Load Balancing for Distributed and Integrated Power Systems using Grid computing
Active In SP
Joined: Mar 2010
09-04-2010, 09:11 PM
Load Balancing for Distributed and Integrated Power Systems using Grid computing.doc (Size: 383 KB / Downloads: 158)
Grid computing is a software technology, which uses the spare computing resources. This grid computing is now seen as a powerful and important tool for the power system applications. These days the power system operation and control involves large data-intensive, time-intensive applications. Therefore it can get a better solution by using this new technology. Also the power system is moving towards the renewable energy sources. In this case it helps them in providing cheap and efficient electrical energy supply.
The demand for conservation of the current conventional energy has been increased due to the growing concern on the deteriorating condition of our environment. The control and operations of modern power systems are becoming data-intensive, information-intensive, communication-intensive and computation-intensive due to much of the reliance of these power systems on computerized communications and control.
Therefore, since Grid Computing is considered as an inexpensive means for super-computing for dealing with heterogeneous and distributed computing. Grid Computing is the best option in providing the necessary storage media for the huge data, intensive computational processes and also the required computing resources. Thus, grid computing can be utilized to fulfill the requirement for efficient load balancing.
1.1 Background Theory
In a distributed and integrated power systems, it is vital to ensure that each power source (generator, wind turbine, etc) is working within its allowed parameters. These parameters are normally based on the current power load that are sometimes have been forecasted within regular intervals (weekly, monthly or yearly). Anyway, these non real-time forecasts have their drawbacks and may not be supplying correct information when any of these events occur:
1) Sudden failure of any of the power sources.
2) Unexpected increase or decrease of power demand within a short timeframe.
During the occurrence of any of these events, power load balancing within these power sources are required within immediate timeframe in order to ensure that there will be no power interruptions.
1.1.1 Load Balancing
In computer network, load balancing is a technique to distribute workload evenly across two or more computers, network links, CPUs, hard drives, or other resources, in order to get optimal resource utilization, maximize throughput, minimize response time, and avoid overload. For parallel applications, load balancing attempts to distribute the computation load across multiple processors or machines as evenly as possible with objective to improve performance. Generally, a load balancing scheme consists of 3 phases: information collection, decision making and data migration. During the information collection phase, the information of workload distribution and the state of computing environment is gathered. The decision making phase focuses on calculating an optimal data distribution, while the data migration phase transfers the excess amount of workload from overloaded processors to under loaded ones.
In power applications, the transmission system provides for base load and peak load capability, with safety and fault tolerance margins. The peak load times vary by region largely due to the industry mix. In very hot and very cold climates home air conditioning and heating loads have an effect on the overall load. They are typically highest in the late afternoon in the hottest part of the year and in mid-mornings and mid-evenings in the coldest part of the year. This makes the power requirements vary by the season and the time of day. Distribution system designs always take the base load and the peak load into consideration.
The transmission system usually does not have a large buffering capability to match the loads with the generation. Thus generation has to be kept matched to the load, to prevent overloading failures of the generation equipment.
Multiple sources and loads can be connected to the transmission system and they must be controlled to provide orderly transfer of power. In centralized power generation, only local control of generation is necessary, and it involves synchronization of the generation units, to prevent large transients and overload conditions. In distributed power generation the generators are geographically distributed and the process to bring them online and offline must be carefully controlled.
CONCEPT AND DESIGN
Grid computing is the combination of computer resources from multiple administrative domains applied to a common task, usually to a scientific, technical or business problem that requires a great number of computer processing cycles or the need to process large amounts of data.
One of the main strategies of grid computing is using software to divide the pieces of a program among several computers, sometimes up to many thousands. Grid computing is distributed, large-scale cluster computing, as well as a form of network-distributed parallel processing
2.1 PROPOSED MODEL FOR INTEGRATING GRID COMPUTING WITH POWER APPLICATIONS
In this section we focus on utilizing Grid Computing for real-time power load balancing operations that are based on the model for integrating Grid Computing with power applications as illustrated in Figure 1.
From Figure 1, further descriptions of the model are as follows:
a) The Grid Computing model contains 2 main servers (server01 and server02) and 8 computing resources (client01 â€œ client08). Server01 stores the list for all power system applications and server02 acts as a resource broker.
b) The Grid Computing is integrated with the power systems applications via an application interface.
c) The power applications interact with server01 to obtain the current list of all power applications connected to the Grid Computing and therefore, these power applications will interact with each other for real-time data transfer and load-balancing operations as depicted with the white arrows.
d) Server01 interacts with server02 to find available Grid Computing resources meeting certain specifications in order to perform real-time load forecasting, as depicted with the black arrows.
2.1.1 Power World Simulator
Power World Simulator simulation software has been used for simulating the power applications in this model. Power World Simulator is an interactive graphical-based power application for simulating high voltage power systems that runs only on MS Windows-based OS. It is a power simulation package designed from ground up to user-friendly and highly interactive. This simulator has the power for engineering analysis and its interactive and graphical that can be used to explain the power system operation. It is a standalone package for power flow. This simulator allows the users to visualize the system through the use of full colour animated one line diagram with full zooming facility. Thus, this simulator provides a convenient medium for simulating the evolution of power system over time.
2.1.2 Globus Toolkit version 4.0.2 (GT4)
For the Grid Computing middleware, Globus Toolkit version 4.0.2 (GT4) is selected since
a) It is an open source software for development issue.
b) It is considered the de facto standard of Grid Computing.
Grid Resource Management involves coordination of a number of components, including resource registries, factories, discovery, monitoring allocation and data access. The globus toolkit includes a set of components to help users have a standard set of interfaces for the coordination of the above activities. The Globus Toolkit provides a number of components for doing data management. The data management services provide standard means for helping to manage the grid computing environment. The GridFTP is a standard extension to normal FTP(File Transfer protocol) that works with grid computing data requirements. This is a high performance, secure, reliable, data transfer protocol that is optimized for high bandwidth across wide area networks. This is a standard that provides security, parallel transfer capabilities, and channel reusability and efficient transfer of bulk data. The Globus Toolkit provides the most commonly used implementation of this protocol.
The following section describes the main tools used in the development of integrating Grid Computing with the power applications.
Wine (the short form for Wine Is Not an Emulator) allows MS Windows applications to be executed on Linux OS. Using Wine, Power World Simulator software has been successfully installed and executed on Linux machines.
2.2.2. Java Robot Class
Since Power World Simulator is a graphical-based simulation, it is not possible to save the simulation data relating to the generatorsâ„¢ performances on a continuous basis. Therefore the Java Robot Class has been used to control, save the simulation data and also to load new simulation data. This class is used to generate native system input events for the purposes of test automation, self-running demos, and other applications where control of the mouse and keyboard is needed. In order to achieve this purpose, we have to:
a) Capture the image screen for each mouse movement in order to save the simulations data and loading new data.
b) Identify the exact x and y coordinates of where the mouse needs to be moved and clicked.
Using this class to generate input events differs from posting events to the AWT event queue or AWT components in that the events are generated in the platform's native input queue. For example, Robot.mouseMove will actually move the mouse cursor instead of just generating mouse move events. Basically, a Robot object makes it possible for our program to temporarily take over control of the mouse and the keyboard.
2.2.3. Karajan Workflow
Karajan is a workflow language and workflow engine. Karajan is a Grid parallel task management language and an execution engine. It aims to provide the scientific community with an easy-to-use tool to define and manage complex jobs on computational Grids, while keeping scalability and offering some advanced features, such as failure handling, check pointing, dynamic execution, and distributed execution. Karajan also provides a web-service interface, which can be used to remotely control and monitor the execution. An instance of the Karajan service can be used to execute multiple specifications at one time. The Karajan specifications are written in an XML based language. XML has the advantage of a very strict and well defined structure.
Through Karajan, grid users can easily define set of complex tasks by creating the appropriate XML file. Karajan Workflow has a number of libraries and one of these libraries is the System Library. The System Library contains the elements <parallelFor> and <parallel> that allow parallel execution of tasks as illustrated in Figure 2.
Thus, by using Karajan Workflow, we can submit jobs and transfer files in parallel.
ANALYSIS AND APPLICATIONS
3.1 DEMONSTRATION OF REAL-TIME LOAD BALANCING USING POWER APPLICATION GRID INTERFACE (PAGI)
To demonstrate the functionalities of the interface between the Power World Simulator and the Grid Computing, a Java-based GUI program called ËœPower Application Grid Interfaceâ„¢ or PAGI is developed as illustrated in Figure 3.
From Figure 3, PAGI is divided into two parts:
a) Set appropriate settings for running the Power World Simulator.
b) Displaying the data received from other power applications (simulators).
During the execution of PAGI, the simulations data (for e.g. generatorsâ„¢ loads) are saved to a fixed filename Ëœdata.auxâ„¢.
3.1.1 Real-time data transfer
PAGI performs real-time data transfers by using the Globus GridFTP function to all other simulators based on the list obtained from server01. This list is continuously obtained from server01 in order to ensure that the simulation data are transferred to all power applications connected to the Grid Computing. PAGI also displays the data received from all other simulators in order to view the current load usage at other simulators (or power plantsâ„¢ sites).
3.1.2 Real-time Load balancing
For the mechanism of real-time load balancing operations, PAGI was developed based on the following process flow as shown in Figure 4.
Real-time load balancing operations are basically executed by means of sending a specific command which is known as the Ëœchange commandâ„¢ from one simulator to the other simulator(s). Here, we substituted the term Ëœsimulatorsâ„¢ with Ëœsitesâ„¢ that has the meaning of where the power plants are actually located.
From Figure 4, PAGI also allow the user to specify the settings for the load balancing operations:
a) Minimum load percentage before a Ëœchange commandâ„¢ is sent to other sites in order to balance the load balance operations.
b) Set the specific sites that shall receive the Ëœchange commandâ„¢ or let PAGI decide which sites automatically receive Ëœchange commandâ„¢.
To calculate how much load is required to be added at sites receiving the Ëœchange commandâ„¢, the following formula is used:
Pa: Additional power load
Pc: Current power load
Pm: Maximum power load allowable as set by the user
Ef: Expansion factor
Pg: The maximum power of the generator at the origin site
The expansion factor, Ef, is included in order to increase the additional power load to a certain factor. This will ensure that the power generated at the recipient sites is slightly more than the actual power need.
3.1.3 Load Forecasting
There are important differences in load between weekdays and weekends. The load on different weekdays also can behave differently. For example, Mondays and Fridays being adjacent to weekends, may have structurally different loads than Tuesday through Thursday. This is particularly true during the summer time. Weather conditions influence the load. In fact, forecasted weather parameters are the most important factors in load forecasts. Various weather variables are considered for load forecasting. Temperature and humidity are the most commonly used load predictors.
Since all power application simulations are submitting their data to server01, a mechanism that utilizes Grid Computing resources has been developed in order to perform load forecast calculations for each simulation. The pseudo code for this mechanism is as follows:
Using the resource broker function at server02, find available Grid Computing resources.
For each simulation data, the following tasks are executed in parallel
1. Split the data based on the number of available resources.
2. Submit the split data to each resource.
3. Submit jobs in parallel to perform load forecast on all available resources.
4. Return the forecast value to relevant simulation.
5. Submit a job the relevant simulation to analyze the forecast value and take appropriate action if required.
The ËœFor...Loopâ„¢ stated above is executed by using Karajan Workflow that allows parallel jobs submissions and data transfers. For task no. 5, the power application simulations shall send the Ëœchange commandâ„¢ to other simulation(s) if the forecast value reaches or exceeds the working limit of the power systems. This particular task is included in order to create decentralized distributed power systems, moving away from a traditional centralized control management for power systems.
3.2 GRID COMPUTINGâ„¢S PRIORITY AND ADVANTAGES IN POWER UTILITIES COMPANIES
In order to find the level of priority of grid computing technology to the power utilities companies, institutions and academies, first it should identify what will be the benefits of this technology to them. These benefits are as follows:
a) Real-time data transmission among all power plants which dissolves the function of a centralized control centre.
b) Real-time load balancing process among all power plants to ensure that the power load are evenly distributed. In crucial situations such as power outages at one of the power plants, real-time load balancing could be performed to minimize the damage caused by the power outage.
c) Shorter time required for load forecast calculation which allows the power plant to react faster for the near future changes in power load demands.
COMPARISON AND EVALUATION
4.1 RESULTS AND ANALYSIS
In analyzing the performance for real-time data transfer, the time taken for submitting simulations data to a different number of sites by using parallel mode (based on Karajan Workflow) is compared with the sequential mode. The results of the comparisons are as shown in Figure 5.
From Figure 5, it is clear that the time taken for submitting data in sequential mode takes longer time as compared to the parallel mode. Also, in the parallel mode it could be recognized that time curve doesnâ„¢t show smooth proportional flew, the possible reasons for that are:
a) The start-up time required during the execution of the workflow.
b) Full bandwidth utilization for each data transfer.
For the real-time load balancing operations, it has been successfully tested on sending the Ëœchange commandâ„¢ to other simulators (sites) receiving the Ëœchange commandâ„¢. These recipient sites will immediately load the new simulation data received from the sender that will increase the power load at those locations. Also, utilizing all the available Grid Computing resources was successfully done, in order to perform load forecasting operations. The forecast values were then returned to each relevant power systems where, if these values exceed a specific limit, decentralized load balancing operations are then executed. The load forecasting performance on different number of resources calculations are analyzed and the results are as illustrated in Figure 6.
Figure6 illustrates that the time taken for load forecasting reduces quite significantly when more Grid Computing resources are selected. Thus, by having the required resources, hourly load forecasting can be calculated where the forecast value shall be return to each power system for decentralized load balancing operations.
4.2 TRADITIONAL AND GRID METHODOLOGY
As workloads fluctuate during the course of a month, week, or even through a single day, the grid computing infrastructure analyzes the demand for resources in real time and adjusts the supply accordingly.
Implementing the specific task within both technologies traditional and grid was achieved to compare and analyze new technology benefits and differences, traditional method by using normal unique machine, then job was submitted and time is measured. The same task was submitted into grid system successfully. Time comparison shows that grid system consumed much less time with high efficiency and high accuracy. From figure 7, it is clear that traditional method consumed 260 minutes (blue cylinder) while grid computational process needed 75 minutes to achieve same job (red cylinder).
The model of integrating the Grid Computing with distributed power systems has been managed to build and operate the target to perform real-time load balancing operations. Using Java Robot Class as means to run interactive graphical-based power systems simulation Power World Simulator also achieved and managed to save the simulation data which are then transferred to other simulations in real-time by using the Globus GridFTP function. Real-time load balancing operations are implemented by sending a Ëœchange commandâ„¢ from one power system to the other power system(s) once the load exceeds the specific limit. By utilizing the available Grid Computing resources, load forecasts are calculated where the forecast values are then returned to the relevant power systems in order to take appropriate actions if required. The application is split into various sub tasks and are submitted to grid, each node performs their work and send back the result to the job submitting machine. As the jobs are processed in parallel, the results are obtained in less time.
The process of integrating Grid Computing into the power industry can be concluded as economically and technically feasible. It means that power utilities companies should not have any problems in financial factors and in meeting the technical requirements.
1. R. Al-Khannak, B. Bitzer, Load Balancing for Distributed and Integrated power systems using grid computing, IEEE 2007.
2. Michael Hategan, Gregor von Laszewski, Java CoG Kit Karajan/Gridant Workflow Guide, cogkitrelease/4_0_a1/manual/workflow.pdf.
3. Joshy Joseph, Craig Fellenstein, Grid Computing, Pearson, 2007, page-239.
4. Yawei Li and Zhiling Lan,A survey of load balancing in Grid computing, Springer Berlin / Heidelberg,2005,pages:280-285
Use Search at http://topicideas.net/search.php wisely To Get Information About Project Topic and Seminar ideas with report/source code along pdf and ppt presenaion