See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/340183797

A DER Management System for Distributing Power among Discretely and

Continuously Controlled Assets within DER Group in Microgrid

Preprint · September 2019

CITATIONS READS

0 293

4 authors, including:

Chu Sun Syed Qaseem Ali

McGill University Quanta Technology
21 PUBLICATIONS   168 CITATIONS    38 PUBLICATIONS   347 CITATIONS   

SEE PROFILE SEE PROFILE

Geza Joos
McGill University
413 PUBLICATIONS   14,733 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

InnovEE project View project

Smart Inverter Modeling Toolbox View project

All content following this page was uploaded by Chu Sun on 26 March 2020.

The user has requested enhancement of the downloaded file.

 CIGRE-195 2019 CIGRE Canada Conference
Montréal, Québec, September 16-19, 2019

A DER Management System for Distributing Power among Discretely and

Continuously Controlled Devices within DER group

Chu Sun1, Syed Qaseem Ali2, Geza Joos1, Francois Bouffard1

1McGill University, 2OPAL-RT Technologies

Montreal, Canada

SUMMARY
The integration of distributed energy resource (DER) poses new control challenges to distribution
system which are originally comprised mainly of passive loads. To ensure harmonious operation, a
software application capable of monitoring and controlling the distribution system in conjunction with
the grid operator, must be developed to provide DERs with appropriate set points predetermined by a
sophisticated control algorithm subject to grid requests and constraints.
Given DERs with diverse characteristics dispersed widely in distribution system, managing DERs in a
collective and hierarchical way becomes increasingly important to address the large quantity and high
complexity of DERs. Among the high numbers of DERs, there may be several devices of the same type
or termed as DER group in IEC 61968-5, which can be generally categorized into continuously (termed
as C-devices) and discretely controlled (termed as D-devices). The continuously controlled DERs can
be adjusted continuously in a specific range, while the discretely controlled DERs can only be turned
on or off. For this, the control system should not only dispatch the DER groups on higher-level, in light
of the overall objectives such as cost minimization, high resilience and reliability, fuel consumption
reduction, but also should ensure reasonable power distribution among C-devices and D-devices within
the DER group by a lower-level control.
To meet this requirement, a DER management system capable of aggregating DERs as groups and then
distributing power within DER group is proposed, which can serve as supplementary measures of utility
distribution management system (DMS) or microgrid controller. For active power dispatch, a weighted
and queuing (W&Q) method is employed to manage power distribution within DER group. For reactive
power distribution, an iterative backtracking method is adopted for switching ON or OFF D-devices
while weighted method is used for distribution among C-devices. Theoretical basis which interprets the
power distribution as Knapsack problem from the perspective of computer engineering is also
established. The proposed control strategy paves ways for developing advanced distribution
management system with hierarchical architecture, by firstly generating dispatch command on
aggregated DER group level and then distributing power among specific devices on DER device level.

KEYWORDS
DER Management System, Distributed Energy Resource Group, Discretely and Continuously
Controlled Devices, Weighted and Queuing Approach, Knapsack Problem, Greedy Algorithm,
Iterative Backtracking Algorithm

[email protected]
1. MICROGRID AND DER GROUP COMPOSITION
With common functions specificified for inverter-interfaced DERs [1], standardization of grid code,
device communication interface and test procedure of DERs [2], there is a growing need of transferring
the numerous individual devices with complex settings and DER-type-specific interactions into DER
groups with harmonious control actions [3]. One path to reach this objective is DER management system
which can aggregate numerous DERs into several groups, thus simplifying the overall control actions.
As per the imminent guide IEEE Std. 2030.11 entitled “Distributed Energy Resources Management
Systems (DERMS) Functional Specification” and common function specifications for DER group [4],
the basic control functions of DER group include real power and reactive power dispatch. Integrating
DER group management with higher-level controller such as microgrid controller or DMS will make
dispatch of distribution system with numerous DERs compact, modular and scalable.

One architecture of deployment of the DER management system is sketched in Fig. 1 where multiple
DER groups are controlled. On the DER group level, the characteristics and interactions of each devices
can be ignored but focus only needs to be given to the aggregated group behavior. Taking active power
dispatch as an example, a general DER composition is shown in Fig. 1 where controllable devices are
categorized into four types: dispatchable generators such as diesel generator and biomass, controllable
loads such as continuous demand response (if allowed by contract), discretely sheddable load or load
with hybrid control characteristics, energy storage system (ESS), and non-dispatchable generators (wind
and photovoltaic (PV)). The conventional control system designed for individual DER dispatch can still
be used for power dispatch among DER groups based on certain objectives [5].

On the device level, DER management system capable of aggregation and dissemination, is responsible
for power distribution between C-devices and D-devices with general representation given in Fig 2. This
is realized by a few control algorithms designed to distribute the total power command given by group
level control. In this paper, control algorithms for both active and reactive power distribution between
C-devices and D-devices are investigated and tested with controller hardware-in-the-loop approach.

Dispatchable Generator Group #1

Group
management Dispatchable Generator
Controlable Load
Group
Adjust Controlable Load Group #2
management
PCC Controlable Load

Shed
Critical Load
DER Energy Storage
Charge/discharge
Management
Curtail
System Renewable Generator
Group
management Renewable Generator Group #3
Renewable Generator

Fig. 1. DER management system for multiple DER groups

ON/OFF

Synchronous generator PD,j Discretely curtailable renewables

Type III&IV WTG,PV PC,i Sheddable load
(QD,j)
Continuous adjustable load (QC,i) TSC
STATCOM, TCR

(a) (b)
Fig. 2. General illustration of devices within DER group. (a) Continuous controlled. (b) Discretely controlled.

2
2. ACTIVE POWER DISTRIBUTION
2.1. Problem formulation

The power distribution within DER group should first meet power balance constraint. It is assumed that
Pplan is the total planned power command for the considered DER group, which is specified on group
level by optimization or rule (priority)-based control [5]. Let PCi,f and PDj,f be the forecasted power of
the ith C-device and the jth D-device, PC,i and PD,j be the practical power of the ith C-device and the jth D-
device after control, Pci,min and Pci,max be the minimum and maximum power of the ith C-device, the power
balance constraint can be formulated by (1) where variable xi is the adjustment percentage of the ith C-
device, yj is the ON/OFF command of the jth D-device with yj =0 denoting ON and yj =0 denoting OFF.

 P +  j =1 PD , j − Pplan =0
m n
i =1 C , i

Pc ,i = (1 − xi ) PCi , f , PD , j = y j PDj , f
(1)
xi  [ PCi ,min / PCi , f ,1](for decrease case) or [1, PCi ,max / PCi , f ](for increase case)
y j  0,1 , i = 1, 2, , m; j = 1, 2, ,n
The power distribution problem is to find xi and yj. Given (1) is an indefinite equation, other constraints
and objectives should be included to narrow down the solution set. In reference to the principle of power
curtailment within wind farm [6] and computer engineering algorithms to solve dispatch problem [7],
an algorithm combining weighted and queuing (W&Q) method for power distribution within DER group
is proposed to find power command PCi and PDi. Notice that other algorithms such as dynamic
programming or branch and bound method, can also be adopted for power distribution among D-devices,
but higher computational capability is required [7].

2.2. Weighted method

For C-devices within DER group, weighted averaging based on certain criteria is employed for
continuous power adjustment, as given in (2)-(6) where PC,f = ∑m n
i=1 PCi,f and PD,f = ∑j=1 PDj,f are the
total forecasted active power of C-devices and D-devices. Notice the value of xi in (2) may exceed the
ranges defined in (1). In that case, xi will take the value at the terminal of the interval while the remaining
devices will balance the power mismatch as per the same weighted criterion. In (5), the power margin
Pmarg,i which represents the adjustable power range, is defined as PCi,min-PCi,f or PCi,max-PCi,f, depending
on power adjustment direction. A combined consideration of cost and power margin is given by (6)
where i is the incremental cost near the operating point PCi,f.

(a) Uniform (or even) distribution: xi = (PC , f − Pplan ) / (m * Pci , f ) (2)

(b) Nameplate capacity (Pname,i) weighted averaging: xi = (PC , f − Pplan )Pname ,i / ( i =1 Pname ,i PCi , f )
m
(3)
(c) Forecasted power (PCi,f) weighted averaging: xi = (PC , f − Pplan ) /  i =1 PCi , f
m
(4)
(d) Capacity-margin (Pmarg,i)weighted averaging: xi = (PC , f − Pplan )Pm arg,i / ( PCi , f  i =1 Pm arg,i )
m
(5)
(e) Power-cost ratio weighted averaging: xi = (PC , f − Pplan )(Pm arg,i / i ) / ( PCi , f  i =1 Pm arg,i / i )
m
(6)

2.3. Queuing method

For D-devices within DER group, a queuing method is adopted. Combining with weighted method,
complete algorithm of the W&Q method to set power command PCi and PDi of a hybrid DER group with
C-devices and D-devices is developed. It is assumed that C-devices are prioritized over D-devices given
the lower cost than D-devices (λi < λj ). Objective of the algorithm is to minimize the total number of
ON/OFF actions of D-devices (∑nj=1 yj ) while power curtailment of C-devices are weighted with their
forecasted power (PCi,f). Procedure of the W&Q algorithm for power curtailment is described below.

Step (1): If PD,f+∑m

i=1 PCi,min≤Pplan<PD,f +PC,f, set PC,i = max{PCi,min, PCi,f /PC,f (Pplan-PD,f)}. The distribution
ends and no ON/OFF actions of D-devices are needed.
3
Step (2): If the condition PD,f+∑m
i=1 PCi,min≤Pplan in (1) does not hold, the strategy is shown as below, by
which the D-devices will be tripped off from large to small in a queuing manner to solve the
problem (7) which is essentially a combinatorial optimization problem [8].

n
min j =1
yj

 PD , j  Pplan −  i =1 PCi ,min

n m
s.t. j =1
(7)
PD , j = y j PDj , f , y j  0,1 , j = 1, 2, ,n
2.1) Find and trip off the hth D-device that satisfies PDh,f=min{PDr,f |PDr,f>(-Pplan+PD,f + ∑m i=1 PCi,min)}
and recalculate PD,f, then go to 2.3).
2.2) If no such D-device exists, find the kth D-device that satisfies PDk,f =max{PDs,f |PDs,f <(-Pplan+PD,f
+∑m m
i=1 PCi,min )}. Trip it off and recalculate PD,f. If Pplan -PD,f - ∑i=1 PCi,min >0, return to 2.1);
otherwise stop tripping D-devices and go to 2.3).
2.3) The total remaining power of C-devices is calculated to be Pplan-PD,f where PD, f is the power
of D-devices after tripping in 2.1) and 2.2). After 2.3), go to Step (3).
Step (3): For C-devices, the power in the next interval is PC,i=max{PCi,min, PCi,f/PC,f (Pplan-PD,f)} based
on Pci,f weighted averaging. After that, the algorithm of W&Q method ends.

The above procedure to solve (7) in W&Q algorithm is quite similar to the greedy algorithm to solve 0-
1 Knapsack problem [8]. In the above process, it is assumed that all devices have the same incremental
cost which may be different in practical situations. In that case, the power distribution problem will be
reformulated as minimizing {∑m n
i=1 λi x𝑖 PCi,f + ∑j=1 λj yj PDi,f } with suboptimal problem changed to
minimizing ∑nj=1 λj y PDi,f . A greedy algorithm with j PDi,f queuing from small to large can be developed.
j

3. REACTIVE POWER DISTRIBUTION

For reactive power management, DER groups can be separated based on zone-division approach [9],
which may be unnecessary for networks covering a small region only. The reactive power distribution
within DER group can also be categorized into 0-1 Knapsack problem. However, the priority will be
given to D-devices such as thyristor-controlled capacitor (TSC), for their lower incremental cost relative
to C-devices such as inverter-interfaced generators. The objective here will be minimizing the use of C-
devices for reactive power regulation while avoiding over-compensation by D-devices.

Replacing the symbol “P” with “Q”, removing “f ” (the capacity of D-devices are known) and using the
same label (i, j, m, n), the power distribution for D-devices can be formulated as (8), which can be
interpreted as Container Loading problem and related algorithms can be leveraged [7]. With reactive
power compensation for inductive load as an example, iterative backtracking algorithm based on Depth-
First Search (DFS) principle is adopted here to determine the ON/OFF action of D-devices, while power
of C-devices is weighted by their reactive power capability.

n
max j =1
QD , j z j


n
s.t. j =1
QD , j z j  Qplan , (8)
z j  0,1 , j = 1, 2, ,n
Step (1): If min{QD,j}≤Qplan holds, the power distribution is determined by the following two steps.
1.1) Iterative backtracking algorithm will be implemented to obtain the ON/OFF commands (zj) of
D-devices. The details of the algorithm can be found in [8] and are not given here for brevity.
Calculate the total reactive power of D-devices QD=∑nj=1 QD,jzj .
1.2) Set the total power command for C-devices as QC =min {∑m i=1 QCi,capmax, Qplan-QD} where QD
is calculated from (1). QCi,capmax is the maximum adjustable capacitive power of C-devices,
which can be determined from power capacity of C-devices. For STATCOM, QCi,capmax is
equal to the inverter capacity. For renewable generators, QCi,capmax depends on the output
active power and constraints of power factor and reactive power [2]. The reactive power

4
 command of C-devices QC,i is given by (9) where power margin weighting is considered.
QC ,i = QC QCi ,cap max /  i =1 QCi , cap max
m
(9)
Step (2): If the condition in (1) does not hold, two scenarios are considered.
2.1) If Qplan≤ ∑mi=1 QCi,capmax, all D-devices will be OFF. QC,i is calculated from 1.2).
2.2) If Qplan> ∑mi=1 QCi,capmax, two sub-scenarios will be further considered.
2.2.1) If min{QD,j}-Qplan< ∑m i=1 QCi,indmax, the D-device with reactive power equal to min{QD,j}
will be ON and QC is set as min{QD,j} -Qplan. After that, QC,i will be calculated by (9)
with QCi,capmax replaced by QCi,indmax which represents the maximum adjustable inductive
power of C-devices. Notice that reactive power is over-compensated in this case.
2.2.2) If condition in 2.2.1) not rue, power is determined by the following steps.
2.2.2.1) If min{QD,j}-Qplan-∑m m
i=1 QCi,indmax>Qplan-∑i=1 QCi,capmax, go to 2.1).
2.2.2.2) If the condition in 2.2.2.1) does not hold, the D-device with reactive power equal
to min{QD,j} will be ON, QC is set as ∑m i=1 QCi,indmax and power will be distributed
as per 2.2.1).

4. IMPLEMENTATION AND CASE STUDIES

4.1. CHIL testing platform

The DER management system for power distribution within DER group is tested with controller-
hardware-in-the-loop (CHIL) approach. Simplified model of power network is run on real-time
simulator (RTS) while control algorithm is implemented on microcomputer Raspberry Pi. Ethernet
connection based on UDP/IP protocol is used for communication between RTS and Raspberry Pi.

Host PC

TCP/IP
Network

Raspberry Pi
Real time simulator

Internet network socket Internet network socket

Control commands for dispatch

Signal measured from DER groups

Fig. 3. CHIL test platform of DER management system

4.2. Active power management

To verify the effectiveness of W&Q method for active power management, power shedding for
continuously controlled load (C-devices) and discretely controlled load (D-devices) is taken as an
example. A simplified DER group model with two C-devices and three D-devices depicted in Fig. 4 is
adopted for the test. For convenience, the original power outputs of the devices are assumed constant,
as listed in Table. 1. The total shedding command Pplan increases incrementally from 0.1 to 1.5 p.u. The
simulation results in Fig. 5 show that the practical power can closely track the desired profile. The
continuously controlled loads (C#1-2) are shed in proportion to their original power, assuming the cost
of C-devices are identical in the case study. The discretely controlled loads (D#1-3) are shed from small
to large power in a queuing manner which meets power balance while avoiding unordered actions.

5
 PCC Simplified model of DER group
Pplan

PC1,f PC,1 PD,1 PD,2 PD,3

PC2,f DER PC,2 PC,1 PC,2
PD1,f management PD,1
PD2,f system PD,2
PD3,f PD,3 C#1 C#2 D#1 D#2 D#3

Fig. 4. Simplified model of DER group to test the DER management system

Table 1. Original forecasted load power

Devices D-devices C- devices
Label D#1 D#2 D#3 C#1 C#2
Forecasted Active Power 0.3 0.4 0.5 0.6 0.7
(PCi,f)/p.u.
Constraints ON/OFF PCi,f≥Pc,i≥max{PCi,f/2}
ON/OFF
ON/OFF

Fig. 5. Simulation results of DER management system for active power load shedding

4.3. Reactive power management

A similar model as in Fig. 4 with active power replaced by reactive power is adopted. The
interconnecting impedance between PCC and the slack bus is 0.017+j0.1 p.u. The power values are
listed in Table. 2. The total dispatch command to compensate for inductive load increases from 0.1 to
1.5 p.u. The D-devices (D#1-3) can be three TSCs, the first C-device (C#1) can be a PV generator which

6
has 0.44 p.u. reactive power capability while the second C-device (C#2) can be a STATCOM [2]. The
practical reactive power closely matches the inductive load profile, as shown in Fig. 6. The priority of
reactive power balance is given to D-devices which are added from small to large power based on
iterative backtracking algorithm, while C-devices output power in proportion to their reactive power
capability. In contrast to active power dispatch, the number of ON/OFF actions of D-devices is larger.
The voltage profile improvement in Fig. 7 demonstrates the effectiveness of reactive power
compensation.

Table 2. Parameters of reactive power devices

Assets D-assets C-assets
Label D#1 D#2 D#3 C#1 C#2
Reactive Power/p.u. 0.3 0.4 0.5 Qc,1 Qc,2
Constraints ON/OFF Qc,1∈[-0.44,0.44] Qc,2∈[-1,1]
ON/OFF
ON/OFF

Fig. 6. Simulation results of DER management system for reactive power distribution

Fig. 7. Profiles of voltage amplitude at PCC with and without reactive power compensation

7
 5. CONCLUDING REMARKS
This paper presented algorithms of DER management system to distribute power within aggregated
DER group. For active power distribution, the proposed method sets weighting criteria for continuously
controlled devices so that their power is shared based on power capacity, cost or other priority criteria,
while the discretely controlled devices are controlled to reduce the total number of ON/OFF actions or
the associated cost. For reactive power distribution, the discretely controlled devices are switched
ON/OFF based on iterative backtracking algorithm while the remaining power is distributed among
continuously controlled devices with weighted method. The proposed control algorithms provide a
simple way to orchestrate devices within DER group including load shedding, renewable curtailment,
diesel generator adjustment and reactive power management, which will be promising to embrace a
future of digital grid with millions of DERs.

BIBLIOGRAPHY

[1] "Common Functions for Smart Inverters, Version 4," Electric Power Research Institute, Palo Alto,
California, USA, December 2016.
[2] IEEE Standard for Interconnection and Interoperability of Distributed Energy Resources with
Associated Electric Power Systems Interfaces," in IEEE Std 1547-2018 (Revision of IEEE Std
1547-2003), pp.1-138, 6 April 2018.
[3] Walt Johnson, A Standardized Approach to DER Group Management - A Look at IEC 61968-5,
OpenADR Alliance Member Meeting, April 24, 2018,
https://www.openadr.org/devices/Group%20Management%20with%2061968-5.pdf.
[4] "Common Functions for DER Group Management, Version 3," Electric Power Research Institute,
Palo Alto, California, USA, November 2016.
[5] C. Sun, J. N. Paquin, F. Al Jajeh, G. Joos and F. Bouffard, "Implementation and CHIL Testing of
a Microgrid Control System," 2018 IEEE Energy Conversion Congress and Exposition (ECCE),
Portland, OR, 2018, pp. 2073-2080.
[6] L. Xiang, Y. Xie, B. Zhang, M. Ding, Z. Fu and Y. Tang, "Dynamic grouping strategy for active
power control of wind farm," 2013 IEEE International Conference on Cyber Technology in
Automation, Control and Intelligent Systems, Nanjing, 2013, pp. 132-137.
[7] Sartaj Sahni. 2004. Data Structures, Algorithms, and Applications in C++ (2nd ed.). Silicon Press,
Summit, NJ, USA.
[8] S. Q. Ali, S. D. Maqbool, T. P. I. Ahamed and N. H. Malik, "Load scheduling with maximum
demand and time of use pricing for microgrids," 2013 IEEE Global Humanitarian Technology
Conference: South Asia Satellite (GHTC-SAS), Trivandrum, 2013, pp. 234-238.
[9] H. Sun, Q. Guo, B. Zhang, W. Wu and B. Wang, "An Adaptive Zone-Division-Based Automatic
Voltage Control System with Applications in China," in IEEE Transactions on Power Systems,
vol. 28, no. 2, pp. 1816-1828, May 2013.

View publication stats