2010 International Conference on Power System Technology
Exploiting Cloud Computing for Power System Analysis Qiuhua Huang, Mike Zhou, Yao Zhang, Zhigang Wu
state estimation [2-8] . Abstract-Cloud computing provides a new paradigm for
Grid Computing is a good candidate solution, however,
easy access to larger scale computing resources over the Internet,
there are some critical issues must be solved before it can be
thus offering an alternative solution to huge data processing and
turned into real application. Some are technical: the setup,
heavy computational work in power system. In this paper, InterPSS Cloud Edition, a cloud computing based platform for power system analysis, is set up by InterPSS development team
configuration, operation and maintenance of the platform and the applications implemented on it usually demand expertise
using Google App Engine (GAE). At the current release, it
or special IT knowledge. In addition, some experiments or
provides mainly three functions including load flow, contingency
applications have to be reorganized or redesigned to fit into
analysis and ODM (Open Data Model) power system simulation
the
data transformation services, and it is accessible anywhere
implemented as bag of tasks applications, workflows ,and
around the world, 24x7, via internet. An overview of this project is given and tests regarding its functionalities are presented. Finally, the future applications and challenges of exploiting cloud
computing for
power system analysis are
discussed. Keywords:
Power
system
analysis,
Cloud
computing,
InterPSS Cloud Edition, Google App Engine
that application run on Grids are
MPI(Message Passing Interface) parallel process [6].These technical obstacles, with a perfect solution, can prevent it from wider applications both in academics and industries field. Cloud computing[IO-12], a new paradigm for computing technology and IT service, can address many of problems mentioned above. By means of virtualization technologies,
I. INTRODUCTION
W
grid models in
cloud computing provides a flexible mechanism for offering
ith the deregulation and constant expansion over the
past decades, power systems have developed from
end users a variety of services, from hardware to application level, thus engineers can have easy access to large distributed
isolated systems to interconnected interregional lintemational
computing
connections.
execution environment, virtually like working on their local
Also
power
systems
are
moving
towards
resources
and
completely
customize
their
renewable energy sources, most of which are integrated into
computers, without the need of purchase, maintenance or
the systems as distributed generation unites [1]. Therefore, a
even understanding of sophisticated hardware and high
higher
performance computational methods[12]. Other important
level
of
operation,
control
and
coordination
is
imperative, which is based on large data processing and/or
features include its scalability and pay-as-you-go billing
computing intensive simulations, and requires for a higher
model. Embedded with these merits, it is increasingly tested
performance computing solution [2, 4, 6, 8].
and used for scientific applications. The Science Clouds
Traditionally,
power system simulation and analysis uses
project, initiated in mid-2008, has proved the feasibility of
a computer of a set of computers at one physical location.
using cloud computing for scientific computing and provided
When large amount of data are acquired to be processed for
early experiences of such new paradigm from a research point
online
operation
decision
support,
for
example
on-line
contingency analysis, computing resource often is the limiting
of view[13]. It has been applied to the climate research [14] and gene expression and brain imaging [12]. In 2009, the U.S.
computing-intense
Department of Energy(DOE) stated the Magellan Project and
simulations[2,4,8]. The initial proposed solution was parallel
set up a test bed to examine cloud computing as a cost
factor
to
the
meet
such
demand
of
processing, but proved to be hard-to-operate and expensive
effective
[8]. Then the Grid computing [9] was later adopted and used
scientists [15]. Along with this new trend in other fields,
in the researches of simulation, reactive power optimization,
power companies are beginning to show attention and interest
load balancing, stability and security analysis, distributed
and
energy-efficient computing
paradigm
for
on it. Mercury Solar Systems uses a cloud computing CRM (customer relationship management) to better meet the energy
Qiuhua Huang, Zhigang Wu, Yao Zhang is with School of Electric Power, South China U niversity of Technology, Guangzhou, 510641, Ch ina (E·mail:
[email protected]). Mike Zhou is with InterPSS System L LC, Houston, Texas, USA (E·mail:
[email protected] ).
978-1-4244-5940-7/10/$26.00©2010 IEEE
needs
of
its
customers[16]. Engineers
from
the
China
Southern Power Grid have proposed to take advantage of cloud computing to upgrade its power analysis software(PAS) for smart dispatching[17].
2
computing
Measured Service: A usage-based billing model where
application researches in other disciplines, to our knowledge,
Although
there
are
some
early
cloud
users essentially "rent" virtual resources and pay for what
limited (almost none) researches on utilizing its functionality
they use.
in power system analysis have been done. However, Its low
Resource pooling: The provider's computing resources are
cost, agility, reliability and scalability makes it a potential
pooled together, with different physical and virtual resources
approach for future power system application. In this paper, a
dynamically assigned and reassigned according to consumer
platform, InterPSS Cloud Edition [18] host on Google Cloud,
demand.
is set up by InterPSS development team using Google App Engine (GAE) [19], for studying the potential application of cloud computing in power system analysis. Study cases are presented to evaluate the performance of the cloud computing platform. The rest of the paper is organized as follows: in section II of this paper, an overview of cloud computing is provided. Section III presents the details of building of InterPSS Cloud Edition on the Google Cloud. Section IV describes the test cases done on the InterPSS Cloud. Section V discusses the
C. Three Delivery Levels
Based
on
the
level
of abstraction
presented
to
the
programmer and the level of management of the resources, all the Cloud Computing service accessible to the public can be categorized into three delivery levels ,they are Infrastructure as a Service(laaS), Platform as a Service (PaaS),and Software as a Service(SaaS)[7]. Figure 1 gives an overview and provides the corresponding representative companies and their
services
offering.
opportunities and challenges of exploiting cloud computing
�
for power system analysis. The paper is concluded in section VI with reference presented in section VII.
SaaS
II. AN OVERVIEW OF CLOUD COMPUTING A.
PaaS
The term "cloud computing" refers to any computing that
Docs
G.9.,Qgle
Definition
capability
Coogle
,.ueforcc.com
is
delivered
as
a
service
over
Azure
the
internet .While it encompasses many aspects(i.e., distributed
vrnwore
IaaS
computing resources, virtualization ,network, datacenters, etc) and there is no authoritatively accredited definition, it is possible to identify some key features that characterize this
Fig.
I. Three delivel)' levels of cloud computing
technology through the most frequently used definitions[12]. A Berkeley view of cloud computing is that it refers to both
IaaS, as showed in fig.1 , lies at the bottom of the cloud
the applications delivered as services over the Internet and the
stack, and it usually refers to a practice of delivering IT
hardware and systems software in the datacenters that provide
infrastructure based on virtual or physical resources where the
the services, then the datacenter hardware and software is
consumer
what they call a Cloud[IO].According to the NIST definition
representative IaaS solutions provider is Amazon, with its
can
deploy
and
run
arbitrary
software.
A
of cloud computing, it is a model for enabling convenient, on
Elastic Compute Cloud (EC2) providing computing service
demand network access to a shared pool of configurable
and Simple Storage Service (S3) providing storage service.
computing
some researches were done to experiment with EC2 and S3
resources
(e.g.,
networks,
servers,
storage,
applications, and services) that can be rapidly provisioned
for scientific computing [12, 14] ,showing the potential of
and released with minimal management effort or service
IaaS , or EC2-style cloud computing, as a high-performance
provider interaction[II].
solution.
B.
PaaS provides a platform where users ,or customers can
Five Essential Characteristics
create and run their applications or programs .It usually
There are five key characteristics for Cloud computing that
provide an application framework and a set of API that can be
distinguish it from previous computing model and provide a
used by the users to develop their applications for the cloud
basis
for
understanding
it
and
its
role
and
potential
[10-12]. Following this model, Google and Microsoft both set
application in academic and research sectors. The five key
up their application platforms, Google App Engine [19] and
characteristics are[ll]:
Windows Azure [20], respectively. Google App Engine is a
Rapid elasticity: Capabilities can be rapidly and elastically
provisioned to have scalability. On-demand self-service:
A consumer can unilaterally
provision computing capabilities. Broad network access: large scale resources are available
over the network and accessed through standard mechanisms.
platform that enables you to build and host web apps on the same systems that power Google applications. It features fast development and deployment; simple administration, with no need to worry about hardware, patches or backups; and effortless scalability. Now it supports applications written in Java and Python.
3
SaaS provides the users with provider's applications
DataStore, Google's distributed database system, is also used
running on a cloud. Since SaaS is mainly designed for
to save the intermediary file or data as well as the study case
commerce
if the users select the option.
and
business,
and
limited
control
and
configuration is allowed, thus it is not suitable for research, and no scientific research based on SaaS has been reported or
Application:
published so far.
InterPSS Cloud Edition
III. INTERPSS CLOUD EDITION ON GOOGLE APP
Google App Engine
ENGINE InterPSS Cloud Edition is a cloud-based, or GAE-based
Google
specifically, implementation of InterPSS that leverages the
DalaSlun::
Google App Engine to perform computing intensive power system
analysis.
The
details
of
this
work
include
the
following: A.
An Overview ofInterPSS
InterPSS[21] is an open-source , Internet technology based software system for design , analysis and simulation of power systems . It is designed and developed with component based
Fig.2. interPSS Cloud Edition on Google App Engine
development approach, therefore it features an open and loosely coupled plug-in architecture, which allows users extend its functionalities easily by plug-in, and equally
D. Functionalities in InterPSS Cloud Edition
important, allows the components to be integrated into other
InterPSS Cloud , at current release , provides mainly three
system to provide power system simulation and analysis
functionalities, including Load Flow Analysis, Contingency
service. Specifically,
Analysis based on complete AC load flow and Open Data
its
main
power
system
simulation
functions are packaged into the power system simulation
Model(ODM)[22]
framework as the core library and can be integrated into other
transformation service. Figure 3 shows these functionalities
systems as a power system simulation engine. B.
based
power
system
data
format
provided in InterPSS Cloud (already logged in with a user account).
Why Google App Engine?
InterPSS Cloud Edition
As described in the section II, Google App Engine is a
[ user.lhuang J
platform for developing scalable applications, and it is built on the infrastructure of Google, so it merits the high reliability, performance and security of Google's system. App � Upload New StudyCase
Engine's Java runtime environment supports standard Java
[ LoadtIowJ
technologies, including the NM, Java servlets, and the Java
Sensltlvl1y
Contingency
ODM
case:SZE00924_2_3Trans.raw
programming language. Furthermore, now it provides a free
Loadflow [ NR NR+ .EQ PO+ J
but limited service (free quota of 500 MB of storage and 1.3
Loss PJlocatlon
million requests daily) for developers to build applications, thus moving the obstacle of cost and operation for our
Copyright C2OO9lnterPSS ,AJI rights reserved I Feedback .,..., .. eo", ... _
scientific application. C. Implementation
Fig3 .The functionalities in interPSS Cloud Edition
The Architecture of the proposed InterPSS Cloud Edition in this paper is showed in figure 2. As InterPSS core
ODM is an open model for exchanging power system
simulation engine provides the computing and analysis
simulation data, and InterPSS has a good support of it.
capability, it runs within the Java Virtual Machine
(NM) of
Several Xformat-to-ODM adapters have been developed,
GAE once it is deployed in the Google cloud, therein it
herein the Xformat includes PSS/E, UTEC, BPA, PSAT,
provides the capabilities to response to the requests, which, in
InterPSS, and the ODM-to-InterPSS adapter is also developed.
this paper, regarding to different kinds of power system
In
analysis, i.e. ac Idc power flow, contingency analysis, etc,
exchanging model is set up; therefore, any load flow data of
from
these formats is acceptable by InterPSS Cloud.
the
users.
With
the
help
of
the
Application
Programming Interfaces (APIs) provided by GAE, a website
this
The
way,
an
contingency
"Xformat->ODM->InterPSS"
analysis
includes
three
types:
data
N-I
(http: //cloud.interpss.com) is set up as a front-end, enabling
analysis, N-I-I analysis and N-2 analysis. N-I Analysis
users to upload the data for processing, define the study case.
conducts complete AC load flow for each contingency of a
4
branch open; in the N-I-I analysis mode, first open each
the perspective users an easy access to test and verification.
branch in the power network and run complete AC load flow
Since the contingency analysis depends on the complete
analysis for each contingency, then for each N-I contingency,
AC load flow, a contingency test can verify both the Load
open each branch with branch Mva rating violation and run
Flow Analysis and the Contingency Analysis functions.
complete AC load flow analysis for each N-I-I contingency;
According to the tests, N-I analysis of this lIS-bus system
N-2 analysis open two branches each time and run load flow
takes about 2 seconds (time may vary, but not significantly,
for all possible contingency.
considering the Internet condition), thus the effectiveness and
As for the ODM transformation, an application for real
high performance is proved, especially when considering it is
world data conversion can be found in reference [23]. For
achieved within the limited quota of free service. With
what the users concern, InterPSS Cloud provides such service,
Google's pricing service, more computing power is available
and through it, one can easily transform the supported data
and accessible, if necessary, with the corresponding billing. In
format to an ODM XML file[I8] .
addition, the detailed results are posted back once the analysis completed, part of the bus voltage margin report and branch
E. How it works?
Since InterPSS simulation engine has been deployed in GAE and running there 24x7, users can access to it anywhere around
the
world
via
the
internet
.More
importantly,
especially for the users, InterPSS Cloud is easy-to-use and user-friendly. Figure4.
The user
After
logging
operation process is showed in on the
InterPSS
Cloud
website
(http://cloud.interpss.com), users can upload the simulation data in one of the supporting data formats by selecting the corresponding data file adapter first, and then choose one analysis function, all the rest is done by InterPSS Cloud and results are automatically posed back to them through the browser.
Mva margin report is showed in figure S and 6, respectively. The voltage margin report lists the lowest voltage for the bus for all contingencies and compares it with a Bus Low Voltage Limit.
The
description
tells which
contingency
caused the lowest voltage. For example, as is reported in figure
S, [x]BusllO->Bus280(l) indicates that the open
branch BusllO->Bus280(l) contingency caused the lowest voltage. The branch Mva margin report lists the most severe branch Mva rating violation, indicating with a negative sign. The description tells which contingency caused the most severe branch rating violation. For example, as is showed in figure 6, the open branch BusllI->BusllO causes the branch BusISI->BusISO most severe rating violation. Bus Vol tage Limi t: Bus Id
o
InterPSS Cloud sends results back to user after processing the request
Fig.4 .The working mechanism of interPSS Cloud Edition
IV. TEST CASES One of the goals of InterPSS Cloud Edition was to
O. O.
9331 9331 0.8774 0.9774 0.9579
Bus140
8437
85]
O. O.
7850
Description
8.3% 8. 3% 2.7% 12.7% 10.8%
[x]Bus110-)Bus280(1) [x]Bus110-)Bus280(1) [x]Bus110-)Bus280(1) [x]Bus110-)Bus280(1) [x]Bus110-)Bus280(1)
-0. 6%
[x]Bus110-)Bus280(1)
-6. 5%
[x]Bus110-)Bus280(1)
Fig.5. Bus Voltage Margin of a liS-Bus system
f2\
InterPSS Cloud perform s analysis on Google \:.J Cloud infrastructure
O.
LowVolt LowMargin
Bus342 Bus341 Bus340 Bus241 Bus240
Bus141
[1. 10,
Branch Id MvaFl"" �vaRating P t jQ �argin Description f=====--===============--========t==============
Bus260->Bus26!([) Bus31->Bus30(1) BusI70->BusI80([) Bus220->Bus221 (1) BusI60->BusI61(l) Bus260->Bus263 (1) Bus2l1->Bus210(1)
q\lIslSI->BlIslSolll
10.2 70.5 �8. 2 33. 5 66.2 56.2 63.7 369.0
150.0 150.0 65.0 150.0 150.0 65.0 150.0 300 0 .
( 9.5tj ( 70. Hj ( �3. 1+j ( 32. 1+ j ( 61.O+j ( 53. 4tj ( 61.3tj (362.O+j
3.7) 3.6) 21.6) 9.6) 25.6) 17 . 4) 17.5)
71.4)
9311 5311 26% 78\1 56% 14%
� L:mJ
[dBus261->Bus350([) [']BusI0->Bus3020([) [dBusllO->Bus280(1) [,]BusllO->Bus280 (I) [']BusllO->Bus280([) [']BusllO->Bus280(1) [']Bus210->Bus230([)
hIBllslll ->BlIsl1011ll
demonstrate that an existing, not cloud-purpose designed Fig.6. Branch Mva Margin of a liS-Bus system
software or package can work well on a cloud platform (its programming language should be support by the platform). Considering its implementation and functions provided, this object has been achieved, with InterPSS Cloud, at its early stage, running quite well on GAE, which is also supported by the tests presented below. As a further goal, its performance should be evaluated. In this section, tests of its main functionalities (i.e., load flow, contingency analysis and ODM transformation) with a IIS bus system in PSS/E V30 format are performed on InterPSS Cloud Edition and the test data is provided along with the user guide [24] on the InterPSS community website, offering
A test on ODM data transformation is also performed with the
same network data, figure 7 shows the load flow
data«a).BuslO and (b). Branch "BuslO_to_Bus20") in
the
returned ODM XML file. It can be seen from the result, compared with traditional text format, the data in such xml file is more meaningful and easy to read and visualize, making it as a better way to share data among different operators with a common ODM schema. The ODM is still at an early stage, and the adapters for transforming data from ODM to other data formats are under
5
ODM-to
development of smart grids[27]will likely have to back on the
InterPSS adapters finished[22], thus the function of data
development,
with
only
ODM-to-PSAT
and
cloud computing technology to deal with the increasing large
exchanging between different formats with ODM as the
scale
intermediary is not completed and available at this moment,
interoperation [28].
but it will be one of next phases. Finally, more voluntary participations scientific
and
and contributions from the power system research
community
are
welcomed
and
definitely needed for progress and further applications.