mar. 2013.
XIA et al.: FULLY DISTRIBUTED HIERARCHICAL CONTROL FRAMEWORK FOR COORDINATED OPERATION OF DERS IN ADPN 5197
Ka Wing Chan (M’98) received the B.Sc. (Hons) and Bin Zhou (S’11–M’13–SM’17) received the B.Sc.
Ph.D. degrees in electronic and electrical engineering degree in electrical engineering from Zhengzhou Uni-
from the University of Bath, Bath, U.K., in 1988 and versity, Zhengzhou, China, in 2006, the M.S. degree
1992, respectively. He is currently an Associate Pro- in electrical engineering from the South China Uni-
fessor and Associate Head with the Department of versity of Technology, Guangzhou, China, in 2009,
Electrical Engineering, the Hong Kong Polytechnic and the Ph.D. degree from The Hong Kong Polytech-
University, Hong Kong. His general research interests nic University, Hong Kong, in 2013. Afterwards, he
include power system stability, analysis and control, was as a Research Associate and subsequently a Post-
power grid integration, security, resilience and opti- doctoral Fellow with the Department of Electrical En-
mization, and demand response management. gineering, The Hong Kong Polytechnic University.
He is currently an Associate Professor with the Col-
lege of Electrical and Information Engineering, Hunan University, Changsha,
China. His main fields of research include smart grid operation and planning,
renewable energy generation, and energy efficiency.