Graph Computing based Parallel Power Flow Algorithm and Graph Visualization for Power Distribution Networks Jun Tan*, Yi Lu § , Kewen Liu**, Hong Fan*, Guangyi Liu*, Renchang Dai*, and Zhiwei Wang* *Global Energy Interconnection Research Institute North America, San Jose, CA 95134, USA **Global Energy Interconnection Research Institute, Beijing, 102209, China §Sichuan Electric Power Corporation, China [email protected]Abstract—With the emergence of the graph database and the graph computing technologies, the power system is facing a new era of technology development. Meanwhile, the fast-growing power distribution systems with increased size and complexity require more efficient data management systems and faster power flow solving algorithm. These challenges could be well solved by applying the graph data model (GDM) and graph computing technologies as the GDM provides more efficient data management methods and the graph computing is suitable for the node based parallel computing. This paper proposes a GDM based power distribution network modeling approach. Then a graph computing based parallel power flow algorithm and a graph visualization based power flow software have been developed based on it. The proposed parallel power flow algorithm is implemented on a graph database platform and the simulation results show that it is able to effectively reduce the computing time of the power flow with large test systems. Moreover, the power flow software is able to provide vivid data visualization and perform flexible data analysis and data management functions. Index Terms—Graph computing, parallel power flow, graph visualization, backward forward sweep. I. INTRODUCTION With the fast-growing penetration of the renewable resources such as solar and wind generations, the power distribution systems are becoming more complex and it requires higher efficient data management methods and faster power flow solutions [1]. To achieve the goal of efficient data management and faster computation, new modeling methods for the power systems are required to break through the limits brought by traditional relational database. Graph computing is a new technology which constructs the power network from the viewpoint of a graph. The graph database also provides more efficient data management approaches by storing the information directly on vertices and edges. Its graph structure for connecting the vertices and edges also demands less processing time for retrieving data at any depth [2]. Different types of graph platforms have been developed in the market such as Pregel [3], Neo4j [4], Giraph [5], TigerGraph [6], etc. for various applications. Many research has been conducted in the field of graph computing and its application in power systems [7]-[11]. Reference [7] adopts a graph theory based network flow analysis in real time power system operations to improve network connectivity visualization. A graph based computational framework for coupled infrastructure networks’ optimization has been proposed in [8]. Paper [9] proposes a graph partition based mixed integer linear programming approach for power system islanding operation. A graph theory based optimal power quality monitors planning approach is presented in [10]. Reference [11] applies factor graphs for distributed power system state estimation. However, very few study has been carried out in the field of graph computing based parallel power flow algorithm and the graph based visualization of power distribution networks. Thus, this paper will propose a graph data model (GDM) based power distribution network modeling approach which is able to provide fast parallel power flow platforms, efficient data management approaches, and graph based data visualizations. Base on the GDM of the power distribution network, a graph computing based parallel power flow algorithm has been proposed. The parallel power flow algorithm adopts the hierarchical group synchronization (HGS) parallel computing mechanism in bulk synchronous parallel (BSP) [12] computing model and it is able to effectively reduce the computing time for the power flow when dealing with large systems. Finally, a software package has been developed based on the GDM of the power distribution network. The graph visualization based software is able to provide vivid data visualization and perform flexible data analysis and data management functions such as voltage profile along the feeder, network reconfiguration, etc. This study has made contributions in several major aspects by: (1) proposing a GDM based power distribution network modeling approach; (2) proposing a graph computing based This work was supported by State Grid Corporation technology project 5455HJ180018.
5
Embed
Graph Computing based Parallel Power Flow Algorithm and ... · requires higher efficient data management methods and faster power flow solutions [1]. To achieve the goal of efficient
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Graph Computing based Parallel Power Flow Algorithm and Graph Visualization for Power
Distribution Networks
Jun Tan*, Yi Lu§, Kewen Liu**, Hong Fan*, Guangyi Liu*, Renchang Dai*, and Zhiwei Wang*
*Global Energy Interconnection Research Institute North America, San Jose, CA 95134, USA **Global Energy Interconnection Research Institute, Beijing, 102209, China
Abstract—With the emergence of the graph database and the graph computing technologies, the power system is facing a new era of technology development. Meanwhile, the fast-growing power distribution systems with increased size and complexity require more efficient data management systems and faster power flow solving algorithm. These challenges could be well solved by applying the graph data model (GDM) and graph computing technologies as the GDM provides more efficient data management methods and the graph computing is suitable for the node based parallel computing. This paper proposes a GDM based power distribution network modeling approach. Then a graph computing based parallel power flow algorithm and a graph visualization based power flow software have been developed based on it. The proposed parallel power flow algorithm is implemented on a graph database platform and the simulation results show that it is able to effectively reduce the computing time of the power flow with large test systems. Moreover, the power flow software is able to provide vivid data visualization and perform flexible data analysis and data management functions.
Index Terms—Graph computing, parallel power flow, graph visualization, backward forward sweep.
I. INTRODUCTION
With the fast-growing penetration of the renewable resources such as solar and wind generations, the power distribution systems are becoming more complex and it requires higher efficient data management methods and faster power flow solutions [1]. To achieve the goal of efficient data management and faster computation, new modeling methods for the power systems are required to break through the limits brought by traditional relational database. Graph computing is a new technology which constructs the power network from the viewpoint of a graph. The graph database also provides more efficient data management approaches by storing the information directly on vertices and edges. Its graph structure for connecting the vertices and edges also demands less processing time for retrieving data at any depth [2]. Different types of graph platforms have been developed in the market
such as Pregel [3], Neo4j [4], Giraph [5], TigerGraph [6], etc. for various applications.
Many research has been conducted in the field of graph computing and its application in power systems [7]-[11]. Reference [7] adopts a graph theory based network flow analysis in real time power system operations to improve network connectivity visualization. A graph based computational framework for coupled infrastructure networks’ optimization has been proposed in [8]. Paper [9] proposes a graph partition based mixed integer linear programming approach for power system islanding operation. A graph theory based optimal power quality monitors planning approach is presented in [10]. Reference [11] applies factor graphs for distributed power system state estimation.
However, very few study has been carried out in the field of graph computing based parallel power flow algorithm and the graph based visualization of power distribution networks. Thus, this paper will propose a graph data model (GDM) based power distribution network modeling approach which is able to provide fast parallel power flow platforms, efficient data management approaches, and graph based data visualizations. Base on the GDM of the power distribution network, a graph computing based parallel power flow algorithm has been proposed. The parallel power flow algorithm adopts the hierarchical group synchronization (HGS) parallel computing mechanism in bulk synchronous parallel (BSP) [12] computing model and it is able to effectively reduce the computing time for the power flow when dealing with large systems. Finally, a software package has been developed based on the GDM of the power distribution network. The graph visualization based software is able to provide vivid data visualization and perform flexible data analysis and data management functions such as voltage profile along the feeder, network reconfiguration, etc.
This study has made contributions in several major aspects by: (1) proposing a GDM based power distribution network modeling approach; (2) proposing a graph computing based
This work was supported by State Grid Corporation technology project 5455HJ180018.
parallel power flow algorithm; (3) developing a graph visualization based power flow software.
II. GRAPH BASED SYSTEM MODELING
A. Graph Database
In the context of graph computing, a graph database is defined as a database that applies semantic queries with vertices, edges and attributes to store data in a graph structure [13]. In a graph database, each vertex or edge represents an entity, and the relationships among these entities are represented by the graph structure. Thus, various kinds of scenarios such as power system network, transportation network, social network, etc. can be modeled into graphs by properly defining the vertices and edges.
Different from the relational database, data in a graph database is directly related and linked together stored in a graph as shown in Fig. 1. For relational databases, the data tables retrieved from the database are the same as they were first stored. Its data retrieving process needs complex join operations of the tables. However, for graph databases, the relationships among the data are constructed into a graph structure in the database. When the data is retrieved from the graph database, the relationships among the data are also retrieved with one simple operation. Thus, the data retrieval, data update and data communication efficiency is greatly enhanced by using the graph database.
Figure 1. Data storage in relational database and graph database.
The benefits of using graph database is obvious, especially in handling the topology traversal problems in systems with complex hierarchical structures. One of such applications is to analyze the voltage profile along a feeder in a power distribution system. This kind of analysis is very important to power distribution system operators/planners to perform necessary operations to improve the power quality and it can be efficiently realized in a graph database as shown in Fig. 2(c).
Fig. 2 illustrates that how a power distribution network is represented in both relational database and graph database. For the example of the IEEE 13 node test feeder as shown in Fig. 2(a), the relational database needs to build a node table with 13 records, a branch table with 12 records and a bridge table with 24 records while the graph database only needs to create 13 vertices and 12 edges. As shown in Fig. 2(b), a bridge table needs to be created to represent the relationship between the nodes and branches as the relational database does not support Many-to-Many relationship between tables. However, in the graph database, the connectivity information is directly
represented in the graph structure as shown in Fig. 2(c). To retrieve the voltage information along the feeder from node 650 to node 675 as an example, the graph database only needs to traverse the highlighted path by finding the father node of each traversed node in the path. It only takes 4 steps to locate all the nodes in the path from node 675. However, this process is very complex for relational database as shown in the highlighted path in Fig. 2(b). It takes 8 join operations to find all the nodes in the path. These join operations are both compute and memory intensive and its time cost grows exponentially with the increase of the system size. Thus, the graph database is very effective in the application of data management in power distribution systems.
646 645 632 633 634
650
692 675611 684
652
671
680
Figure 2. Power distribution network data representation in relational database
and graph database. (a) IEEE 13 node test feeder, (b) data expression in relational database, (c) data representation in graph database.
B. Modeling Power Distribution Networks with Graph Data Model
The power distribution network is essentially a graph, thus it can be easily modeled as a graph network in graph computing as shown in Fig. 3. To construct the GDM for a distribution network, we need to define the network as �(�, �) where � is the set of vertices and � denotes the set of edges. Intuitively, the load points, voltages regulators, switches and shunt capacitors can be modeled as vertices while the line segments and transformers can be modeled as edges. Both the vertices and edges have a set of attributes denoted as �(��, ��) . Thus, the data for power flow computing and the computed results can be stored in the graph database. As the graph database is already loaded in the computer memory, the graph computing does not need to waste time on communication between the power flow program and the database. Moreover, the data retrieval is more effective for graph database as it does not need the time consuming on join operations in the traditional relational database. For instance, all the data associated with the load point such as load demand, voltage, connected bus, etc. are stored in the vertices and it does not need to join the tables of load demand, network topology, and power flow results to retrieve the information for the load point. The graph structure also provides an opportunity for the application for the node synchronization based parallel computing. As a result, graph
database is able to benefit the power system by providing fast parallel power flow calculations together with efficient data management and data analysis.
n1
n2
n3
n4
S
Z12S12ΔV12
Z23S23ΔV23
Z24S24ΔV24
T
AttributesV1
S12(S1)
AttributesV2
S2
AttributesV4
S4
AttributesV3
S3
n1
n2
n3
n4
S
Z12S12ΔV12
Z23S23ΔV23
Z24S24ΔV24
T
AttributesV1
S12(S1)
AttributesV2
S2
AttributesV4
S4
AttributesV3
S3
Figure 3. Converting the distribution network into a computing graph.
III. GRAPH COMPUTING BASED PARALLEL POWER FLOW
A. Graph Computing Formulation with MapReduce based Parallel Mechanism
The graph computing is efficient in data retrieval and storage, flexible in modeling network connections and effective in model exploration. Graph computing is carried out directly on the vertices and edges of the graph. Thus, we need all the vertices and edges to be able to perform power flow calculations. Additional, we need to add virtual nodes S and T to indicate the start point and terminal point of a distribution network. These virtual nodes can also be viewed as the separation points for different sections of the power distribution network. First we define two relative concepts, child node set and father node set as shown in Fig. 4. Child node set is the nodes we are currently working on, while the father node set are preceding nodes directly connecting with child node set. For instance, if child node set is T0, then father node set is T1.
Figure 4. The mechanism of graph based parallel computing.
After obtaining the GDM, the parallel computing is able to be carried out on the graph as shown in Fig. 4. As shown in the figure, the generation of the GDM and resource allocation is in phase 1, the MapReduce [14] based parallel computing is in phase 2, and phase 3 is used to obtain the results. The BSP model has a Master to control the operation processes in each phase. First, the graph is divided into several parts by partition process and each partition is assigned to an independent worker in phase 1. Then, each worker divides the job into multiple maps in phase 2. There are two different parallel computing mechanism in this phase. The first one is the node
based parallel mechanism. As shown in the figure, each map has one father node and several child nodes. The power flow calculations for the child nodes are carried out in parallel and the result information is provided to their father node. This is also the process of MapReduce. The other parallel mechanism is the hierarchical parallel. We can observe from the figure that the maps at the same level are paralleled. Thus, the child nodes at the same level (for instance: nodes 4, 5, 6 in T0, and nodes 2, 3 in T1) are calculated in parallel.
B. Graph computing based three-phase unbalanced backward-forward power flow algorithm
Fig. 5 shows the process of MapReduce in the three phase unbalanced power flow calculation of radio distribution network with backward forward sweep [15]. In the backward forward sweep, the mapping process is to calculate the current node voltage and current, and the reduction process is to find out the father node voltage and current. Note that currents injecting to farther node set are calculated by aggregating branch currents which can be considered as Reduce phase. In forward sweep, node voltages are updated in a concurrent way as well while there is no current calculation involved in the sweep.
SabcV ][
*
][][][][][
nabc
nmabcmabcnabc
V
SIdVcI
mabcmabcnabc IbVaV ][][][][][
T
4
5
6
2
3
1S
T0
T1
T3
mabcnabcmabc IBVAV ][][][][][
mabcV ][nabcV ][
nabcI ][ mabcI ][
nS
Figure 5. The working principle of the graph computing based three-phase unbalanced power flow algorithm for power distribution networks.
IV. GRAPH BASED VISUALIZATION OF DISTRIBUTION
NETWORK POWER FLOW
In this section, we will introduce the graph computing based power flow software which is developed based on the proposed parallel power flow approach in Section III. We will also demonstrate its advantages for both data visualization and data analysis.
A. Software Architecture
The developed graph computing based fast distribution system power flow software is a web-based software. It includes 15 different web pages for case retrieval, power flow setting, results display, data analysis, graph management, etc. As shown in Fig. 6, the developed software adopts a frontend-backend design and it features a 3-level structure including frontend, backend and database. The frontend uses Angular as framework and leverage HTML, CSS and JavaScript to develop components for different applications. The backend is responsible for communications between sever, applications
and database.the graph database ucan realize fast data communication.
B.
directly stored in the graph which makes thperceptualGDM also makes the software very flexible in data analysis and data instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
case of thethe left. trace the path from the selected node to the substation node as highlighted the highlighted feederand load distrshown in the bar chart. The pie chart on the load
and database.the application program interfaces (graph database ucan realize fast data communication.
B. Graph
As directly stored in the graph which makes thperceptualGDM also makes the software very flexible in data analysis and data instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
Figure
Figcase of the graph structure of the test system is shown the left. trace the path from the selected node to the substation node as highlighted the highlighted feederand load distrshown in the bar chart. The pie chart on the load
and database.application program interfaces (
graph database ucan realize fast data communication.
Graph
As the softwaredirectly stored in the graph which makes thperceptualGDM also makes the software very flexible in data analysis and data instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
Figure
Fig. 7 shows the power flow results case of the
graph structure of the test system is shown the left. Atrace the path from the selected node to the substation node as highlighted the highlighted feederand load distrshown in the bar chart. The pie chart on the load (the upper pie chart)
and database.application program interfaces (
graph database ucan realize fast data communication.
Frontend
Backend
Database
Figure
Graph based
the softwaredirectly stored in the graph which makes thperceptually GDM also makes the software very flexible in data analysis and data managementinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
Figure
Figure 8. Network reconfiguration page for the distribution system
. 7 shows the power flow results the IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown After clicking a node on the
trace the path from the selected node to the substation node as highlighted in the graph. the highlighted feederand load distrshown in the bar chart. The pie chart on the
(the upper pie chart)
and database. It is based on a Flask framework and iapplication program interfaces (
graph database ucan realize fast data communication.
Frontend
Backend
Database
Figure
based
the softwaredirectly stored in the graph which makes th
intuitiveGDM also makes the software very flexible in data analysis
managementinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
Figure 7. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as in the graph.
the highlighted feederand load distribution along the feeder shown in the bar chart. The pie chart on the
(the upper pie chart)
It is based on a Flask framework and iapplication program interfaces (
graph database used in this software is TigerGraph [can realize fast data communication.
Frontend
Backend
Database
Figure 6. Architecture
based Visualization
the softwaredirectly stored in the graph which makes th
intuitiveGDM also makes the software very flexible in data analysis
managementinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as shown in Fig. 8.
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as in the graph.
the highlighted feederibution along the feeder
shown in the bar chart. The pie chart on the (the upper pie chart)
It is based on a Flask framework and iapplication program interfaces (
sed in this software is TigerGraph [can realize fast data
Component
Module
Architecture
Visualization
the software is designed based on GDM, the data is directly stored in the graph which makes th
intuitive from the perspective of a graph. GDM also makes the software very flexible in data analysis
managementinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as in the graph.
the highlighted feeder ibution along the feeder
shown in the bar chart. The pie chart on the (the upper pie chart)
It is based on a Flask framework and iapplication program interfaces (
sed in this software is TigerGraph [can realize fast data retriev
Component
http request
Module
Architecture
Visualization
is designed based on GDM, the data is directly stored in the graph which makes th
from the perspective of a graph. GDM also makes the software very flexible in data analysis
management as explained iinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as in the graph. Meanwhile, the voltage
lateralibution along the feeder
shown in the bar chart. The pie chart on the (the upper pie chart)
It is based on a Flask framework and iapplication program interfaces (
sed in this software is TigerGraph [retriev
Component Component
TigerGraph
http request
data
Architecture of the power flow software.
Visualization
is designed based on GDM, the data is directly stored in the graph which makes th
from the perspective of a graph. GDM also makes the software very flexible in data analysis
as explained iinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as Meanwhile, the voltage
lateral ibution along the feeder
shown in the bar chart. The pie chart on the (the upper pie chart) and loss
It is based on a Flask framework and iapplication program interfaces (
sed in this software is TigerGraph [retrieval
Angular
Component
TigerGraph
http request
Flask
Module
data
of the power flow software.
Visualization and its Advantages
is designed based on GDM, the data is directly stored in the graph which makes th
from the perspective of a graph. GDM also makes the software very flexible in data analysis
as explained iinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as Meanwhile, the voltage
will be shown in the line chart, ibution along the feeder
shown in the bar chart. The pie chart on the and loss
It is based on a Flask framework and iapplication program interfaces (APIs
sed in this software is TigerGraph [al, data update and data
Angular
Component
TigerGraph
data
Flask
Module
query
of the power flow software.
and its Advantages
is designed based on GDM, the data is directly stored in the graph which makes th
from the perspective of a graph. GDM also makes the software very flexible in data analysis
as explained iinstance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the
trace the path from the selected node to the substation node as Meanwhile, the voltage
will be shown in the line chart, ibution along the feeder
shown in the bar chart. The pie chart on the and loss
It is based on a Flask framework and iAPIs)
sed in this software is TigerGraph [, data update and data
Component
Module
query
of the power flow software.
and its Advantages
is designed based on GDM, the data is directly stored in the graph which makes th
from the perspective of a graph. GDM also makes the software very flexible in data analysis
as explained in Section instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results IEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown er clicking a node on the graph
trace the path from the selected node to the substation node as Meanwhile, the voltage
will be shown in the line chart, ibution along the feeder
shown in the bar chart. The pie chart on the and loss (the lower pie chart)
It is based on a Flask framework and i) for
sed in this software is TigerGraph [, data update and data
Component
Module
of the power flow software.
and its Advantages
is designed based on GDM, the data is directly stored in the graph which makes the data visualization
from the perspective of a graph. GDM also makes the software very flexible in data analysis
n Section instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
. 7 shows the power flow results analysisIEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown graph, the software will
trace the path from the selected node to the substation node as Meanwhile, the voltage
will be shown in the line chart, ibution along the feeder lateral
shown in the bar chart. The pie chart on the (the lower pie chart)
It is based on a Flask framework and ifor the
sed in this software is TigerGraph [, data update and data
Component
of the power flow software.
and its Advantages
is designed based on GDM, the data is e data visualization
from the perspective of a graph. GDM also makes the software very flexible in data analysis
n Section instance, the software is able to perform voltage the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
analysisIEEE 123 node test feeder. As shown in the figure,
graph structure of the test system is shown as, the software will
trace the path from the selected node to the substation node as Meanwhile, the voltage
will be shown in the line chart, lateral
shown in the bar chart. The pie chart on the right gives the (the lower pie chart)
It is based on a Flask framework and ithe frontend. The
sed in this software is TigerGraph [, data update and data
htt
p re
ques
t
of the power flow software.
and its Advantages
is designed based on GDM, the data is e data visualization
from the perspective of a graph. GDM also makes the software very flexible in data analysis
n Section instance, the software is able to perform voltage analysisthe feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results
Network reconfiguration page for the distribution system
analysis page for the IEEE 123 node test feeder. As shown in the figure,
as the graph on , the software will
trace the path from the selected node to the substation node as Meanwhile, the voltage profile
will be shown in the line chart, lateral will
right gives the (the lower pie chart)
It is based on a Flask framework and it provides frontend. The
sed in this software is TigerGraph [6], and it , data update and data
htt
p re
ques
t
is designed based on GDM, the data is e data visualization
from the perspective of a graph. GDM also makes the software very flexible in data analysis
n Section II.Aanalysis
the feeder as shown in Fig. 7 and network reconfiguration as
. The data analysis page for power flow results.
Network reconfiguration page for the distribution system
page for the IEEE 123 node test feeder. As shown in the figure,
the graph on , the software will
trace the path from the selected node to the substation node as profile
will be shown in the line chart, will also be
right gives the (the lower pie chart)
provides frontend. The
], and it , data update and data
is designed based on GDM, the data is e data visualization
from the perspective of a graph. The GDM also makes the software very flexible in data analysis
.A. For analysis along
the feeder as shown in Fig. 7 and network reconfiguration as
Network reconfiguration page for the distribution system.
page for the IEEE 123 node test feeder. As shown in the figure,
the graph on , the software will
trace the path from the selected node to the substation node as profile along
will be shown in the line chart, also be
right gives the (the lower pie chart)
provides frontend. The
], and it , data update and data
is designed based on GDM, the data is e data visualization
The GDM also makes the software very flexible in data analysis
. For along
the feeder as shown in Fig. 7 and network reconfiguration as
page for the IEEE 123 node test feeder. As shown in the figure,
the graph on , the software will
trace the path from the selected node to the substation node as along
will be shown in the line chart, also be
right gives the (the lower pie chart)
provides frontend. The
], and it , data update and data
is designed based on GDM, the data is e data visualization
The GDM also makes the software very flexible in data analysis
. For along
the feeder as shown in Fig. 7 and network reconfiguration as
page for the IEEE 123 node test feeder. As shown in the figure,
the graph on , the software will
trace the path from the selected node to the substation node as along
will be shown in the line chart, also be
right gives the (the lower pie chart)
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfigBTbe automatically updated to provide new graph power flow
10and degree of unbalanceselectthe graph will be shown in table the software very effective in providing visualization
A.
forwardpower flowbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain computest system[123 node test feeders.oon 6.8GHz
B.
increase of computing threadsfeeder
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfigBy changing the status of the switches inThe topology of the graph be automatically updated to provide new graph power flow
The node 10. Itand degree of unbalanceselectthe graph will be shown in table the software very effective in providing visualization
A. Simulation Environment
forwardpower flowbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain computest system[16]123 node test feeders.on a regular server, and the graph compution TigerGraph6.8. GHz
B. Simulation Results
Fig.increase of computing threadsfeeder
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph power flow
The node . It shows the magnitudes and phases of the node voltages
and degree of unbalanceselect athe graph will be shown in table will have a direct collection with the graph which makes the software very effective in providing visualization
Simulation Environment
In this paper, we developed a forwardpower flowbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain computest system
]. The larger test system123 node test feeders.
a regular server, and the graph computiTigerGraph. The server has
GHz with
Simulation Results
Fig. increase of computing threadsfeeder. It can
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph power flow
The node shows the magnitudes and phases of the node voltages
and degree of unbalancea certain
the graph will be shown in Fig.
will have a direct collection with the graph which makes the software very effective in providing visualization
Figure
Simulation Environment
In this paper, we developed a forward sweep algorithmpower flowbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain computest systems
. The larger test system123 node test feeders.
a regular server, and the graph computiTigerGraphThe server has with 64 GB memory.
Simulation Results
11 increase of computing threads
. It can
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph power flow calculation
The node voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
and degree of unbalancecertain
the graph will be Fig.
will have a direct collection with the graph which makes the software very effective in providing visualizations.
Figure
Figure
Simulation Environment
In this paper, we developed a sweep algorithm
power flow and itsbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
s are exactly based . The larger test system
123 node test feeders.a regular server, and the graph computiTigerGraphThe server has
64 GB memory.
Simulation Results
shows the increase of computing threads
. It can be observe
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph calculation
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
and degree of unbalancecertain node
the graph will be autom 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
s.
Figure 9. Power
Figure 10
Simulation Environment
In this paper, we developed a sweep algorithm
and itsbe tested. The computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
are exactly based . The larger test system
123 node test feeders.a regular server, and the graph computiTigerGraph v0.8.1 withThe server has
64 GB memory.
Simulation Results
shows the increase of computing threads
be observe
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph calculation
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
and degree of unbalancenode
autom10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
. Power
10. Power flow results of the node voltages.
V.
Simulation Environment
In this paper, we developed a sweep algorithm
and its performances be tested. The simulationcomputing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
are exactly based . The larger test system
123 node test feeders.a regular server, and the graph computi
0.8.1 withThe server has 2 CPUs × 6 Cores × 2 Threads @ 2.10
64 GB memory.
Simulation Results
shows the increase of computing threads
be observe
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfig
y changing the status of the switches inhe topology of the graph
be automatically updated to provide new graph calculations.
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
and degree of unbalance in the table on the right. in the
automatically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
. Power flow results of the node voltages.
. Power flow results of the node voltages.
V.
Simulation Environment
In this paper, we developed a sweep algorithm
performances simulation
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
are exactly based . The larger test system
123 node test feeders. The testing programa regular server, and the graph computi
0.8.1 with2 CPUs × 6 Cores × 2 Threads @ 2.10
64 GB memory.
Simulation Results
shows the trendincrease of computing threads
be observed
distributions in different phases of the selected feeder. Fig. 8 shows the network reconfiguration function of the software.
y changing the status of the switches inhe topology of the graph in the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right.in the table
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
C
Simulation Environment
In this paper, we developed a for
performances simulation
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
are exactly based . The larger test systems are combined with multiple IEEE
he testing programa regular server, and the graph computi
0.8.1 with the2 CPUs × 6 Cores × 2 Threads @ 2.10
64 GB memory.
trend increase of computing threads
d from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
y changing the status of the switches inin the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right.table
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
CASE
In this paper, we developed a unbalanced distribution network
performances simulation will include comparing the
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
are exactly based theare combined with multiple IEEE
he testing programa regular server, and the graph computi
the 2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the increase of computing threads
from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
y changing the status of the switches inin the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right.table, the corresponding
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
ASE STUDIES
In this paper, we developed a unbalanced distribution network
performances under will include comparing the
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
the IEEE 123 node test feederare combined with multiple IEEE
he testing programa regular server, and the graph computi
operation2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the for
from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
y changing the status of the switches inin the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right., the corresponding
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
TUDIES
In this paper, we developed a unbalanced distribution network
under will include comparing the
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
IEEE 123 node test feederare combined with multiple IEEE
he testing programa regular server, and the graph computi
operation2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the for the IEEE 123 node t
from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
y changing the status of the switches in thein the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right., the corresponding
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
TUDIES
In this paper, we developed a paralleled unbalanced distribution network
under various will include comparing the
computing time for certain test systemcomputing threads as well as comparing the computing time for different size systems with a certain compu
IEEE 123 node test feederare combined with multiple IEEE
he testing programa regular server, and the graph computing platform is based
operation 2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node t
from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
the graphin the middle shown in Fig. 8
be automatically updated to provide new graph
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right., the corresponding
atically zoomed in and 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing
flow results of the node voltages.
. Power flow results of the node voltages.
paralleled unbalanced distribution network
various will include comparing the
computing time for certain test systems computing threads as well as comparing the computing time for different size systems with a certain computing thread.
IEEE 123 node test feederare combined with multiple IEEE
he testing programs are implemented ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node t
from the figure that
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
graphin the middle shown in Fig. 8
be automatically updated to provide new graph structure
voltage results are displayed in Fig.shows the magnitudes and phases of the node voltages
in the table on the right. When users , the corresponding
atically zoomed in and highlighted as 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes the software very effective in providing user
flow results of the node voltages.
. Power flow results of the node voltages.
paralleled unbalanced distribution network
various scenarioswill include comparing the
with different computing threads as well as comparing the computing time
ting thread. IEEE 123 node test feeder
are combined with multiple IEEE are implemented
ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node t
from the figure that the
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
graph on the left. in the middle shown in Fig. 8
structure
voltage results are displayed in Fig. 9 and Fig.shows the magnitudes and phases of the node voltages
When users , the corresponding
highlighted as 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes user-
flow results of the node voltages.
. Power flow results of the node voltages.
paralleled backunbalanced distribution network
scenarioswill include comparing the
with different computing threads as well as comparing the computing time
ting thread. IEEE 123 node test feeder
are combined with multiple IEEE are implemented
ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node t
the computing
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
on the left. in the middle shown in Fig. 8
structure
and Fig.shows the magnitudes and phases of the node voltages
When users , the corresponding node
highlighted as 10. As the data is stored in the graph, the
will have a direct collection with the graph which makes -friendly
backwardunbalanced distribution network
scenarioswill include comparing the
with different computing threads as well as comparing the computing time
ting thread. IEEE 123 node test feeder
are combined with multiple IEEE are implemented
ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node t
computing
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
on the left. will
structure for
and Fig.shows the magnitudes and phases of the node voltages
When users node in
highlighted as 10. As the data is stored in the graph, the data
will have a direct collection with the graph which makes friendly
wardunbalanced distribution network
scenarios will will include comparing the
with different computing threads as well as comparing the computing time
ting thread. The IEEE 123 node test feeder
are combined with multiple IEEE are implemented
ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the the IEEE 123 node test
computing
distributions in different phases of the selected feeder. Fig. 8 uration function of the software.
on the left. will for
and Fig. shows the magnitudes and phases of the node voltages
When users in
highlighted as data
will have a direct collection with the graph which makes friendly
ward-unbalanced distribution network
will will include comparing the
with different computing threads as well as comparing the computing time
The IEEE 123 node test feeder
are combined with multiple IEEE are implemented
ng platform is based system of CentOS
2 CPUs × 6 Cores × 2 Threads @ 2.10
of the computing time with the est
computing
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads
Figure
Figure
test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing dynsizesequential computithe increase of the system sizeparallel computing, much slower the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads
Figure
Figure
Figure
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing dynamics of the computing time size when using different computing threadssequential computithe increase of the system sizeparallel computing, much slower the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads
Figure 11. The performance of the parallel computing algorithm for IEEE 123
Figure
Figure 13.
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing
amics of the computing time when using different computing threads
sequential computithe increase of the system sizeparallel computing, much slower the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads
. The performance of the parallel computing algorithm for IEEE 123
12. The computing performance of the algorithm with different computing threads for dif
. The computing performance of the algorithm with different size of
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing
amics of the computing time when using different computing threads
sequential computithe increase of the system sizeparallel computing, much slower the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different computing threads for dif
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing
amics of the computing time when using different computing threads
sequential computithe increase of the system sizeparallel computing, much slower withthe computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is notcomputing threads reach
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different computing threads for dif
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing
amics of the computing time when using different computing threads
sequential computing, tthe increase of the system sizeparallel computing,
with the increase of the system size. the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
reach
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different computing threads for dif
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing
amics of the computing time when using different computing threads
ng, tthe increase of the system sizeparallel computing, the increase rate of the computing time is
the increase of the system size. the computing speed of the algorithm is much faster
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
reach a certain number.
. The performance of the parallel computing algorithm for IEEE 123 node test feeder.
The computing performance of the algorithm with different computing threads for dif
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more significant when dealing with
amics of the computing time when using different computing threads
ng, the computing time increasesthe increase of the system size
the increase rate of the computing time is the increase of the system size.
the computing speed of the algorithm is much faster tha
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
a certain number.
. The performance of the parallel computing algorithm for IEEE 123 node test feeder.
The computing performance of the algorithm with different computing threads for dif
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
with large systems. Fig. 13 amics of the computing time when using different computing threads
he computing time increasesthe increase of the system size
the increase rate of the computing time is the increase of the system size.
the computing speed of the than the sequential computing for the
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
a certain number.
. The performance of the parallel computing algorithm for IEEE 123 node test feeder.
The computing performance of the algorithm with different computing threads for different size of
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power for larger test systems are shown 12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 amics of the computing time when using different computing threads
he computing time increasesthe increase of the system size. However, for the case of
the increase rate of the computing time is the increase of the system size.
the computing speed of the prthe sequential computing for the
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
a certain number.
. The performance of the parallel computing algorithm for IEEE 123 node test feeder.
The computing performance of the algorithm with different ferent size of
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 amics of the computing time with the increase of system when using different computing threads
he computing time increases. However, for the case of
the increase rate of the computing time is the increase of the system size.
proposed parallel computing the sequential computing for the
time decreases fast with the increase of the computing threads at the beginning while the decrease is not
a certain number.
. The performance of the parallel computing algorithm for IEEE 123 node test feeder.
The computing performance of the algorithm with different ferent size of
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 with the increase of system
when using different computing threadshe computing time increases
. However, for the case of the increase rate of the computing time is
the increase of the system size. oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads at the beginning while the decrease is not significant when the
a certain number.
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different ferent size of the
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 with the increase of system
when using different computing threadshe computing time increases
. However, for the case of the increase rate of the computing time is
the increase of the system size. oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different systems
The computing performance of the algorithm with different size of the systems using different computing threads
The large systems combined with multiple IEEE 123 node test feeders. The performances of the power flow algorithm
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 with the increase of system
when using different computing threads. For the case of he computing time increases
. However, for the case of the increase rate of the computing time is
the increase of the system size. oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different systems.
The computing performance of the algorithm with different size of the systems using different computing threads.
The large systems combined with multiple IEEE 123 node flow algorithm
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 with the increase of system
For the case of he computing time increases
. However, for the case of the increase rate of the computing time is
the increase of the system size. As a result, oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different
The computing performance of the algorithm with different size of
The large systems combined with multiple IEEE 123 node flow algorithm
in Fig. 12 and Fig. 1312 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
large systems. Fig. 13 shows the with the increase of system
For the case of he computing time increases fast
. However, for the case of the increase rate of the computing time is
As a result, oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different
The computing performance of the algorithm with different size of
The large systems combined with multiple IEEE 123 node flow algorithm
in Fig. 12 and Fig. 13. Fig. 12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computithreads and the computing speed improvement is more
shows the with the increase of system
For the case of fast with
. However, for the case of the increase rate of the computing time is
As a result, oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The computing performance of the algorithm with different
The computing performance of the algorithm with different size of
The large systems combined with multiple IEEE 123 node flow algorithm
Fig. 12 shows the computing time of the proposed algorithm with different computing threads for different size of the system. The computing time decreases with the increase of computing threads and the computing speed improvement is more
shows the with the increase of system
For the case of with
. However, for the case of the increase rate of the computing time is
As a result, oposed parallel computing
the sequential computing for the
time decreases fast with the increase of the computing threads significant when the
. The performance of the parallel computing algorithm for IEEE 123
The large systems combined with multiple IEEE 123 node flow algorithm
Fig. 12 shows the computing time of the proposed algorithm with different computing threads for different size of the system.
ng threads and the computing speed improvement is more
shows the with the increase of system
For the case of
. However, for the case of the increase rate of the computing time is
As a result, oposed parallel computing
the sequential computing for the
cresults that our proposed parallel power flow algorithm is very effective when dealing with large
GDM which provideefficient data Based on power flow algorithm and a power flow software have been developed. effectively reduceprovidevisualization
the
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
case of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
ThiGDM which provideefficient data Based on power flow algorithm and a power flow software have been developed. effectively reduceprovidevisualization
Thethe coding work at the early stage of this research.
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
This paper modelsGDM which provideefficient data Based on power flow algorithm and a power flow software have been developed. effectively reduceprovidesvisualization
The coding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,” Energy
C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in Conf.
G. Malewicz processing,” in2010, pp. 135
“Neo4jhttp://neo4j.com/
C. Avery, “Giraph: LargehadoopAvailable:http://www.slideshare.net/averyching/20110628giraphhadoop-summit
“TigerGraph: The first native parallel graphhttps://www.tigergraph.com/.
T. Werhoconnectivity monitoring using a graph theory network flow algorithmIEEE Trans. Power
J. Jalving, S. Abhyankar, K. Kim, M. graphof coupled infrastructure networks," in& D
T. Ding, K. Sun, C. Huang, Z. Bie, and F. Li, programmingoperation considering network connectivitypress
D. J. Won and S. quality monitors considering system topologyDelivery,
P. Chavali and A. Nehorai, using factor graphspp. 2864
Leslie G. Valiant, Communications of the ACM
G. Ravikumar andoriented graph database framework for power systemsPower Systems,
J. Dean and S. Ghemawat, large clustersJan. 2008.
W. H.2002.
“Distribution Test Fieee.org/
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
s paper modelsGDM which provideefficient data Based on power flow algorithm and a power flow software have been developed. effectively reduce
s efficientvisualization
authors are gracoding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,” Energy, vol. 8, no. 4, pp. 1351C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in Conf.,2010, Art. no. 42G. Malewicz processing,” in2010, pp. 135“Neo4j: An open source graph databasehttp://neo4j.com/C. Avery, “Giraph: Largehadoop,” in Available:http://www.slideshare.net/averyching/20110628giraphhadoo
summitTigerGraph: The first native parallel graph
https://www.tigergraph.com/.Werho
connectivity monitoring using a graph theory network flow algorithmIEEE Trans. Power J. Jalving, S. Abhyankar, K. Kim, M. graph-based computational framework for simulation and optimisation of coupled infrastructure networks," in& DistributionT. Ding, K. Sun, C. Huang, Z. Bie, and F. Li, programmingoperation considering network connectivitypress. D. J. Won and S. quality monitors considering system topologyDelivery,P. Chavali and A. Nehorai, using factor graphspp. 2864Leslie G. Valiant, Communications of the ACMG. Ravikumar andoriented graph database framework for power systemsPower Systems,J. Dean and S. Ghemawat, large clustersJan. 2008.
H. Kersting,2002. Distribution Test F
ieee.org/
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
s paper modelsGDM which provideefficient data managementBased on graph database and graph computingpower flow algorithm and a power flow software have been developed. The simulation reseffectively reduce
efficientvisualizations.
authors are gracoding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
, vol. 8, no. 4, pp. 1351C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in
2010, Art. no. 42G. Malewicz processing,” in2010, pp. 135
: An open source graph databasehttp://neo4j.com/C. Avery, “Giraph: Large
,” in Available:http://www.slideshare.net/averyching/20110628giraphhadoo
summit/. TigerGraph: The first native parallel graph
https://www.tigergraph.com/.Werho, V. Vittal, S. Kolluri, and S. M. Wong,
connectivity monitoring using a graph theory network flow algorithmIEEE Trans. Power J. Jalving, S. Abhyankar, K. Kim, M.
based computational framework for simulation and optimisation of coupled infrastructure networks," in
istributionT. Ding, K. Sun, C. Huang, Z. Bie, and F. Li, programmingoperation considering network connectivity
D. J. Won and S. quality monitors considering system topologyDelivery, vol. 23 no. 1, pp. 288P. Chavali and A. Nehorai, using factor graphspp. 2864-2876, Jun. 2015.Leslie G. Valiant, Communications of the ACMG. Ravikumar andoriented graph database framework for power systemsPower Systems,J. Dean and S. Ghemawat, large clustersJan. 2008.
Kersting,
Distribution Test Fieee.org/ soc/pes/dsacom/testfeeders/
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
s paper modelsGDM which provide
managementgraph database and graph computing
power flow algorithm and a power flow software have been The simulation res
effectively reduceefficient data management services and
authors are gracoding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
, vol. 8, no. 4, pp. 1351C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in
2010, Art. no. 42G. Malewicz et al.processing,” in Proc. 2010 ACM SIGMOD Int. Conf. Manage. data2010, pp. 135–146
: An open source graph databasehttp://neo4j.com/.C. Avery, “Giraph: Large
,” in Proc. Hadoop SummitAvailable:http://www.slideshare.net/averyching/20110628giraphhadoo
TigerGraph: The first native parallel graphhttps://www.tigergraph.com/.
, V. Vittal, S. Kolluri, and S. M. Wong, connectivity monitoring using a graph theory network flow algorithmIEEE Trans. Power J. Jalving, S. Abhyankar, K. Kim, M.
based computational framework for simulation and optimisation of coupled infrastructure networks," in
istribution, vT. Ding, K. Sun, C. Huang, Z. Bie, and F. Li, programming-based splitting strategies for power system islanding operation considering network connectivity
D. J. Won and S. quality monitors considering system topology
vol. 23 no. 1, pp. 288P. Chavali and A. Nehorai, using factor graphs
2876, Jun. 2015.Leslie G. Valiant, Communications of the ACMG. Ravikumar andoriented graph database framework for power systemsPower Systems, vol. 32, no. 4, pp. 2560J. Dean and S. Ghemawat, large clusters”, Communication of the ACM
Kersting, Distribution system modeling and analysis
Distribution Test Fsoc/pes/dsacom/testfeeders/
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
VI.
s paper modelsGDM which provide
managementgraph database and graph computing
power flow algorithm and a power flow software have been The simulation res
effectively reduces the computing time of power flow and data management services and
Aauthors are gra
coding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
, vol. 8, no. 4, pp. 1351C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in
Proc. Hadoop SummitAvailable:http://www.slideshare.net/averyching/20110628giraphhadoo
TigerGraph: The first native parallel graphhttps://www.tigergraph.com/.
, V. Vittal, S. Kolluri, and S. M. Wong, connectivity monitoring using a graph theory network flow algorithmIEEE Trans. Power Systems,J. Jalving, S. Abhyankar, K. Kim, M.
based computational framework for simulation and optimisation of coupled infrastructure networks," in
, vol. 11, no. 12, pp. 3163T. Ding, K. Sun, C. Huang, Z. Bie, and F. Li,
based splitting strategies for power system islanding operation considering network connectivity
D. J. Won and S. II quality monitors considering system topology
vol. 23 no. 1, pp. 288P. Chavali and A. Nehorai, using factor graphs”,
2876, Jun. 2015.Leslie G. Valiant, “Communications of the ACMG. Ravikumar and S. A. Khaparde, oriented graph database framework for power systems
vol. 32, no. 4, pp. 2560J. Dean and S. Ghemawat,
Communication of the ACM
Distribution system modeling and analysis
Distribution Test Feederssoc/pes/dsacom/testfeeders/
ase of larger test systems.results that our proposed parallel power flow algorithm is very effective when dealing with large
VI.
s paper models the power GDM which provides fast parallel power flow platform,
managementgraph database and graph computing
power flow algorithm and a power flow software have been The simulation res
the computing time of power flow and data management services and
ACKNOWLED
authors are grateful to Dr. Wenlei Bai for hiscoding work at the early stage of this research.
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
, vol. 8, no. 4, pp. 1351C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in
2010, Art. no. 42. et al., “Pregel: A system for largeProc. 2010 ACM SIGMOD Int. Conf. Manage. data
TigerGraph: The first native parallel graphhttps://www.tigergraph.com/.
, V. Vittal, S. Kolluri, and S. M. Wong, connectivity monitoring using a graph theory network flow algorithm
Systems,J. Jalving, S. Abhyankar, K. Kim, M.
based computational framework for simulation and optimisation of coupled infrastructure networks," in
ol. 11, no. 12, pp. 3163T. Ding, K. Sun, C. Huang, Z. Bie, and F. Li,
based splitting strategies for power system islanding operation considering network connectivity
Moon, quality monitors considering system topology
vol. 23 no. 1, pp. 288P. Chavali and A. Nehorai,
, IEEE Trans. Signal Processing,2876, Jun. 2015.
“A bridging model for parallel computationCommunications of the ACM
S. A. Khaparde, oriented graph database framework for power systems
vol. 32, no. 4, pp. 2560J. Dean and S. Ghemawat,
Communication of the ACM
Distribution system modeling and analysis
eederssoc/pes/dsacom/testfeeders/
ase of larger test systems. Thus, iresults that our proposed parallel power flow algorithm is very effective when dealing with large
C
the power fast parallel power flow platform,
management, and graph database and graph computing
power flow algorithm and a power flow software have been The simulation res
the computing time of power flow and data management services and
CKNOWLED
teful to Dr. Wenlei Bai for hiscoding work at the early stage of this research.
REFERENCES
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
, vol. 8, no. 4, pp. 1351-C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in
, “Pregel: A system for largeProc. 2010 ACM SIGMOD Int. Conf. Manage. data
: An open source graph database
C. Avery, “Giraph: Large-scale graph processing infrastructureProc. Hadoop Summit
TigerGraph: The first native parallel graphhttps://www.tigergraph.com/.
, V. Vittal, S. Kolluri, and S. M. Wong, connectivity monitoring using a graph theory network flow algorithm
Systems, vol. 31, no. 6, pp. 4945J. Jalving, S. Abhyankar, K. Kim, M.
based computational framework for simulation and optimisation of coupled infrastructure networks," in
ol. 11, no. 12, pp. 3163T. Ding, K. Sun, C. Huang, Z. Bie, and F. Li,
based splitting strategies for power system islanding operation considering network connectivity
Moon, “Optimal number and locations of power quality monitors considering system topology
vol. 23 no. 1, pp. 288P. Chavali and A. Nehorai, “Distributed power
IEEE Trans. Signal Processing,2876, Jun. 2015.
A bridging model for parallel computationCommunications of the ACM, v
S. A. Khaparde, oriented graph database framework for power systems
vol. 32, no. 4, pp. 2560J. Dean and S. Ghemawat, “MapReduce: simplified data processing on
Communication of the ACM
Distribution system modeling and analysis
eeders”, 2017. soc/pes/dsacom/testfeeders/
Thus, iresults that our proposed parallel power flow algorithm is very effective when dealing with large
CONCLUSIONS
the power fast parallel power flow platform, , and graph based
graph database and graph computingpower flow algorithm and a power flow software have been
The simulation results show that the software the computing time of power flow and
data management services and
CKNOWLED
teful to Dr. Wenlei Bai for hiscoding work at the early stage of this research.
EFERENCES
C. Yuan, M. Illindala, A. Khalsa, “Coresource planning in community microgrids,”
-1360, Oct. 2017.C. Vicknair, M. Macias, Z. Zhao, X. Nan, Y. Chen, and D. Wilkins,comparison of a graph database and a relational database: A dataprovenance perspective,” in Proc. 48th Annu. Sout
, “Pregel: A system for largeProc. 2010 ACM SIGMOD Int. Conf. Manage. data