Fast and Reliable Load-Shedding Scheme for

Wastewater Treatment Plant – A Case Study

G. M. Asim Akhtar, Muhammad Waqar Ahmed, Will Allen, and Sujay Dasgupta
Schweitzer Engineering Laboratories, Inc.

Perry Zhang
The City of Calgary

Kyle Jensen
Stantec Inc.

© 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained
for all other uses, in any current or future media, including reprinting/republishing this material
for advertising or promotional purposes, creating new collective works, for resale or redistribution
to servers or lists, or reuse of any copyrighted component of this work in other works.

This paper was presented at the 2022 IEEE Electrical Power and Energy Conference and can be
accessed at: https://doi.org/10.1109/EPEC56903.2022.10000080.

For the complete history of this paper, refer to the next page.
 Revised edition released February 2023

Originally presented at the

2022 IEEE Electrical Power and Energy Conference, December 2022
 Fast and Reliable Load-Shedding Scheme for
Wastewater Treatment Plant – A Case Study
G. M. Asim Akhtar Muhammad Waqar Ahmed Will Allen Perry Zhang
Schweitzer Engineering Schweitzer Engineering Schweitzer Engineering The City of Calgary
Laboratories, Inc. Laboratories, Inc. Laboratories, Inc. Calgary, Canada
Calgary, Canada Toronto, Canada Calgary, Canada [email protected]
[email protected] muhammad_waquar_ahmed [email protected]
@selinc.com

Kyle Jensen Sujay Dasgupta

Stantec Inc. Schweitzer Engineering
Calgary, Canada Laboratories, Inc.
[email protected] Pullman, U.S.A.
[email protected]

Abstract—Innovations in the fields of automation and

A PMCS is ideal for industries with
networking have helped traditional power system substations
evolve. Intelligent electronic devices (IEDs) accompanied by onsite generation and/or that are grid
optimized and smartly engineered communications networks have connected. It contains automated control
provided engineers with opportunities to better design and functions specifically designed to
implement various algorithms. Therefore, in the event of a prevent, detect, and mitigate system
disturbance or fault, the power system stability and process blackouts in either grid-connected or
survivability are maintained. islanded mode. Automated functions
within a PMCS control major power
Power systems are proven to have more stable operation while system assets for optimal economic
connected to a utility; however, the challenge arises when the
operation. By properly collecting,
power system is islanded and suffers from a loss or an excess of
generation. In an islanded configuration, fast and selective processing, and presenting power system
shedding of loads and/or generators based on system topology is data as usable information, the PMCS
critical in responding to system disturbances to avoid blackouts system enables operators, maintenance
and ensure minimum process downtime. personnel, and engineering staff to
diagnose system events, predict
This paper presents a real-world implemented load-shedding equipment failures, and minimize
scheme (LSS) for a North American wastewater treatment plant. unnecessary maintenance [1].
The LSS was deployed in two tiers of primary and secondary
controls via redundant substation-hardened controllers. The This paper presents a case study for one of the largest
primary shedding system is based on calculation of a predictive wastewater treatment plants in North America. The plant under
power deficit or surplus for various predetermined contingency consideration was constructed in 1932 and has had various
events. The primary system issues shedding decisions upon updates during its life span. In 2017, the plant went through a
contingency detection, whereas the secondary shedding system is major electrical infrastructure upgrade with the installation of a
based on triggers asserted by underfrequency and/or
PMCS to ensure robust, reliable, and operator-friendly control
overfrequency protective relays.
and monitoring of the power system in the facility. As part of
The paper also provides an overview of the implemented the complete PMCS suite, one important feature was to
network scheme; however, a detailed discussion regarding implement a fast, reliable, and robust load-shedding system that
engineering and performance will be included in the authors’ would guarantee power system stability and process
future work. survivability. In addition, it was essential to recognize that any
misoperation could lead to the release of toxic waste or untreated
Keywords—fast load shedding, contingency, underfrequency, water into the downstream river, posing a serious health and
software-defined networking, power management and control safety hazard for the public.
system, frequency response, wastewater treatment plant
The paper describes the load-shedding scheme (LSS)
I. INTRODUCTION implemented to meet the project requirements using the field
proven engineering algorithms, design techniques, and testing
A power monitoring and control system (PMCS) is defined methodologies. Section II examines the simplified power
in [1] as: system of the wastewater treatment facility. Section III provides
an overview of the implemented system architecture using primary sources of high-voltage distribution to the various areas
software-defined networking (SDN). Section IV presents the of the plant. Major process loads are fed at 600 V, except for the
details of the two types of high-speed load-shedding algorithms aeration blowers in the plant, which are fed at 4.16 kV. The
and their implementation using the defined system architecture. 13.2 kV, 4.16 kV, and 600 V switchgears throughout the plant
Section V explores implementation details for intelligent and are configured as double-ended main-tie-main systems. Fig. 1
selective load shedding. Section VI evaluates testing scenarios provides a simplified version of the facility power system as an
and associated results. overview for readers; however, to maintain the end user’s
system confidentiality, the asset tags and details have not been
II. OVERVIEW OF FACILITY POWER SYSTEM disclosed.
The power system for the wastewater treatment plant is an
industrial microgrid with multiple power sources, including III. SIMPLIFIED SYSTEM ARCHITECTURE BASED UPON
onsite power generation and numerous high- and low-voltage SDN TECHNOLOGY
buses distributed throughout the facility. Modern numerical relays and metering IEDs not only protect
and monitor the plant power system, but also form the
The plant is interconnected to the electrical utility via two
foundation of a load-shedding system. The IEDs are networked
13.2 kV primary voltage service feeders. In addition to the utility
over Ethernet communications using two independent networks
feeders, there is 8.8 MW of onsite cogeneration, which includes
in a ring configuration utilizing parallel redundancy protocol
three 4.16 kV/1.6 MW reciprocating engines and one
(PRP) based on SDN technology.
4.16 kV/4 MW gas-fired combustion turbine generator (GEN).
The 13.2 kV utility feed is transformed down to 4.16 kV at
two different switchgears. These two switchgears are the

Fig. 1. Simplified Single-Line Diagram of the Power System for the Wastewater Treatment Plant
 Data for load-shedding purposes are gathered by local relays, For the wastewater treatment plant under discussion, two
installed on various equipment across the power system. These types of FLS schemes have been designed and deployed. The
data are then sent via International Electrotechnical Commission primary FLS scheme is called contingency-based load shedding
(IEC) 61850 Manufacturer Messaging Specification (MMS) (CLS); whereas, the secondary FLS scheme is called
and Generic Object-Oriented Substation Event (GOOSE) underfrequency-based load shedding (UFLS). The details of
communications protocols over the Ethernet network to data both types of load-shed schemes are explained in the following
concentrators (DCONs) that operate on redundant hardware as subsections.
A and B devices. DCON-A and DCON-B are installed in two
physically distant locations. These DCONs then distribute data A. Contingency-Based Load-Shedding Scheme
to centralized controllers, which operate on redundant hardware “A CLS algorithm sheds load to maintain the power system
installed in the same locations as DCON-A and DCON-B, balance by reducing the total plant electrical load to less than
respectively. Data are sent to the fast load-shedding (FLS) the calculated available turbine and generator capacity after a
controllers by DCONs via User Datagram Protocol (UDP) over contingency occurs” [1]. A contingency is defined as the
the Ethernet network. Each of the controllers performs its opening of a breaker that interrupts system power flow.
specific functions and outputs control signals. These control Contingency triggers are communicated from protective relays
outputs are distributed through the DCONs via UDP to the DCONs using GOOSE messages.
communications protocol. From the DCONs, the signals are sent When a contingency breaker opens, the CLS controller
via IEC 61850 GOOSE signals over the Ethernet network to triggers load shed “based on the contingency status and
relays and trip units. From the relays, hardwired output contacts metering, user-settable load-shedding priorities, user-settable
transmit controls directly to the desired control point. incremental reserve margin (IRM) values, topology status, load
Fig. 2 presents a simplified version of the system status, and metering. The CLS controller sends the load trip
architecture used for implementation of an FLS system. signals to the respective” intelligent electronic devices (IEDs,)
via DCONs, using GOOSE messages, “the output contacts of
IV. HIGH-SPEED LOAD SHEDDING – FUNCTIONAL OVERVIEW which are wired to trip coils of the breakers” [1] Fig. 3 shows
“High-speed [load-]shedding control functions are required the overall CLS algorithm.
to mitigate power system unbalance and prevent blackout[s]” The algorithm runs on real-time automation controllers, in
[1]. The main purpose of these control functions is to maintain advance of the event trigger taking place; therefore, the power
balance between power generation and demand and/or load by deficit can be seen by the operator before any event occurs. If
intelligently selecting and tripping load breakers based on a there is not sufficient plant load to balance the loss of power
triggered contingency that may include a generator, utility tie from a contingency source, an alarm will convey this to the
line, and bus tie. operator, which allows the operator to take corrective action for
avoiding an operation state that is vulnerable to a blackout.

Fig. 2. Simplified System Architecture Based on SDN Technology

Fig. 3. Block Diagram for Contingency Load-Shed Algorithm

For the operator to understand the expected loads to be shed, Fig. 4. Crosspoint Switch Matrix
based on the prospective contingency trigger, a crosspoint
switch is populated and displayed to the operator by means of a This scheme backs up the primary CLS scheme by detecting
human-machine interface (HMI). Fig. 4 shows a simplified frequency decay that was not prevented by the CLS, which was
version of the crosspoint logic. A crosspoint switch matrix is due to an alarmed breaker opening, overestimated generator
essentially a table constructed as contingency versus load. IRM, or a load-shedding failure due to wiring and/or trip coil
issues.
Each contingency has an associated trigger. Trigger 1
corresponds with Contingency 1, and so forth. Each load is An added benefit of this centralized underfrequency load-
selected based on contingency and may be preselected for shedding scheme is the ability to isolate events on an island-by-
multiple contingencies. However, if a load was shed in one island basis. This ensures that only relevant loads are shed for
contingency, it will be inhibited from being selected for frequency excursions throughout the island and that other
shedding by any other contingencies. The result of the islands remain unaffected. This is in contrast to relay-based
crosspoint switch multiplication is trip signals that are sent underfrequency load-shed schemes in which feeders are tripped
directly to the DCONs at various locations. From there, they are as soon as the UF trigger asserts, without any consideration of
sent to the field IEDs with output contacts wired to the trip coil the required amount of load to be shed. Fig. 5 explains the UFLS
of the associated load. algorithm.

B. Underfrequency-Based Load-Shedding Scheme

According to [1], a UFLS scheme is an FLS:

algorithm that maintains the power

system balance by reducing the total
plant load by fixed amounts of load
power at [four] separate underfrequency
(UF) levels. The UF level detection
occurs in [protective] relays located at
each bus/generator.

When the [relay] detects a UF event, it

sends a high-speed signal to the UFLS
system. The UFLS controller determines
the load to shed based on the UF trip
Fig. 5. Block Diagram for Contingency Load-Shed Algorithm
level, user-settable load-shedding
priorities, topology status, load status,
and metering. The UFLS system sends
the load trip signals to [the] IED[s. The]
output contacts of [the IEDs] are wired
to trip coils of the breakers [1].
V. HIGH-SPEED LOAD SHEDDING – IMPLEMENTATION DETAILS
A. Selected Contingencies for Primary CLS Scheme
For the power system shown in Fig. 1, 27 breakers were
identified that may initiate primary load-shedding
contingencies. The contingencies are split into three different
classifications: generator, tie line (loss of utility source), and bus
coupler (loss of a link between system buses). Table I shows the
contingencies identified for the specific case study.
B. Selected Contingencies for Secondary UFLS Scheme
The secondary LSS is based on the frequency degradation of
the system and the corresponding power decrease for which it is
stabilized. For the power system shown in Fig. 1, the
underfrequency system operates on the 4.16 kV generator buses
and will only operate when islanded from the utility. The
underfrequency system will shed load based on a power setting
entered for each level on the HMI. Underfrequency triggers are
generated via relays that perform bus frequency measurements
at generator buses. In addition, the relays installed in the 13.2 kV
distribution buses are also programmed to provide
underfrequency triggers to controllers for the identified
underfrequency contingencies.
Table II lists the underfrequency contingencies identified for
the specific case study.
C. Underfrequency Coordination
Fig. 6 shows a simplified visual representation of the
frequency coordination for the UFLS scheme with the
underfrequency protection settings of the two types of Fig. 6. Coordination Between UFLS Levels and Generator Frequency
generators installed in the power system shown in Fig. 1. Protection Settings

The underfrequency load-shedding set points are Coordination of the UFLS and CLS should be performed so
coordinated with the utility decoupling requirements, such that that when both algorithms are enabled, it would not cause any
the plant can ride through voltage and frequency excursions if unnecessary load tripping. The CLS action takes place first;
there are disturbances on the utility network. To overcome this, whereas, the UFLS waits for a trigger from the field. Because of
the controller topology tracking informs the operator about the the vast differences in the inertia and rating of machines, it was
connected state of the utility in the plant power system. If the critical that the load-shedding system be designed to cover the
plant is utility connected, then the controller dynamically maximum range of possible operation scenarios. Based on the
disables Level 1 and Level 2 of the underfrequency scheme. aforementioned facts, the following conditions were established
to achieve coordination between the CLS and UFLS:
TABLE I. PRIMARY CONTINGENCIES
• Block Underfrequency Levels 1 and 2 when the CLS is
Contingency Number Type
enabled. Levels 1 and 2 were designed as a backup for
1–5 Generator the CLS and will only come into effect when the primary
6–7 Utility tie line load-shedding scheme is disabled.
8–27 Bus tie
• Keep Underfrequency Levels 3 and 4 active all the time,
even with the CLS enabled. Levels 3 and 4 were
TABLE II. UNDERFREQUENCY CONTINGENCIES designed to serve as a last resort for saving the system
Contingency Number Bus Underfrequency Level from a blackout.
1 Generator Bus-A Level 1
D. Sheddable Loads
2 Generator Bus-A Level 2
Based on recommendations from the customer’s operations
3 Generator Bus-A Level 3 department, 66 loads were identified as sheddable loads for the
4 Generator Bus-A Level 4 power system shown in Fig. 1. These 66 loads are situated
5 Generator Bus-B Level 1 throughout the entire power system at voltage levels of 4.16 kV
and 0.6 kV.
6 Generator Bus-B Level 2
7 Generator Bus-B Level 3 E. Load Selectivity Based on Load Group Priorities
8 Generator Bus-B Level 4 All of the sheddable loads are organized into ten groups of
loads. Each load has a group associated with it, which can be set
by an operator through the HMI. The CLS and/or UFLS VI. TESTING SCENARIOS
algorithm will try to select the optimal amount of load to satisfy
each contingency. Each load within a group will have equal A. Test-1: Islanded From Utility With Three Generators
priority to shed. The group assigned to each load does not need Running and One Type-2 Generator Tripping
to be unique. The power system is islanded from the utility and two
Type-1 generators each running at 1.15 MW and one Type-2
All load-shedding actions use an optimal load selection generator running at 3.6 MW are kept online. The UFLS is
algorithm, which tries to select minimum number of loads to enabled, the load on the island is approximately 5.9 MW, and
satisfy a contingency. Loads are selected for shedding based on the Type-2 generator is tripped at t = 1 second. The frequency
their present power and their predefined group. Loads with a response of the island for the UFLS is represented by the solid
power value of zero, negative power value, or group value of blue line in the plot shown in Fig. 8. The observations are as
zero will be inhibited and not selected for shedding. Operators follows:
may set load groups to zero to intentionally inhibit them from
shedding. Loads with lower numerical groups are selected for • Level 1 was triggered at t = 1.330 seconds and UF L1
shedding first, starting with the number 1 and moving up the list required to shed (RTS) = 1.6 MW.
as loads are available, until the total amount of load selected for
shedding matches or exceeds the amount of load required for • Level 2 was triggered at t = 1.417 seconds and UF L2
shedding. RTS = 2 MW.

F. Load-Shedding Signals • The minimum frequency was 53.86 Hz.

Load-shedding signals are sent out by the redundant • The Type-1 generators did not trip and the remaining
controllers, propagated through the DCONs, sent to the field load on the island was 2.3 MW.
relays that are installed on load feeders, and then sent to the load
trip coils. When a load is selected for shedding, the load trip The same scenario was repeated with CLS enabled and
signal is sent out by the controller and held for 60 seconds. This UFLS disabled. The load on the island was approximately
ensures that the signal is propagated through all devices and 5.9 MW and the Type-2 generator was tripped at t = 1 second.
reaches the load. The trip signal is then propagated through the At t = 1.200 seconds, the CLS shed 3.6 MW of load on the
load-specific relay to the breaker trip coil. The average round- island. The frequency response of the island for CLS is
trip time from contingency detection to load-shed command represented by the dashed black line in the plot shown in Fig. 8.
transmit was observed to be in the range of 25–30 milliseconds The observations are as follows:
in this case study. This time does not include contingency and • The minimum frequency was 55.48 Hz.
load-breaker opening time, which can vary from
45 milliseconds (three-cycle breakers) to as much as • The Type-1 generators did not trip and the remaining
200 milliseconds. Fig. 7 shows a typical data flow, including load on the island was 2.3 MW.
detection of the contingency trigger and sending out a load-shed B. Test-2: All Generators Running With Maximum Import
command. and an Upstream Fault on the Utility
The power system is connected to the utility with a
maximum import of 4.8 MW and all generators are kept online.
The UFLS Level 1 and Level 2 are disabled and CLS is enabled.
A three-phase fault on the utility-side transmission line is
simulated at t = 0.917 seconds. The transmission line breaker is
opened approximately three cycles after the fault at t = 1 second
to isolate the faulted section of line.
The island-side main utility breaker is opened at
t = 1.115 seconds and CLS operates based on the status of this
specific breaker. Two hundred milliseconds after the main
breaker opened at t = 1.315 seconds, the CLS controller sheds
4.8 MW of load. The minimum frequency was 57.44 Hz.
The frequency response of the island is plotted in Fig. 9.
C. Test-3: Islanded With Two Type-1 and One Type-2
Generators and One Steam Turbine Running, and One
Type-2 Generator and a Steam Turbine Tripping With
the CLS Undershedding
The power system is islanded from the utility and two
Type-1 generators each running at 1.15 MW, one Type-2
generator running at 3.6 MW, and a steam turbine running at
Fig. 7. Load-Shedding Timing Diagram
0.6 MW are kept online with the contingency-based load-
shedding processor (CLSP) enabled and UFLS disabled. The
load on the island is approximately 6.5 MW. The Type-2 • The Type-1 generators did not trip and the remaining
generator and steam turbine are tripped at t =1 second and at load on the island was 2.1 MW.
t = 1.200 seconds, and the CLS shed 3.2 MW of load on the
island. However, the required-to-shed amount was 4.2 MW. • The maximum frequency for this scenario was 63.84 Hz
Therefore, undershedding by CLS caused the system frequency and the overfrequency protection of the Type-1
to decay. The frequency response of the island for CLS is generators was 65 Hz for 5 seconds.
represented by the dashed black line in the plot shown in Fig. 10. Based on the designed coordination between CLS and
The observations are as follows: UFLS, the controller should block UF Levels 1 and 2 to prevent
• The minimum frequency was 51.27 Hz. overshedding of the load on the island. However,
Underfrequency Levels 3 and 4 should be kept active as a safety
• At t = 2.303 seconds, UF L3 picked up with margin for the system in cases where CLS or other UFLS levels
RTS = 1.2 MW, which helped with frequency recovery undershed loads on the island.
and saved the system from collapsing.

Fig. 8. System Frequency Response for Test-1

Fig. 9. System Frequency Response for Test-2

Fig. 10. System Frequency Response for Test-3

 VII. CONCLUSION
A wastewater treatment plant is a critical infrastructure
that is required to always operate with reliable and trusted
operations. These operations can only be guaranteed with the
availability of a reliable electric power system and processes.
Any power system disturbance can easily destabilize the
overall power system and can lead to blackout scenarios. Such
blackouts can cause maloperations that can have a direct effect
on the public health because usually water that is cleaned
through these treatment facilities is released into river
systems.
The work described in this paper explicitly focuses on the
design and engineering of high-speed load-shedding systems,
which are based on the latest SDN-based network
infrastructure. The results presented show the significant
impacts of timely load-shed actions that helped the power
system survive without getting into a blackout situation.
The system developed for this plant will not only make the
plant power supply more robust and reliable, but it also will
allow the plant operator to make decisions in cases of natural
disasters, such as flooding or hurricanes, and manage the
power flow to critical loads, eventually keeping the plant
running and safeguarding both the environment and public
health.
REFERENCES
[1] M. Kazmi and G. M. A. Akhtar, “Power Management and Control
System – Insights Into Design and Testing,” proceedings of the 8th
Saudi Arabia Smart Grid 2018 Conference, Jeddah, Saudi Arabia,
December 2018.
[2] S. Manson, B. Nayak, and W. Allen, “Robust Microgrid Control
System for Seamless Transition Between Grid-Tied and Island
Operating Modes,” proceedings of the 44th Annual Western Protective
Relay Conference, Spokane, WA, October 2017.
[3] S. M. Manson, A. Upreti, and M. J. Thompson, “Case Study: Smart
Automatic Synchronization in Islanded Power Systems,” proceedings
of the 51st IEEE/IAS Industrial & Commercial Power Systems
Technical Conference, Calgary, AB, May 2015.
[4] K. G. Ravikumar, T. Alghamdi, J. Bugshan, S. Manson, and S. K.
Raghupathula, “Complete Power Management System for an Industrial
Refinery,” proceedings of the 2015 IEEE Petroleum and Chemical
Industry Committee Conference (PCIC), Houston, TX, October 2015.
[5] W. Allen and T. Lee, “Flexible High-Speed Load Shedding Using a
Crosspoint Switch,” proceedings of the 32nd Annual Western
Protective Relay Conference, Spokane, WA, October 2005.
[6] J. W. Bartlett and W. F. Allen, “Contingency-Based Load Shedding,”
U.S. Patent 9,519,301 B2, December 2016.

Previously presented at the 2022 IEEE Electrical Power

and Energy Conference, December 5-7, 2022.
© 2022 IEEE – All rights reserved.
20230213 • TP7091