TPEL

1

Abstract—Electric spring (ES) was originally proposed as

a distributed demand-side management (DSM) technology

for stabilizing power distribution network in the presence

of intermittent power generation without using

communication. This paper explores the practical use of

consensus control for a cluster of electric springs (ESs)

through a WiFi communication layer for new functions not

previously realized in practice. This approach can be

considered as a form of DSM for smart grid technology. A

novel consensus control is introduced to enable distributed

ES circuits to provide local voltage and system frequency

regulations in a microgrid with shared responsibility of

active and reactive power compensation. The practical

implementation details of consensus control for a cluster of

ESs are addressed. New plug-and-play functions of ESs are

practically demonstrated for the first time under consensus

control. Practical results indicate that droop control

(without communication) and consensus control (with

communication) are complementary. Under normal

condition when the communication network is available,

distributed ESs can perform with shared power

compensation efforts based on consensus control. If the

communication network fails, ESs can revert to perform

under droop control.

Index Terms—Consensus control, droop control, demand-side

management, distributed control, electric spring, microgrid.

I. INTRODUCTION

NSTANTANEOUS balance of electric “power generation”

and “power demand” is a fundamental requirement for power

system stability. If the percentage of renewable power

generation is negligible, utility companies can adopt the

traditional control paradigm of “power generation following

power demand”. Controlling power generation/supply is a type

of “supply-side management” (SSM) with which the utility

companies generate electric power to meet the load demand. In

the emerging power grids with increasing penetration of

Manuscript received November 23, 2017; revised February 19, April 16 and

August 14, 2018; accepted September 20, 2018. This work was supported in part by the Hong Kong Research Grant Council under the theme-based project

T23-701/14-N.

J. Chen, S. Yan, T. Yang and S.C. Tan are with the Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong (e-mail:

distributed renewable energy generation of intermittent nature

(known or unknown to the utility companies), the total power

generation/supply becomes difficult to predict and control in

real time. This situation will be worsen when the penetration of

intermittent renewable energy becomes substantial. Under such

situation, the power generation on the supply side is changing

dynamically with the wind and solar power profiles. Any

mismatch of power generation and demand could lead to

fluctuations in the mains voltage and frequency. Generally,

mismatch of active power leads to frequency variations while

mismatch of reactive power results in voltage variations.

Therefore, utility companies would have difficulty in ensuring

that the local ac mains voltage can meet the typically +/-5%

tolerance, because tap changing of local transformers is too

slow to cope with fast transients. For a weak power grid or a

microgrid using small generators with low inertia, the

frequency fluctuation could become a serious issue and a threat

of power system collapse.

An alternative control paradigm of “power demand

following power generation” has been suggested for future

power grid with a high percentage of intermittent renewable

energy generation [1]. If power consumption of some loads can

vary adaptively to follow the fluctuating renewable power

generation profile, the power demand and power generation can

be balanced in real time [2]. Modulating the load consumption

adaptively to achieve power balance in a power grid is a type of

“Demand-Side Management” (DSM). In practice, SSM and

DSM are two approaches that can play a part in achieving

power balance. They are not mutually exclusive and can be

complementary. For example, consensus control of distributed

generators has been proposed to maintain voltage stability [13].

Fig.1 (a) shows a simplified schematic of using distributed

generators for providing regulated mains voltage at the point of

common coupling (PCC). Curtailment of wind and solar power

is one example of SSM [3]. However, using SSM itself is

insufficient in maintaining local voltage stability particularly

when the load is remotely connected to the local transformer

through a long distribution line (e.g. in rural areas) as shown in

Fig.1 (b). If DSM can be applied on the remote load in Fig.1(b),

the local load voltage can also be regulated. This important

jiechen@eee.hku.hk; yanshuo@connect.hku.hk; yang2014@connect.hku.hk;

sctan@eee.hku.hk). S. Y. R. Hui is with the Department of Electrical and Electronic Engineering,

The University of Hong Kong, Hong Kong, and also with the Department of

Electrical and Electronic Engineering, Imperial College London, London SW7 2AZ, U.K. (e-mail: ronhui@eee.hku.hk; r.hui@imperial.ac.uk).

Practical Evaluation of Droop and Consensus

Control of Distributed Electric Springs for Both

Voltage and Frequency Regulation in Microgrid

Jie Chen, Student Member, IEEE, Shuo Yan, Member, IEEE, Tianbo Yang, Student Member, IEEE,

S.C. Tan, Senior Member, IEEE, S.Y. (Ron) Hui, Fellow, IEEE

I

TPEL

2

point has recently been highlighted by a power company in [23].

This paper focuses on the control and use of distributed ESs

as a DSM method for achieving instantaneous power balance

and hence mitigation of fluctuations in both frequency and

voltage in a microgrid. Several forms of ES technologies have

been evaluated by various research groups [2],[4]-[9] as a new

form of distributed DSM for maintaining voltage and/or

frequency stability in smart grid fed with intermittent renewable

energy sources. ESs are power electronics circuits designed to

be distributed over the distribution power grid for providing

local active and reactive power compensation for regulating the

local mains voltage and/or system frequency [10]. The original

versions of ES were designed to work independently under

droop control [11] and without the requirement of wired or

wireless communication. Being able to work independently

without communication and internet control, ESs can provide

local support for the power grid even when major central

control structures fail. However, under normal conditions,

coordination of ES operations could offer new features not

previously contemplated. This project aims at exploring the

practical implementation issues of a distributed control for

coordinating a cluster of ESs for regulating both ac mains

voltage and frequency in a mircogrid.

Among various distributed control methods, consensus

control has gained substantial attention recently. It is an

algorithm developed for ensuring a global agreement regarding

a common quantity of interest through limited information

exchange between neighbours [12]. Agents in this kind of

system are typically equipped with distributed computing and

communication devices. Without any global information, each

agent has access to its local measurement and neighbours’

information only. For energy related applications, consensus

control has been applied to coordinate the operation of

distributed generators [13], speed regulation of induction

motors [14], energy storage systems [15] and sharing of

distributed reactive power compensation [16]. Consensus

control of distributed generators (DGs) is a type of SSM

because the power inverters involved are connected to the

“renewable energy sources” such as solar panels or wind

generators that supply electrical power. In this paper, we focus

on both droop and consensus control of distributed ESs, which

is a type of DSM because the ESs are controlled to modulate

the power consumption of their associated “noncritical loads”.

While droop control does not require communication and

consensus control require communication, they are also not

exclusive and are in fact complementary. It will be shown in

this paper that consensus control of distributed ESs provide

extra features such a good sharing of responsibility in the

process of voltage and frequency regulation, while droop

control can be activated when the communication layer for

consensus control fails.

Simulation studies of consensus control on ESs have been

reported in several previous works. The first simulation study

of using consensus control of ES is reported in [17]. A

hierarchical control scheme is proposed to reduce the grid

voltage steady state error. However, the use of simplified model

of ES only allows the implementation of reactive power

compensation. Moreover, the reactive power sharing ability of

ES with consensus control is not realized. In [18], a leader-

following consensus control is developed to coordinate

distributed ESs to support the chosen critical buses voltage. The

work is further improved to form a two-level voltage

management scheme in [19]. The proposed control shows good

reactive power sharing for voltage regulation performance.

However, the method is restricted to reactive power

compensation for voltage regulation only and does not cover

system frequency fluctuation issue. Also, the control system

could be at risk and would fail upon the loss of the leader of

consensus control.

The common drawback of previous works using simplified

models of ES in system level simulation is the insufficient

consideration of utilizing the active power capability of ES in

tackling frequency fluctuation. Meanwhile, practical

implementation of consensus control of ES involving both

active and reactive power compensation for regulating both ac

mains voltage and frequency in microgrid has not been

previously reported. The new contributions of this paper (over

existing works of distributed control of ES) include the

following novel aspects:

1) This paper is the first report on the practical design and

implementation of “consensus control of ES” for “BOTH grid

voltage AND frequency regulation”. The experimental setup

consists of (i) a physical layer comprising the hardware setup

of several ESs distributed in a small-scale power grid

environment, (ii) a WiFi communication network and (iii) a

control layer.

2) The proposed consensus control of ES has a good

voltage/frequency regulation and active/reactive power sharing

performance. For the first time, the plug-and-play capability of

ES is practically demonstrated to show the control robustness

against hardware failure.

3) An experimental comparison study between droop

control and consensus control is included to verify the

advantages of consensus scheme. Previously, a cluster of ESs

have been demonstrated to work independently through droop

control, but good sharing of reactive power compensation is not

guaranteed [11]. The use of consensus control allows new

functions such as improved sharing of compensation efforts

among ESs in the cluster. Hence, the utilization of consensus

control adds new functions to the coordinated control of

distributed ESs, while droop control can stay as the last resort

of control when the communication network fails.

G

G

feeder

DGs PCC

local load

(a)Local load voltage control

G

G

feeder

DGs PCC

distribution line

remoteload

(b)Remote load voltage control

voltage control pointload voltage point

Fig. 1. Schematic of a distribution line fed by distributed generators with (a)

a load closely connected to the PCC and (b) a load remotely connected to the

PCC through a long distribution line.

TPEL

3

II. DISTRIBUTED ES IN MICROGRID

A. Microgrid setup

Consider an islanded microgrid comprising a weak ac power

source and a cluster of loads as shown in Fig.2. The ac power

source is programmed to emulate a weak grid fed with

intermittent renewable energy sources. A pre-recorded mains

voltage profile is stored in the programmable power source to

create a time-varying voltage fluctuation along the distribution

line. To emulate the frequency response of a weak grid with

small inertia, the AC power supply is programmed to behave

like a small-capacity spinning machine. Mismatch between

mechanical power and electrical power may cause the

frequency fluctuation. The cluster of loads consist of a mixture

of smart loads and critical loads. Each smart load consists of an

ES connected in series with a noncritical load [2]. The

noncritical load is one that can tolerate certain voltage variation

without causing consumer inconvenience. An example is an

electric water heater. The critical load requires a well-regulated

mains voltage and frequency within a tight tolerance.

The challenge in this study is how to coordinate the control

of the active power and reactive power consumption of a cluster

of distributed smart loads to regulate the voltage and frequency

(that is subject to the disturbance of the intermittent renewable

power generation). Distributed ESs operating under droop

control enjoy the underlying advantage of being independent of

wireless communication and internet control. This feature is

important because the droop control enables them to continue

to work even if the wireless communication network and

internet of the power company fail. The distributed feature of

ESs makes the system robust against potential hardware failure

of ES. It is therefore beneficial to retain the droop control for

abnormal situations while exploring new functions of

distributed control under normal conditions.

The schematic of the second version of ES is illustrated in

Fig.3. Implemented by a half bridge power inverter, this ES

provides a sinusoidal voltage (PWM voltage of the inverter

filtered by an LC filter) in series with a noncritical load. A

bypass switch will disconnect the ES out of the system in case

of hardware failure. The amplitude and phase of ES voltage (Ves)

directly affects the power fed into the noncritical load. A proper

power variation range of noncritical load is still necessary for

the safe operation of the electric devices. The energy storage

(such as batteries) in the DC link makes the bidirectional power

flow (±Pes, ±Qes) possible. To regulate the local voltage and

system frequency, the ES can operate in eight different modes

to modulate the total active and reactive power dynamically

[20]. Output voltage of batteries clamps the dc link voltage at a

stable value.

B. Analysis of Active and Reactive Power of the Smart Load

The key issue in AC microgrid control is the decoupled

control of active and reactive power for regulating grid

frequency and voltage respectively. Consider the case when the

noncritical load is a resistive load, the vector diagram of ES

voltage and current is shown in Fig.4 to demonstrate the

adjustment mechanism of active and reactive power of smart

load. Vs is the mains voltage of AC bus. A decoupled single-

phase d-q framework is developed by choosing the noncritical

load current (Inc) as the reference d vector. θ is the angle

between ES voltage and non-critical load current. The total

active power of smart load (Psl) consists of noncritical load

active power (Pnc) and ES active power (Pes). All reactive power

of smart load (Qsl) is contributed by the ES (Qes). Active and

reactive power of smart load can be calculated as: 2

_nc nc es d

sl nc es

nc

V V VP P P

R

+= + = (1)

_nc es q

esslnc

V VQ Q

R= = (2)

where Rnc is the resistance of noncritical load.

As shown in (1) and (2), the active power of smart load Psl is

directly controlled by the d-component of ES voltage (Ves_d),

while reactive power Qsl is mainly decided by q-component of

ES voltage (Ves_q).

C. Basic Control Scheme of Single ES

The primary level control of ES is devoted to regulating the

local voltage and frequency by tracking the reference points

generated by upper level control. Fig.5 shows the detailed

primary control diagram of ES_i, where voltage and frequency

control loops are highlighted in different colours. V s,nom

i and

ω nom

i are correspondingly the nominal voltage and frequency

points assigned by upper level control. V s

i is the local voltage

and ωi is the measured frequency of ES_i.

Fig. 2. Microgrid with a cluster of smart loads.

DistributionLine

DistributionLine

DistributionLine

RenewableSource

SmartLoad

SmartLoad

SmartLoad

CriticalLoad

Fig. 3. Schematic of an ES.

Battery

Battery

ncV

esV

ncI

NoncriticalLoad

sVesP esQ

LC Filter

BypassSwitch

Fig. 4. Vector diagram of ES in case of resistive load.

d

q

ncI

ncV

sV

esV es_qV

es_dV

θ

TPEL

4

As discussed in Section II-B, a synchronized rotating d-q

reference frame is developed with load current I nc

i chosen as the

reference vector in order to separate the control of active and

reactive power. The subscripts d and q represent the d and q

components of the corresponding variables in Fig.5. The

frequency fluctuation can be mitigated by controlling the d

component of the ES voltage (V es_d

i ) while the local voltage can

be regulated by modulating the q component of ES voltage

(V es_q

i ).

Both the voltage and frequency regulators have a dual-loop

structure. The inner loops are ES-voltage tracking loops and the

outer loops are the system voltage/frequency control loops. In

the outer loops, the voltage and frequency error signals are fed

into PI controllers to generate references for the inner ES-

voltage loops. (V es_d

i )* and (V es_q

i )* are the reference signals for

d and q components of ES voltage respectively. The inner ES-

voltage control loops then track the references and generate the

modulation signals u es_d

i and u es_q

i .

Most of the previous works focus on the primary level (droop)

control, which is independent, decentralized and

communication-free. Consider a cluster of ESs with only

primary level control located in various places, ESs are

designed to regulate local voltage and the system frequency

towards a common reference value. However, since the control

inputs (especially the local voltage values along the

distribution line) of the ESs and their system dynamics are

different, the operating states of ESs will be different. To

preserve the power sharing ability of ESs, a distributed

secondary control is introduced. The consensus-based

secondary control is implemented inside ES local controller. By

collecting both local and neighbour’s information, consensus

control then generates a compensation value to update the

nominal set point V s,nom

i and ω nom

i for the primary level control

of the ith ES (ES_i). A dynamic voltage observer is also

developed to regulate the global average voltage.

III. PROPOSED DISTRIBUTED SECONDARY CONSENSUS

CONTROL

As the primary level control of ES mainly deals with the local

voltage and frequency regulation problems, the power sharing

ability of distributed ESs has not been considered carefully in

practical implementation. A distributed secondary consensus

control is proposed to perform accurate proportional power

sharing control during the voltage and frequency regulation

processes. The nominal set point for the primary level control

V s,nom

i and ω nom

i will be dynamically updated in each secondary

control cycle. The basic control objectives of both secondary

control and primary control can be summarized as:

(1) Global average voltage regulation and accurate proportional

reactive power sharing for all ESs 1,2,..., N , i.e.,

{1,2,..., }i j N

1 ,( )

lim 0

N s

i i s ref

t

V tV

N

=

→− =

(3)

,max ,max

( )( )lim 0

esesji

es esti j

Q tQ t

Q Q→− = (4)

(2) Frequency regulation and accurate weighted active power

sharing for all ESs 1,2,..., N , i.e., {1,2,..., }i j N

lim ( ) 0ref

it

t →

− = (5)

,max ,max

( )( )lim 0

esesji

es esti j

P tP t

P P→− = (6)

Remark 1: P es

i,max and Q es

i,max are correspondingly the maximum

active and reactive power output of ES_i. The boundary of

power compensation is usually pre-determined by the electrical

power ratings of the inverters [13]. However, the maximum

power output of ES is also affected by the noncritical load

impedance, as discussed in [21]. Consider the case when

noncritical load R nc

i is a pure resistance, P es

i,max and Q es

i,max can be

derived as: , 2

,max

( )=

4

s refes

i

nc

VP

R (7)

, 2

,max

( )=

2

s refes

i

nc

VQ

R (8)

Assuming that the rated power of implemented inverters is large

enough, then P es

i,max and Q es

i,max are determined by (7) and (8) only.

The flow chart structure of the proposed consensus control is

illustrated in Fig.6 and a detailed block diagram is shown in

Fig.7. Each ES transfers a set of data ,max ,max

[ , , ]es es

s i i

i es es

i i

P QV

P Q to its

neighbours. �̅�𝑖𝑠 is ES_i’s estimation of the global average

Cri

tica

l

LoadNoncritical

Load

fLi

fCi

Li Ri

R inc

MainBus

VDC

R ic

ω i

nom

ω iω Δ iPI Frequency

Controller

PLLd/q

θ PLL

Ii

nc

PI Voltage

Controller

Q Axis

Controller

D Axis

Controller

SPWM

Synchro

Voltage Regulator

Frequency Regulator

Smart Load

( )Vies_q *

( )Vies_d *

Δ ViVi

s,nom

Vies

VDC

Vi

sVies_q

uies_d

uies_q

Fig. 5. Primary control loops of single ES.

Neighbor ES j

,V i

s

,P i,max

es

P i

es

Qi,max

es

Qi

es

,V j

s

,P j,max

es

P j

es

Qj,max

es

Qj

es

ES i

Consensus Control

Initialize

Pi,max

esV

s,ref ref, , Q i,max

es

,

V is

Qi

esPi

esI i

nc, , , [ES i data]

[ES j data]

Send/Receive data

Send [ES i data]Receive [ES j data]

Measure

Updateupdate

update Q

i P

i,

V is (9)

(13),(22)

Drive reference for

primary control

V i

s,nomωi

nom

,

Primary Control

Voltage observer

Replace measured

voltage by estimated

average voltage V is

V i

s,nom

ωi

nom

Frequency regulator

Voltage regulator

Inner ES-voltage

controller

SPWM

Modulation

signal

ES i

Fig. 6. Flow chart structure of the proposed consensus control.

TPEL

5

voltage. P es

i /P es

i,max and Q es

i /Q es

i,max are per-unit terms which

represent the active and reactive output compensation

percentage of ES_i, respectively.

As shown in Fig.6, the proposed secondary control starts with

initialization of the system reference voltage Vs,ref , reference

frequency ωref and maximum ES power capacity. After

measuring the local data, ES_i will send the consensus

information to its neighbours and receive neighbours’ data.

Voltage observer then updates the estimated global average

voltage according to (9). For power sharing purposes,

consensus algorithm generates the active and reactive power

error signals (δ P

i and δ Q

i ) according to (13) and (22). The

original system reference points will be compensated by these

two error signals, as shown in Fig.7. The shifted nominal points

V s,nom

i and ω nom

i are then transferred to primary level control as

references. For the primary level control, the measured local

voltage will be replaced by the estimated global average voltage

in order to achieve the global voltage regulation (will be

described in the following section).

A. Global voltage regulation and reactive power sharing

based on a new distributed voltage observer

Reactive power sharing control is required to prevent

overloading of individual ES. A distributed voltage observer

method is introduced into the ES controller for the first time.

Instead of controlling the local voltage directly (as in a

traditional droop control), each ES will have an estimated value

(�̅�𝑖𝑠) of the global average voltage and compare this estimated

value with the voltage reference. The estimated value is updated

through a dynamic consensus algorithm and all local estimated

values will converge to the global average voltage.

1) Voltage Observer

The voltage observer module is based on the dynamic

consensus algorithm as shown in Fig.8. Each ES receives the

neighbours’ estimated information ( )s

j iV j N . The voltage

observer will update its own estimate �̅�𝑖𝑠 by combining the

local measured voltage V s

i and neighbours’ information as

follow:

0( ) ( )+ ( ( ) ( ))

i

ts s s s

i i ij j i

j N

V t V t a V V d

= − (9)

Let 1 2[ , ,..., ]s s s s

NV V V V= ,1 2[ , ,..., ]s s s s

NV V V V= and

( )ij N NL l = (the Laplacian matrix of communication network),

the global dynamic of voltage observer can be formulated as

�̇̅�𝑠 = �̇�𝑠 − 𝐿�̅�𝑠 (10)

In the frequency domain, (10) can be written as 1( )s s −= +s s

NV I L V (11)

where

N N NI R . s

V and sV are the Laplace transforms of

voltage estimation vector �̇̅�𝑠 and voltage measurement vector sV , respectively. If the proposed communication graph is

undirected and connected, then L is irreducible and -1( )s s +

NI L is stable. A special left eigenvector

[1/ ,1/ ,...,1/ ]N N N = corresponds to the zero eigenvalue of

matrix L. Based on the proof in [12], all estimation values will

converge to the average value of global voltage,

1 ( )lim (t)= lim

N s

k ks

it t

V tV

N

=

→ →

(12)

2) Reactive power sharing and voltage regulation

A cooperative reactive power control loop can be

implemented in the voltage regulator. In each ES, the local

controller processes local per-unit reactive power and

neighbours’ per-unit reactive power to find the error signal δ Q

i :

,max ,max

=c ( )i

es esjQ Q i

i ij es esj N j i

Q Qa

Q Q

− (13)

where cQ is the coupling parameter between voltage and

reactive power regulator. The error signal is added to the

original voltage reference as a compensation to generate the

new nominal reference point V s,nom

i , ,=s nom s ref Q

i iV V + (14)

Fig. 7. Detailed block diagram of the proposed consensus control scheme.

DC

Sour

ce

SPWM LC Filter

Smart Load

Crit

ical

LoadNoncritical

Load

Mic

rogr

id

Communication Network

N2

1Electric Spring

Link

i

ES i information

,[V is

P i

es

P i,max

es

/ ,Qi

esQ

i,max

es

/ ]

Proposed Consensus Control

ES j information

,[V js

P j

es

P j,max

es

/ ,Qj

esQ

j,max

es

/ ]

=

i

P es es es es

i ij j j,max i i,max

j N

δ a (P (t) / P - P (t) / P )

i

Q es es es es

i ij j j,max i i,max

j N

δ = a (Q (t) / Q - Q (t) / Q )

ts s ss

i j ii ij0

V = V + a (V (t) -V (t))dt

Vs,ref

cQ

cP

iv

i

ref

s

iV

i

Inner Voltage Control Loop

PWM

,s nom

iVs

iV

s

i

Voltage Regulator

nom

iω

Frequency Regulator

Fig. 8. Voltage observer: dynamic consensus algorithm.

TPEL

6

As shown in Fig.7, the voltage estimation s

iV is compared

with the reference V s,nom

i and the error signal is then fed to a PI

controller to generate the ES q-component voltage reference (

V es_q

i )*

_ * , ,( ) = ( ) ( )s

es q V s nom s V s nomii P i i I iV K V V K V V− + − (15)

where K V

P and K V

I are the proportional and integrator gain of

voltage regulator, respectively. By differentiating (15)

(�̇�𝑖𝑒𝑠_𝑞

)∗

= 𝐾𝑃𝑉 (�̇�𝑖

𝑄− �̅�𝑖

𝑠̇ ) + 𝐾𝐼𝑉(𝑉𝑠,𝑟𝑒𝑓+𝛿𝑖

𝑄 − �̅�𝑖𝑠 ) (16)

In steady operation, it can be assumed that (�̇�𝑖𝑒𝑠_𝑞

)∗

= �̇�𝑖

𝑄=

�̅�𝑖𝑠̇ = 0 since the control inputs are the root-mean-square

values with DC components. Let , [1,1,...,1]s ref TV=s,refV ,

1 1,max 2 2,max ,max[ / , / ,..., / ]es es es es es es

N NQ Q Q Q Q Q=q and reorganize (16)

to describe the global steady state of the system:

0sL V− − =s,refV q (17)

Since [1/ ,1/ ,...,1/ ]N N N = is a left eigenvector of L

corresponding to zero eigenvalue. Multiplying the equation (17)

by λ on the left yields

,

1- - 0

sNs s ref i

i

VV V

N

== =s,ref

V (18)

As mentioned before, the estimated voltage value will reach a

consensus on the average measured voltage. By integrating (12)

into (18),

,

1

s

s ref N i

i

VV

N== (19)

The control objective of global voltage regulation is then

achieved. Combining (19) and (17) yields:

0L =q (20)

If the communication graph is connected and undirected,

then [1,1,...,1]=q is a stable equilibrium of system (20) [12].

Equivalently, the solution of (20) can be expressed as:

{1,2,..., }i j N

,max ,max

0

esesji

es es

i j

QQ

Q Q− = (21)

The proposed control can successfully carry out (i) global

average voltage regulation and (ii) proportional reactive power

sharing.

B. Frequency regulation and active power sharing

In this experiment, the programmable AC power supply is

controlled to emulate the frequency characteristic of a small

generator with low inertia. As defined by the swing equation,

any imbalance between mechanical power input and electrical

power output could cause the generator frequency to vary.

Typically, the primary frequency control operates at a timescale

up to tens of seconds. Then a governor is used to adjust the

mechanical power of the generator and restore its frequency to

the nominal value [22]. A cluster of ESs can monitor the

frequency deviation and reduce power imbalance by

manipulating the active power consumption on the load side.

The fast dynamic of ES will enable a much faster system

frequency recovery process in a time scale of several seconds.

A frequency control loop is introduced in the consensus control.

As seen in Fig.7, the mismatch of the active power is

expressed as an error signal δ P

i :

,max ,max

=c ( )i

es esjP P i

i ij es esj N j i

P Pa

P P

− (22)

where cP is the coupling parameter between frequency and

active power regulator. This signal is added to the frequency

reference to generate the new nominal reference point ω nom

i .

Similar to the voltage regulator, the error between the measured

frequency and nominal reference will be fed to a PI controller

to generate the ES d-component voltage reference (V es_d

i )*,

which has a differential form of:

(�̇�𝑖𝑒𝑠_𝑑

)∗

= 𝐾𝑃𝑓

(�̇�𝑖

𝑃− 𝜔𝑖̇ ) + 𝐾𝐼

𝑓(𝜔𝑟𝑒𝑓+𝛿𝑖𝑃 − 𝜔𝑖 ) (23)

where K f

P and K f

I are the proportional and integrator gain of

frequency regulator.

Let [1,1,...,1]ref T=refω ,

1 2[ , ,..., ]T

N =ω and

1 1,max 2 2,max ,max[ / , / ,..., / ]es es es es es es

N NP P P P P P=p , the global system

dynamic in steady state can be calculated as:

0L− − =refω p ω (24)

It is reasonable to assume that the measured frequency is

unique, i.e., 1 2= =...= N . Multiplying the equation (24) by λ

on the left yields ref

1 2ω = = =...= N (25)

By integrating (25) and (24) yields: {1,2,..., }i j N

,max ,max

0

esesji

es es

i j

PP

P P− = (26)

The control objectives of frequency regulation and proportional

active power sharing can therefore be derived.

IV. PRACTICAL IMPLEMENTATION AND EVALUATION OF

CONSENSUS CONTROL FOR A CLUSTER OF ELECTRIC SPRINGS

A. Hardware Setup for A Cluster of Distributed ESs in a

Microgrid

1) General overview of hardware setup

Fig.9 shows the schematic of the hardware setup of

consensus control for a cluster of distributed ESs in a microgrid.

It consists of three layers: 1) physical layer, 2) cyber layer and

3) control layer. The physical layer describes the electrical

connections among different ES agents. The communication

links between different agents are displayed in the cyber layer

intuitively. The control layer involves local controller of each

ES.

A practical case study of an 110V AC microgrid with five

distributed ESs is used to demonstrate the effectiveness of the

proposed consensus control. The distribution line impedances

are non-uniform. The distribution line-1 has a larger impedance

than the others to represent a longer distance between the power

source and loads (Fig.9). Lead acid battery cells with Model No.

LC-R127R2NA (and ratings of 12V, 7.2 Ah/20 hr) are used in

the experiment. System parameters are listed in TABLE I.

2) Communication layer topology: convergence requirement

and eigenvalues influence

A circular communication topology is applied in the cyber

layer to meet the convergence requirement of consensus

TPEL

7

algorithm. Graph 𝐺 = (𝑽, 𝑬) is a graph with nodes set 𝑽 =[1,2, … , 𝑁] and edges 𝑬 ⊆ 𝑽 × 𝑽. In our case, the nodes are the

distributed ESs and the edges represent the communication

links between different agents. According to the results in

reference [12], the graph G should be a connected undirected

graph in order to guarantee the consensus convergence. A graph

is undirected if the information flow between agents is

bidirectional. A graph is connected when there is a path

between every pair of vertices. The path may consist of several

communication links. In other words, there are no unreachable

agents. In case of circular communication topology, the

communication graph G is a connected undirected graph.

Therefore, the convergence requirement is achieved.

The eigenvalues of graph G will directly affect the time

response of consensus algorithm. As defined in graph theory,

the adjacency matrix A = (𝑎𝑖𝑗)𝑁×𝑁 has non-negative elements.

𝑎𝑖𝑗 = 1 if and only if there is communication link between node

i and j. Graph Laplacian matrix L is defined as 𝐿 = 𝐷 − 𝐴 ,

where D is the degree matrix of graph G. L is a symmetric

matrix with real eigenvalues 1 20= 2n , where

is the max degree of graph G. If the undirected graph is

connected, the zero eigenvalue is isolated. The second-smallest

eigenvalue 𝜆2 is called algebraic connectivity, which is a

measure of consensus convergence rate [12]. A larger 𝜆2 leads

to a faster convergence rate. At this stage of our study, the

communication topology is fixed because the number of

hardware units of ESs is relatively small. In future research for

large-scale system, the communication network will be an

important research topic.

B. Practical Implementation of the Consensus Control and

Communication Links

1) Detailed implementation of the communication-based

consensus control

A photograph of the experimental setup is shown in Fig.10.

Fig.11 shows the system structure of the practical

implementation. The dSPACE control interface has mux ADC

inputs and multiple PWM output channels. It is used to sense

and control five ESs in a single platform (i.e the physical layer).

For the control layer, five virtual control blocks are established

in the dSPACE. Each controller will collect local sensors’ data

and send real-time PWM signal to ES hardware.

To express the distributed feature of consensus control, the

cyber layer is implemented inside dSPACE. WiFi (WLAN

IEEE 802.11) communication modules are used as

communication links between adjacent ESs. Local

measurement data of ES_i will be processed by the WiFi blocks.

The output of WiFi blocks is the consensus information

received by ES_j.

2) WLAN communication link characteristics

In the tests, the distributed ESs are located in a small

microgrid. The WiFi modules allow a wireless communication

distance between adjacent ESs to be up to 150 meters typically.

The time delay of this kind of short distance communication is

very small (< 500ns typically). Compared with the ES’s

switching period (0.05 ms), this communication time delay is

negligible. (In the laboratory tests, the locations of the 5 ESs are

much closer in practice.) As shown in Fig.12, the WiFi module

in dSPACE consists of three parts: transmitter, receiver and

communication channel. The channel parameters are listed in

TABLE II.

AC Power Source

Noncritical Load 1

Critical Load 1

Critical Load 2

Critical Load 3

Critical Load 4

Critical Load 5

5

4 3

2

1

ES 1

Physical Layer

Cyber Layer

To Control Layer

Neighbors Information

Bus 1 Bus 2 Bus 3 Bus 4 Bus 5Distribution Line 1 Distribution Line 2 Distribution Line 3 Distribution Line 4 Distribution Line 5

Smart Load1

Vs1 Vs2 Vs3 Vs4 Vs5

DC

Link

1

Noncritical Load 2

ES 2

DC

Link

2

Smart Load2

Noncritical Load 3

ES 3

DC

Link

3

Smart Load3 Smart Load4

Noncritical Load 4

ES 4

DC

Link

4

Noncritical Load 5

ES 5

DC

Link

5

Smart Load5

Fig. 9. Hardware setup of a cluster of distributed ESs in a microgrid. Fig. 10. Hardware setup devices.

DSPACE

Controller

LoadSmart

Load

Distribution

Line

TABLE I PARAMETERS OF THE EXPERIMENTAL SYSTEM

Digital Controller

dSPACE DS1006 Board

dSPACE DAC/ADC board DS2002 MUX AD

Microgrid

Distribution line 1 resistance 0.4 Ω Distribution line 1 inductance 5 mH

Distribution line 2-5 resistance 0.2 Ω

Distribution line 2-5 inductance 1.2 mH Non-critical load 1-2 resistance 40 Ω

Non-critical load 1-2 rated power 302.5 W

Non-critical load 3-5 resistance 60 Ω

Non-critical load 3-5 rated power 201.7 W

Critical load 1-5 resistance 180 Ω

Critical load 1-5 rated power 67.2 W Power Source voltage (RMS) 120 V

ESs

Switching frequency 20 kHz

DC link voltage 300 V Output voltage range(peak) 0-110 Vac

DC link capacitor 4500 uF

Filter inductor 500 uH Filter capacitor 13.2 uF

PI Controller

Voltage controller proportional gain (K V

P ) 4

Voltage controller integral gain (K V

I ) 2 Frequency controller proportional gain (K

f

P) 20

Frequency controller integral gain (K f

I ) 20

TABLE II COMMUNICATION LINK CHARACTERISTICS

Communication channel

Communication distance < 150 m

Signal-to-noise ratio (Es/N0) 20 dB

Time delay < 500 ns

TPEL

8

C. Performance Assessment of the Proposed Consensus

Control

The proposed distributed consensus control has been

practically evaluated in the 110V micro grid with a cluster of

five ESs for system performance assessment including voltage

regulation, frequency regulation, active and reactive power

sharing.

1) Voltage regulation and accurate reactive power sharing

Fig.13 shows the variations of the local mains voltage at the

locations of the five ESs. Initially (time t from 0 - 30s), the ESs

are not activated. Fig.13(a) shows the wide variation of the

voltage along the line. Some of the local mains voltages fall out

of the acceptable range (105 - 115V). The five ESs are activated

with the proposed control at t = 30s. As a result, the five local

mains voltages are restored to the desired value 110V 2V as

shown in Fig.13(a). Fig.13(b) shows the per-unit reactive power

compensation states of the ESs. After the ESs are activated at t

= 30s, good reactive power sharing is reached at around t = 50s.

At t = 60s, a step change of generator voltage of +3 V is injected.

Fig.13(a) and (b) confirm that good voltage regulation and

sharing of reactive power compensation can be practically

achieved. The variations of the ESs’ estimated global average

voltage values are recorded in Fig.13(c). It can be seen that they

converge to the global voltage, which is tuned at 110V.

2) Frequency regulation and accurate active power sharing

The use of distributed ESs is highly effective in improving

the frequency response of a weak grid. The frequency of the

microgrid implemented with ESs recovers much faster than that

without ES. Step load change tests are then carried out to

evaluate the frequency control and active power sharing. A

critical load of 100Ω is connected at t = 20s. Without using ESs,

the frequency control of this generator will take over 30s for the

frequency to be restored and the peak frequency deviation is

0.009 p.u (Fig.14(a)). The large deviation is caused by a small

inertia of the generator. Fig.14(b) shows the frequency profile

under the same step load change when the ESs are activated

under consensus control. It takes only about 3s for the

frequency to be restored and the peak frequency deviation is

reduced to 0.0048 p.u. Fig.14(c) shows the variations of the

active power of the five ESs. [Note: Negative active power in

Fig.14(c) means that the smart load is reducing active power.]

Two important observations should be noted. First, the ESs

react to the sudden increase in critical load power at t = 20s by

providing active power through a reduction of noncritical load

power and through their battery sources. This fast reaction

reduces quickly the imbalance between power supply and

demand and consequently the system frequency recovery time.

Second, the consensus control ensures good active power

sharing among the ESs in this power balancing mechanism.

DSPACE

ES1

Controller1

SensorPWM

Controller2 Controller3 Controller4 Controller5

ES2

SensorPWM

ES3

SensorPWM

ES4

SensorPWM

ES5

SensorPWM

LocalData

LocalData

LocalData

LocalData

LocalData

Cyber Layer

Control Layer

Fig. 11. Detailed implementation structure of consensus control.

MODULE

Data InTransmitter Channel Receiver

Data Out

DSPACE

ES LOCATION

ES1 ES2

ES3

ES4ES5

Fig. 12. dSPACE WiFi module and ES locations.

0 10 20 30 40 50 60 70 80 90

VsR

MS

(V)

Main Voltage1 RMS

Main Voltage2 RMS

Main Vlotage4 RMS

Main Voltage3 RMS

Main Voltage5 RMS

Time(s)

STEP

ES OFF ES ON

98

106

102

110

114

(a)

0

0.4

0.2

0.6

0.8

1.0

ES1 Reactive Power

ES2 Reactive Power

ES4 Reactive Power

ES3 Reactive Power

ES5 Reactive Power

Rea

ctiv

e P

ow

er(

p.u

.)

ES OFF ES ON

0 10 20 30 40 50 60 70 80 90

Time(s)

STEP

(b)

Est

imat

ed V

olt

age(

V)

ES1 Estimated Voltage

ES2 Estimated Voltage

ES4 Estimated Voltage

ES3 Estimated Voltage

ES5 Estimated Voltage

0 10 20 30 40 50 60 70 80 90

Time(s)

ES OFF ES ON

STEP

98

106

102

110

114

(c)

Fig. 13. Practical performance of voltage regulation and reactive power

sharing. (a) Mains voltages along the distribution line. (b) Per-unit reactive power output. (c) Estimated global average voltages.

TPEL

9

3) Coordinated voltage and frequency regulation with step

load change

Tests on the coordinated voltage and frequency regulation

performance have been conducted. The system begins with ESs

working in reactive power compensation mode. A critical load

of 100Ω is plugged in and out at t = 40s and t = 80s, respectively.

The practical measurements of (a) the local voltages of the ESs,

(b) the reactive power compensation efforts of the ESs, (c) the

system frequency and (d) the active power of the ESs are

included in Fig.15. It can be seen that the consensus control has

enabled this cluster of ESs to successfully achieve simultaneous

local voltage and system frequency regulations in this

microgrid setup. In the frequency response period, the local

voltages are still regulated close to 110V with a shared steady-

state reactive power output ranging from 0.32 p.u. to 0.4 p.u.

D. Plug-and-play Test

The plug-and-play capability of the proposed control has also

been tested for the first time. Initially, ES1 to ES4 are working.

ES5 is plugged in at t = 40s and two communication links ES1-

ES5 and ES4-ES5 are established. Then ES4 is plugged out at t

= 80s, and the communication links of ES3-ES4 and ES4-ES5

are disconnected. Fig.16 shows the (a) the mains voltages at the

locations of the five ESs, (b) the active power and (c) reactive

power of the five ESs. It can be seen from Fig.16(a) that the

mains voltages can be regulated after the plug-in and plug-out

processes. Fig.16(b) and Fig.16(c) show that the ESs can share

their responsibilities of active and reactive compensations well.

These results confirm plug-in and plug-out capability of the ESs

with consensus control.

0 10 20 30 40 50 60

Time(s)

Fre

quen

cy(p

.u.)

0.992

1.000

0.9960.009

t=30.0sΔ

(a)

Time(s)

Fre

qu

enc

y(p

.u.)

0 1 2 3 4 5 6 7 8 9

0.994

0.998

0.996

1.000

1.002

0.0048

t=3.0sΔ

(b)

0 1 2 3 4 5 6 7 8 9-0.8

-0.6

-0.2

-0.4

0

Act

ive

Pow

er(

p.u

.)

Time(s)

ES1 Active Power

ES2 Active Power

ES4 Active Power

ES3 Active Power

ES5 Active Power

(c)

Fig. 14. Practical performance of frequency regulation and active power sharing. (a) Frequency of a.c. generator (i.e grid frequency). (b) Frequency

response with ESs. (c) Per-unit active power output of a group of ESs.

0 20 40 60 80 100 120

106

110

108

112

114

VsR

MS

(V)

Time(s)

V1 RMS V2 RMS

V4 RMSV3 RMS

V5 RMS

LOAD IN LOAD OUT

(a)

0 20 40 60 80 100 120

0.28

0.36

0.32

0.4

0.44

Time(s)

Rea

ctiv

e P

ow

er(

p.u

.)

ES1 Reactive Power

ES2 Reactive Power

ES4 Reactive Power

ES3 Reactive Power

ES5 Reactive Power

(b)

Fre

quen

cy(p

.u.)

0.996

1.000

0.998

1.002

1.004

0 20 40 60 80 100 120

Time(s)

LOAD IN LOAD OUT

(c)

-0.4

-0.3

-0.1

-0.2

0

0.1

ES1 Active PowerES2 Active Power

ES4 Active PowerES3 Active Power

ES5 Active Power

Act

ive

Po

we

r(p

.u.)

0 20 40 60 80 100 120

Time(s) (d)

Fig. 15. Practical performance of coordinated frequency and voltage

regulation with step load change. (a) Voltage response at the locations of a group of ESs. (b) Per-unit reactive power output of a group of ESs. (c)

Frequency response of the grid. (d) Per-unit active power output of a group of

ESs.

TPEL

10

E. Practical Comparison between Droop Control and

Consensus Control

Droop control has been successfully used previously to allow

a group of ESs without communication links to work in a

coordinated manner [11]. Droop control is important in a sense

that it offers the last defense in the control of ESs when the

communication network fails. However, under normal situation,

consensus control can provide much improved performance.

This section presents a practical comparison of the droop

control and consensus control of 5 ESs distributed in a

distribution line with a nominal mains voltage of 110V.

The first set of comparative tests are conducted under 3

consecutive stages: (1) with the five ESs disabled, (2) with the

five ESs activated under the droop control and (3) with the five

ESs activated under consensus control. Each stage lasts for 30

seconds. Fig.17(a) and Fig.17(b) show the measured mains

voltages and reactive power of the distributed ESs, respectively.

When the ES are turned off in the first stage, the five local

voltages decrease with their distance from the power source

because of the voltage drop along the distribution line. During

the first stage, the reactive powers of the ES are zero as shown

in Fig.17(b). When the droop control is activated in the second

stage, the mains voltage values at the locations of the distributed

ESs are regulated close to the nominal value of 110V, but the

ESs do not share their reactive power compensation equally.

After the consensus control is turned on, the measured mains

voltages along the distribution line are regulated close to the

nominal value, while the ESs have good sharing of reactive

power compensation.

106

110

108

112

114

V1 RMSV2 RMS

V4 RMSV3 RMS

V5 RMS

VsR

MS

(V)

0 20 40 60 80 100 120

Time(s)

ES5 PLUG IN

ES4PLUG OUT

(a)

-0.6

-0.4

0

-0.2

Act

ive

Po

we

r(p

.u.)

ES1 Active Power

ES2 Active Power

ES4 Active Power

ES3 Active Power

ES5 Active Power

0 20 40 60 80 100 120

Time(s) (b)

0 20 40 60 80 100 120

0

-0.1

0.1

0.5

0.4

0.3

0.2

Time(s)

Rec

tiv

e P

ow

er(p

.u.)

ES1 Reactive Power

ES2 Reactive Power

ES4 Reactive Power

ES3 Reactive Power

ES5 Reactive Power

(c)

Fig. 16. Practical plug-and-play capability study. (a) Voltage response. (b)

Per-unit active power output. (c) Per-unit reactive power output.

0 10 20 30 40 50 60 70 80 90

VsR

MS

(V)

Time(s)

V1 RMS

V2 RMS

V4 RMS

V3 RMS

V5 RMS98

106

102

110

114 ES OFF ES ON:Droop Control

ES ON:Consensus Control

(a)

0

60

30

Qes

(Var

)

ES1 Reactive Power

ES2 Reactive Power

ES4 Reactive Power

ES3 Reactive Power

ES5 Reactive Power

0 10 20 30 40 50 60 70 80 90

Time(s)

ES OFF ES ON:Droop Control

ES ON:Consensus Control

(b)

Fig. 17. Practical comparison between droop control and consensus control:

One steady operation state. (a) Mains voltage comparison. (b) Reactive power

comparison.

(a)

Qes

(Var

)

ES1 Reactive Power

ES2 Reactive Power

ES4 Reactive Power

ES3 Reactive Power

ES5 Reactive Power

-150

-100

0

-50

50

100

150

100 300 400 500 7006002000

Time(s)

ES OFFES ON:

Droop ControlES ON:

Consensus Control

(b)

Fig. 18. Practical measurements for comparison between droop control and consensus control: Long time operation with fluctuating source. (a)Mains

voltage comparison. (b)Reactive power comparison.

100 300 400 500 7006002000

Time(s)

VsR

MS

(V)

V1 RMSV2 RMS

V4 RMSV3 RMS

V5 RMS

100

95

90

85

105

115

110

120ES OFF

ES ON:Droop Control

ES ON:Consensus Control

TPEL

11

The second set of comparative tests has been conducted with

the power source programmed as a renewable energy source of

intermittent nature. The voltage source is programmed to

fluctuate for 250 seconds. TABLE.III shows the voltage

fluctuation profile. The voltage source will change output

voltage continuously from start-time to end-time in each time

interval. Again, the 3 consecutive stages are adopted, but each

stage lasts for 250 seconds. Fig.18(a) shows the measured

mains voltage along the distributed line where the ESs and their

noncritical loads are connected. The corresponding reactive

power provided by the five ESs are included in Fig.18(b). When

ESs are turned off, the voltages are fluctuating from 90 V to 113

V. The upper and lower boundaries of the nominal voltage

(0.95-1.05 p.u.) are marked by two dotted lines in Fig.18(a). In

the first stage without using ESs, the mains voltages go below

the lower boundary. When droop control is used in the second

stage, the mains voltages are well regulated within the tolerance

range, but the sharing of the reactive power compensation

among the five ESs are not good. In the last stage when

consensus control is adopted, good mains voltage regulation

along the distribution line and sharing of reactive power

compensation among the ESs can be achieved simultaneously.

V. CONCLUSION

ES was originally proposed as a DSM method to provide

local voltage regulation of the distribution network using droop

control without the need for communication network. While

this feature is highly useful during abnormal conditions, this

paper shows that an extra layer of consensus control can offer

new functions not previously realized under the normal

conditions. This paper presents three novel contributions.

Firstly, it is the first practical design and implementation of

consensus control of a group of distributed ESs in a microgrid

environment for sharing both active and reactive power

compensation efforts in mitigating voltage and frequency

fluctuations. Secondly, the study investigates the coexistence of

droop control and consensus control for this group of ESs.

These two control methods are found to be complementary and

not exclusive. Thirdly, new plug-and-play features of ESs are

practically demonstrated for the first time.

A circular communication link in the cyber layer has been

used to implement the consensus control for practical

demonstration. The hardware tests confirm the practical use of

consensus control for a group of ESs in mains voltage and

system frequency regulation in a microgrid environment. The

plug-and-play and the distributed features of the ESs indicate

that a large group of ESs can in principle increase the robustness

of the microgrid because any failure of a small number of ESs

will not affect the overall regulations of the mains voltage and

system frequency. Practical results of two sets of comparative

studies on droop control and consensus control are included to

demonstrate their different features. These experimental results

show that droop control and consensus control are

complementary and not exclusive. Consensus control can be

used under normal conditions when the communication

network is available to provide good sharing of power

compensation efforts among the distributed ESs. Such ESs can

revert to droop control if the communication network fails

and/or become unavailable.

ACKNOWLEDGMENT

This project was supported by the Hong Kong Research

Grant Council under the Theme-based Research Project T23-

701/14-N.

VI. REFERENCE

[1] Felix F. Wu, Pravin P. Varaiya and S.Y. Hui, “Smart Grid with

Intelligent Periphery for the Future: Architecture for the Energy Internet”,

Engineering, Vol. 1, Issue 4, Dec. 2015, Pages 436-446

[2] S. Y. Hui, C. K. Lee, and F. F. Wu, "Electric springs—A new smart grid

technology," IEEE Transactions on Smart Grid, vol. 3, no. 3, pp. 1552-

1561, 2012.

[3] Bentham Paulos, “Too much of a good thing? An illustrated guide to

solar curtailment on California’s grid”, April 3, 2017, Greentech Media,

[online] https://www.greentechmedia.com/articles/read/an-illustrated-

guide-to-solar-curtailment-in-california#gs.oYrfb2U

[4] Qingsong Wang; Ming Cheng; Zhe Chen; Zheng Wang, “Steady-State

Analysis of Electric Springs With a Novel δ Control”, IEEE Transactions

on Power Electronics, 2015, Volume: 30, Issue: 12, Pages: 7159 - 7169,

[5] J. Soni and S. K. Panda, "Electric Spring for Voltage and Power Stability

and Power Factor Correction," in IEEE Transactions on Industry

Applications, vol. 53, no. 4, pp. 3871-3879, July-Aug. 2017.

[6] Qingsong Wang; Ming Cheng; Yunlei Jiang; Fujin Deng; Zhe Chen;

Giuseppe Buja, “Control of three-phase electric springs used in

microgrids under ideal and non-ideal conditions”, IECON 2016 - 42nd

Annual Conference of the IEEE Industrial Electronics Society, 2016,

Pages: 2247 – 2252.

[7] S. Yan, S. C. Tan, C. K. Lee, B. Chaudhuri and S. Y. R. Hui, "Electric

Springs for Reducing Power Imbalance in Three-Phase Power Systems,"

in IEEE Transactions on Power Electronics, vol. 30, no. 7, pp. 3601-

3609, July 2015.

[8] X. Luo, Z. Akhtar, C. K. Lee, B. Chaudhuri, S. C. Tan and S. Y. R. Hui,

"Distributed Voltage Control with Electric Springs: Comparison with

STATCOM," in IEEE Transactions on Smart Grid, vol. 6, no. 1, pp. 209-

219, Jan. 2015.

[9] L. Liang, Y. Hou, D. Hill and S. Y. R. Hui, "Enhancing Resilience of

Microgrids with Electric Springs," in IEEE Transactions on Smart Grid,

vol. PP, no. 99, pp. 1-1.

[10] T. Yang, K. T. Mok, S. C. Tan, C. K. Lee and S. Y. R. Hui, "Electric

Springs with Coordinated Battery Management for Reducing Voltage

and Frequency Fluctuations in Microgrids," in IEEE Transactions on

Smart Grid, vol. PP, no. 99, pp. 1-1.

[11] C. K. Lee, N. R. Chaudhuri, B. Chaudhuri, and S. R. Hui, "Droop control

of distributed electric springs for stabilizing future power grid," in IEEE

Transactions on Smart Grid, vol. 4, no. 3, pp. 1558-1566, 2013.

[12] R. Olfati-Saber, J. A. Fax and R. M. Murray, "Consensus and

Cooperation in Networked Multi-Agent Systems," in Proceedings of the

IEEE, vol. 95, no. 1, pp. 215-233, Jan. 2007.

[13] Q. Sun ; R. Han, H. Zhang, J. Zhou; Josep M. Guerrero, “A Multiagent-

Based Consensus Algorithm for Distributed Coordinated Control of

Distributed Generators in the Energy Internet”, IEEE Transactions on

Smart Grid, Vol: 6, Issue: 6, Nov. 2015, Page(s): 3 006 – 3019.

[14] H. R. Chamorro, M. Nazari, D. Babazadehi, N. R. Malik and M.

Ghandhari, "Consensus control for induction motors speed

regulation," 2014 16th European Conference on Power Electronics and

Applications, Lappeenranta, 2014, pp. 1-6.

TABLE III VOLTAGE FLUCTUATION PROFILE

[Start time, end time] (s) Voltage Change from the

start time value (V)

[0,50] +10

[50,90] +4 [90,130] -16

[130,170] +7

[170,210] -7 [210,250] +10

TPEL

12

[15] J. Khazaei and Z. Miao, "Consensus Control for Energy Storage

Systems," in IEEE Transactions on Smart Grid, vol. PP, no. 99, pp. 1-

1.doi: 10.1109/TSG.2016.2624144

[16] Johannes Schiffer, Thomas Seel, Jorg Raisch, Tevfik Sezi, “A

consensus-based distributed voltage control for reactive power sharing

in microgrids”, European Control Conference (ECC), 24-27 June 2014

[17] X. Chen, Y.H. Hou and Hui S.Y.R., “Distributed Control of Multiple

Electric Springs for Voltage Control in Microgrid”, in IEEE

Transactions on Smart Grid, Vol: 8, Issue: 3, May 2017, 1350 – 1359.

[18] Y. Zheng, D. Hill, K. Meng, and R. Hui, "Voltage Support for Critical

Buses with Consensus Control of Electric Springs in Distribution

Systems," IFAC-PapersOnLine, vol. 48, no. 30, pp. 173-178, 2015.

[19] Y. Zheng, D. J. Hill, K. Meng and S. Y. Hui, "Critical Bus Voltage

Support in Distribution Systems With Electric Springs and

Responsibility Sharing," in IEEE Transactions on Power Systems, vol.

32, no. 5, pp. 3584-3593, Sept. 2017.

[20] S. C. Tan, C. K. Lee and S. Y. Hui, "General Steady-State Analysis and

Control Principle of Electric Springs With Active and Reactive Power

Compensations," in IEEE Transactions on Power Electronics, vol. 28,

no. 8, pp. 3958-3969, Aug. 2013.

[21] C. K. Lee, B. Chaudhuri, and S. Y. Hui, "Hardware and control

implementation of electric springs for stabilizing future smart grid with

intermittent renewable energy sources," in IEEE Journal of Emerging

and Selected Topics in Power Electronics, vol. 1, no. 1, pp. 18-27, 2013.

[22] C. Zhao, U. Topcu, N. Li and S. Low, "Design and Stability of Load-

Side Primary Frequency Control in Power Systems," in IEEE

Transactions on Automatic Control, vol. 59, no. 5, pp. 1177-1189, May

2014.

[23] X. Luo and H. Guo, “Using electric springs to address over-voltage issue

of distribution networks in rural area in China”, 43rd Annual Conference of IEEE Industrial Electronics Society, IEEE IECON, 2017, pp: 471-476

Jie Chen (S’17) received the B.Eng. in electronic information and electrical engineering from the

Shanghai Jiao Tong University, Shanghai, China in

2015. He is currently a Ph.D. candidate in Power Electronics Research Group, Department of Electrical

and Electronic Engineering, The University of Hong

Kong, Hong Kong. His research interests include smart grid technologies, electric springs and

consensus control.

Yan Shuo (S’13-M’16) received his B. Eng. at The

University of South China in 2007, M. Eng. at Southeast University in 2010, and Ph. D. at The

University of Hong Kong in 2016. He is currently a

Postdoctoral Research Fellow in the Department of Electrical and Electronic Engineering at The

University of Hong Kong. His current research

interests include power electronic technologies, smart grid, renewable energy, and Microgrids.

Tianbo Yang (S’15) received the B.Eng. and M.Phil.

degrees in automation and control engineering from

the Harbin Institute of Technology, China, in 2012

and 2014, respectively. He is currently pursuing the Ph.D. degree at the Department of Electrical and

Electronic Engineering, the University of Hong Kong.

His research interest includes power electronic technologies in smart grid and energy storage system.

Siew-Chong Tan (M’06–SM’11) received the B.Eng. (Hons.) and M.Eng. degrees in electrical and

computer engineering from the National University of

Singapore, Singapore, in 2000 and 2002, respectively, and the Ph.D. degree in electronic and information

engineering from the Hong Kong Polytechnic

University, Hong Kong, in 2005. He is currently a Professor in Department of

Electrical and Electronic Engineering, The University

of Hong Kong, Hong Kong. Prof. Tan was a Visiting Scholar at Grainger Center for Electric Machinery and

Electromechanics, University of Illinois at Urbana-Champaign, Champaign,

from September to October 2009, and an Invited Academic Visitor of Huazhong University of Science and Technology, Wuhan, China, in December

2011. His research interests are focused in the areas of power electronics and

control, LED lightings, smart grids, and clean energy technologies. Prof. Tan serves as an Associate Editor of the IEEE Transactions on Power

Electronics. He is a coauthor of the book Sliding Mode Control of Switching

Power Converters: Techniques and Implementation (Boca Raton: CRC, 2011).

S. Y. (Ron) Hui (M’87-SM’94-F’03) received his

B.Sc. (Eng. Hons.) from the University of Birmingham in 1984 and the D.I.C. and Ph.D. degrees

from Imperial College London, London, U.K., in

1987. He has previously held academic positions at the

University of Nottingham (1987–1990), University of Technology Sydney (1991–1992), University of

Sydney (1992–1996), City University of Hong Kong

(1996–2011). Since July 2011, he has been holding the Chair Professorship at the University of Hong Kong. Since July 2010, he

has concurrently been holding the Chair Professor-ship at Imperial College

London. He has published more than 300 technical papers, including more than 180 refereed journal publications and book chapters. Over 50 of his patents have

been adopted by industry.

Dr. Hui is a Fellow of the IET. He has been an Associate Editor of the IEEE TRANSACTIONS ON POWER ELECTRONICS since 1997 and an Associate Editor

of the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS since 2007. He has

been appointed twice as an IEEE Distinguished Lecturer by the IEEE Power Electronics Society in 2004 and 2006. He served as one of the 18

Administrative Committee members of the IEEE Power Electronics Society

and was the Chair-man of its Constitution and Bylaws Committee from 2002-2010. He received the Excellent Teaching Award at CityU in 1998 and the Earth

Champion Award in 2008. He won an IEEE Best Paper Award from the IEEE

IAS Committee on Production and Applications of Light in 2002, and two IEEE Power Electronics Transactions Prize Paper Awards for his publications on

Wireless Charging Platform Technology in 2009 and on LED system theory in

2010. His inventions on wireless charging platform technology underpin key dimensions of Qi, the world’s first wireless power standard, with freedom of

positioning and localized charging features for wireless charging of consumer

electronics. In November 2010, he received the IEEE Rudolf Chope R&D Award from the IEEE Industrial Electronics Society, the IET Achievement

Medal (The Crompton Medal), and was elected to the Fellowship of the

Australian Academy of Technological Sciences and Engineering.