IEEE Power & Energy Magazine - January/February 2020 - 70

In the context of distribution systems, having the OPF as the decisionmaking engine allows for myriad ADMS applications because its
formulation can be adapted to the corresponding objective.
resources. Even though remote-control capabilities have
been enhanced through smart grid efforts, industry prac-
tices remain mostly manual and use mostly backup feeders
for restoration. These methods are, therefore, neither opti-
mal in a decentralized distribution grid paradigm with high
DER penetrations nor resilient to major outages, as central-
ized systems are prone to single-point failures. Realizing
the full potential of the emerging distribution grid for resil-
ience requires a facilitating framework for decentralized
operation within the existing centralized and hierarchical
power distribution systems. Algorithmically, DSR solves
a feeder reconfiguration problem, typically a combinatorial
optimization problem. The growing complexity of distri-
bution grids due to numerous sectionalizing switches, tie
switches, and DERs available for restoration significantly
increases the intricacy of the inherent combinatorial DSR
problem. This calls for an enabling technology that quickly
and optimally restores distribution grids using all available
resources under all fault scenarios.
DSR, although widely studied in current literature, still
requires attention due to growing service reliability and resil-
ience expectations. Earlier DSR methods focused on design-
ing expert systems and heuristic search methods to avoid
solving the combinatorial problem. Soft computing algo-
rithms, including genetic algorithm, particle-swarm opti-
mization, simulated annealing, and fuzzy set approaches,
have also been proposed. Because of the need to solve for
increased levels of system complexity and the added require-
ments for considering resilience to extreme events, model-
based DSR has emerged as a potential means of addressing
service restoration issues. Service restoration using math-
ematical programming for an unbalanced distribution sys-
tem has been proposed as well by several researchers who
formulate the DSR problem as a mixed-integer nonlinear
program (MINLP). Although accurate, MINLP formula-
tions are computationally unattractive because they do not
scale well, i.e., the simulation time increases significantly
as the complexity of the restoration problem increases. This
has led to the development of scalable, linearized formula-
tions with mixed-integer decision variables. As a result, lin-
earized, multiphase ac power flow models have been widely
used by researchers for model-based DSR. Mathematically,
the problem remains nonlinear due to integer/binary decision
variables. Recent advances in solvers for mixed-integer linear
programming (MILP) problems have the potential to scale
the model-based DSR for multifeeder distribution systems
with linearized power flow models.
70

ieee power & energy magazine

In this section, we describe a scalable framework used
for restoring a large-scale, three-phase, unbalanced power
distribution system that combines normal feeder restoration
with intentional islanding methods during extreme events
into one unified framework. This framework has been devel-
oped at Washington State University and is being integrated
into Pacific Northwest National Laboratory's (PNNL) Gri-
dAPPS-D platform, an open-source platform for hosting
advanced distribution system applications. The proposed
generalized framework aims to improve reliability dur-
ing typical outages and resilience during extreme events
by optimally utilizing available backup feeders and DERs.
With this in mind, we determine feeder restoration and a
reconfiguration plan using all available resources: backup
feeders, microgrids, and DERs. Also, intentional islanding
methods are employed to ensure resilience to extreme events
using DERs and remotely controlled smart switches to
restore critical loads during emergency conditions, espe-
cially when the upstream subtransmission/transmission
system is outaged.
A two-stage formulation is used for restoration. In the first
stage, an optimal feeder restoration and reconfiguration plan
using the utility and DER data are obtained by solving an
MILP for the three-phase, unbalanced, multifeeder power
distribution grid. The second stage implements a completely
decentralized framework to control the operation of DER-
supplied restored islands. The schematic for the approach
is presented in Figure 4. For a multifeeder test system with
1,069 buses (~2,500 single-phase nodes), a feasible restora-
tion plan is obtained within a minute. Note that any outage
of fewer than a 5-min interval counts toward the Momentary
Average Interruption Frequency Index instead of the System
Average Interruption Frequency Index (SAIFI); therefore, if
the restoration can be automated, a model-based DSR can
help significantly improve the SAIFI.
The realization of an autonomous restoration using
model-based DSR techniques requires a measurement-
and control-rich environment that provides post-disaster
situational awareness and the ability to remotely deploy
the decisions for restoration. A successful deployment thus
relies on the ADMS to enable real-time communication and
data exchange between several systems employed by distri-
bution companies for distribution system management. A
reliable communication system must exist between differ-
ent utility subsystems including SCADA, DMSs/ADMSs,
outage management system (OMSs), advanced meter-
ing infrastructure (AMI), microgrid energy management
january/february 2020

IEEE Power & Energy Magazine - January/February 2020

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - January/February 2020