IEEE Electrification Magazine - December 2016 - 11

types of controllers. We use GridLAB-D to simulate a distribution feeder and the many houses connected to it. GridLAB-D is an open-source, agent-based, quasi-steady-state
time series power system and load simulator developed by
the Pacific Northwest National Laboratory that can simulate a number of specific end-use loads, including air conditioners, water heaters, dishwashers, clothes washers
and dryers, and distributed energy resources such as PVs,
battery systems, and electric vehicles. The IESM is
designed to run on a high-performance computer to allow
for the parallel execution of hundreds of instances of complex controllers, such as our HEMS algorithms. Smallerscale simulations in IESM can be run on a single computer.
The IESM employs a distributed, real-time, discreteevent modeling and simulation paradigm to manage time.
In each iteration,
1) the cosimulation coordinator requests the time of the
next event (tn) from each of the components
2) the cosimulation coordinator determines a future
time to which all the components should advance
(tadv) by selecting the minimum tn and broadcasts this
to all the components
3) each of the components then advances to t adv ,
updates its internal clocks, and executes any code or
input events
4) each component shares output messages.
The flowchart in Figure 4 depicts the data exchange,
facilitated by the cosimulation coordinator, between a
house simulated in GridLAB-D and a HEMS. Consumer
preferences such as the objective component weights
and desired set-point profiles are input to the HEMS.
Price and weather data are provided to both the HEMS
and the house simulation. In addition, weather and price
forecasts throughout the scheduling horizon (we use one
day) are provided as inputs to the HEMS. The HEMS optimizes the appliance set points throughout the whole
scheduling horizon and outputs the set points for the
next time period to GridLAB-D, which uses them to calculate appliance power use and customer utility values,
such as indoor air temperature and battery state of
charge. These values are provided as inputs to the HEMS
for the next optimization.

IESM Simulation on HPC

Market

Cosimulation Coordinator

Natural Gas
Simulated Distribution Feeder
Figure 3. The IESM cosimulation platform enables the integrated
simulation of physical systems, including distribution feeders, buildings and appliances, controllers, and markets.

25% of the static load on this feeder with 505 houses connected in groups of five through single-phase, centertapped distribution transformers. The houses are broadly
classified into four types: old (constructed prior to 1980)
and small (<2,000 ft2); new (constructed later than 1980)
and small; old and large (>2,000 ft2); or new and large. A
random variation is added to the thermal characteristics

Consumer
Preferences

Setup

Optimize Postprocess

case Study
As a first step toward understanding the impact of
advanced HEMSs on the operation of distribution systems, we consider a case study with a high penetration
of HEMSs that are not coordinated through participation
in a transactive energy market; instead, each operates
independently based on a common TOU price signal. We
model the distribution system using a taxonomic GridLAB-D feeder model. A taxonomic feeder is an artificial
feeder that is representative of a class of real feeders. We
use one (R5-25.00-1) that represents a heavy suburban
and moderate urban load in the southeastern United
States that exhibits a hot and humid climate. We replace

Customer Utility
(e.g., Temperature,
Battery SoC)

Price and Weather

Appliance Power
(e.g., Water Heater,
Air Conditioner)

Set Points
(e.g., AirConditioner
Cooling, EV
Charge Rate)

GridLAB-D
House Simulation
Figure 4. The flowchart of the data exchange between the HEMS and
house simulation.

IEEE Electrific ation Magazine / d ec em be r 2 0 1 6

Table of Contents for the Digital Edition of IEEE Electrification Magazine - December 2016