IEEE Electrification - December 2022 - 56

Forecast Electricity Prices
Situational Awareness
areness
orecast Electricity Prices
Situational
Optimize Own
Generation Dispatch
Forecast Building
Loads
Optimize Demand
Response
Monitor Equipment
Performance
Analytical Applications Behind the Meter
Figure 7. Analytical BTM applications: electricity consumers monitor equipment, forecast electricity
use and renewable generation, and optimize the system performance.
load data into the limited memory of a single machine
and execute a single thread against the data. If the data
did not fit into the memory, they could not be run, or they
would have to spill over to disk storage, generating read
and write operations, swapping the necessary information
in and out of the memory for computation. Multiple
discrete tasks were run serially on a single logical thread,
iteratively solving each task. There was little notion of
parallelization and multithreading, where single physical
cores are split into two logical cores, each of which can
execute discrete processes. In tandem with the proliferation
of memory and parallelization of tasks, computer virtualization
was adopted.
Computer virtualization is the splitting of computing
hardware into abstract virtual resources, simulating the
hardware functionality of having multiple individual
computing devices. Virtualization has led to economies of
Accessibility to High-Frequency Data
Data
Preparation
Variable
Selection
Model
Training
Model
Automated
Detection
Accuracy
Degrades
Over time
Model
Retrain
Model
Assessment
APIs
An API is a common software tool that enables the programmatic
exchange of an application's data with those
external to the application. This is done through an API
call, which either pulls or pushes data from and to the
application. The increasing prevalence of APIs has allowed
for easier data exchange and allowed for real-time predictive
modeling since a manual download of data is no longer
needed but can be obtained through a programmatic
call through the code.
There are now APIs that can serve electricity usage and
Redeployment
Figure 8. An example of model drift monitoring. As equipment ages,
behavior can change, and previously deployed models become less
accurate. Software can monitor the model fit and retrain when models
should be refreshed.
56
IEEE Electrification Magazine / DECEMBER 2022
rate schedule data. Having these data available is useful for
many use cases, some of which include cost optimization,
demand response, and DER generation planning. Also, having
access to weather forecast data can help with wind/
solar production planning. Forecasts that once required a
download of the weather forecast can now rely on an API
that brings in the weather forecast in real time. This is useful
for BTM DER forecasting and makes real-time analytics
Forecast EV Charging
Requirements
Optimize EV Charging
Schedules
Forecast Renewable
Generation
Optimize Storage
Schedules
scale and efficiency in computer
resource allocation. Virtualization
allows for elastic scaling and resizing
of computing power to meet
computing demand. This is a simple
expansion of computing
resources to meet workloads. Combined
with access to high-speed
Internet, virtualization is a key
building block of the public cloud.
Computing has become accessible
and economical with the
proliferation of the public cloud.
Made possible by the Internet,
accessing virtualized computing
environments is now common.
The scale and popularity of the
public cloud has led to further
economies of scale and offered
access to cheap and powerful
computing resources. The rich
cloud marketplace grew new services and products to
assist computational activities. The services in particular
are the serverless storage of data and the ability to
access vast amounts of data quickly, as provided in
products such as Amazon Web Services S3, a cloud storage
device, and Snowflake, a cloud database platform.
This new computing landscape has made the storage of
data simple and economical. While more and more data
are being stored, computing resources have also grown
cheaper and easier to access. The confluence of the
computing accessibility revolution is new data science
opportunities. Highly granular dense data are now the
next horizon for study.

IEEE Electrification - December 2022

Table of Contents for the Digital Edition of IEEE Electrification - December 2022