Documents

Predictive Analysis of Electrical Power Failures in Sri Lanka using Big Data Technologies

Description
Entire world is depending on Energy or power, which is the core of modern technological era. Without power for a short time will cause huge difficulties in the world related to financially, socially and logistically. Therefore electrical power failures are critical in modern technological era. Sometimes it is difficult to identify the causes of outages and simulation of those. So prediction of outages, are even difficult. Most of the power companies do not have methodology of failure analysis and outage prediction system worldwide. Therefore, this research addresses this deficit by introducing big data analytics and machine learning techniques based on outage management and prediction system included an efficient outage management system with these components: Data Warehouse Model Construction, Extraction, Transformation and Loading (ETL) Process and Information Representation. The purpose of this system is to store, integrate, and analyze complex data sets from multiple secondary data and information sources to predict power outages in Sri Lanka by using most accurate big data analytics algorithm.
Categories
Published
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
   M A B Madurapperuma et al, International Journal of Research in Information Technology, Volume 1, Issue 2, Dec 2017, Pg: 65-73 ISSN (Online): 2001-5569 UGC Approved Journal 65 ©Copyrights IJRIT, www.ijrit.net  Predictive Analysis of Electrical Power Failures in Sri Lanka using Big Data Technologies   M A B Madurapperuma, P P G Dinesh Asanka  M.Sc in Information Technology, Faculty of Graduate Studies and Research, Sri Lanka Institute of Information Technology (SLIIT), September 2016    Abstract Entire world is depending on Energy or power, which is the core of modern technological era. Without  power for a short time will cause huge difficulties in the world related to financially, socially and logistically. Therefore electrical power failures are critical in modern technological era. Sometimes it is difficult to identify the causes of outages and simulation of those. So prediction of outages, are even difficult. Most of the power companies do not have methodology of failure analysis and outage  prediction system worldwide. Therefore, this research addresses this deficit by introducing big data analytics and machine learning techniques based on outage management and prediction system included an efficient outage management system with these components: Data Warehouse Model Construction, Extraction, Transformation and Loading (ETL) Process and Information Representation. The purpose of this system is to store, integrate, and analyze complex data sets from multiple secondary data and information sources to predict power outages in Sri Lanka by using most accurate big data analytics algorithm. Keywords: Data warehouse, ETL, Outage Management   1.   Introduction Energy or power is the core of human growth towards technology. Therefore managing energy related to consumptions and outage is so important for better power utilizations. The growth of available data in the electric power industry motivates the adoption of data analytics techniques. However, the companies in this area still face several difficulties to benefit from data analytics. One of the reasons is that analysis  power systems data is an interdisciplinary task. Typically, electrical and computer engineers need to work together in order to achieve breakthroughs, interfacing power systems and data analytics at a mature level of cooperation. Another reason is the lack of freely available and standardized benchmarks. Because of that, most previous research in this area used proprietary datasets, which makes difficult to compare algorithms and reproduce results. Sri Lanka is in the process of industrialization, urbanization rapid development, which leads to constantly rising of energy demand, and the annual electricity consumption is increasing annually. Adding to various researchers related to power consumptions and outages, through this research, a methodology for analysis  power outages and effective outage management system for Sri Lanka is presented. Ceylon Electricity Board used a mechanism of gathering and keeping outage data. When a failure happens most of the times customers are complaining to their hotline and the call center operators enter data into the system and assign jobs to relevant parties. Then monitor the fixing process of the issue with area, grid details, taken time to complete as well as t he identified causes. But the problem is they don’t have proper failure analysis and prediction methodology.   M A B Madurapperuma et al, International Journal of Research in Information Technology, Volume 1, Issue 2, Dec 2017, Pg: 65-73 ISSN (Online): 2001-5569 UGC Approved Journal 66 ©Copyrights IJRIT, www.ijrit.net  In order to address the above issue, in this research we proposed an outage analysis and prediction system  based on big data technology. Traditional outage management database systems, with emphasis on electricity data monitoring, lack the ability to create data aggregation and do not support the analysis of electricity data to deliver reports and actionable information. In addition, information requirements  become complex, as it is difficult to extract information from the data found in power outage databases. Therefore, big data technology is introduced to manage and analyze outage data in an integrated way. 2.   Literature Review Though there are various kinds of researches in the world related to power, we can find out few of those related to power outage analysis and predictions. Therefore lacking much of references related to the topic  but it is very important to refer how the other researchers completed their works under the main steps of  big data analytics. So the survey is mainly interlaced with four major steps of the field of big data related to existing researches, namely Data collection, Data Cleansing, Data modeling and Analysis & Prediction. In these researchers Association Rule, Decision Tree, Naïve Bayes kind of analysis algorithms was used mainly. 2.1 Implementation Methods Related to Data Collection, Cleansing and Modeling Techniques Sowmya R and Suneetha KR [1] have presented a Methodology for big data analytics with three major  parts including data acquisition, Data Processing and Data service which act as an interface for the users to retrieve data and use them from another source without much delay. From an article by Nikunj Thakar [2] related to big data in energy industry, presents the important of data collection by providing examples of investments for both manual and digital data collections by power generated companies. Similarly, it was mentioned the feature of variability in big data by providing the variables which need to be considering in hydro power stations such as water force, quantity, quality depth or height of water terrain type, soil quality and maximum and minimum flow of water etc. This information can be very critical in predicting failures as well. Jefferson Morais, Yomara Pires and et. all[3] from their paper of An Overview of Data Mining Techniques Applied to Power Systems, described about the growth of available data related to power industry and mentioned about selecting the correct necessary information out of those. For that they have worked with both electrical engineers and Computer engineers to do the data collecting and cleansing part along with sort-outing the proprietary issues of data. For data   collecting they have used fault classification  because its nature of easing the generating artificial data using simulators such as alternative Transients  program (ATP).   Xin Zhang, Donghua Li and et. all [4] in their research related to electricity consumption pattern recognition they gathered data based on the shifting of peak load from users ’  consumption pattern. In this research they have considered the peak load shifting strategies with the different varieties of them and described how the results are used to provide a proper continues electricity to users. Viktor Botev and et. all [5] in their research of Detecting Non-Technical Energy Losses through Structural Periodic Patterns in AMI data, introduce a data driven method called Structure & Detect which identify causes in electrical power grid’s non technical losses (NTL). The analysis of structural periodic  patterns related to consumption traces, through spectral analysis is the base for Structure & Detect. They only used consumption data such as trust or credit history, with no need for exogenous data about Customers or open information from domain experts. They presents in their paper that Structure & Detect   M A B Madurapperuma et al, International Journal of Research in Information Technology, Volume 1, Issue 2, Dec 2017, Pg: 65-73 ISSN (Online): 2001-5569 UGC Approved Journal 67 ©Copyrights IJRIT, www.ijrit.net  method results high accuracy and high detection rates b using real world consumption traces, equivalent to methods that require additional, customer-specific information. Various kinds of granularities which enables different monitoring possibilities are used the spectral-analysis basis implies Structure & Detect, as well as , based on incremental Fourier computation the traces can be processed as streams which allows distributed processing, continuous and utilizing efficient synchronization methods. The technique also allows parallel processing, including both parallelizing the different traces processing and for the transformation itself. These are significant for scalability, with deployment in modern infrastructures. Big data Analytics in Power distribution Systems by Nanpeng Yu [6], has present a big data application to distributed systems to reveal their massive value related to system planning and operations as part of the research. Parallel they have bringing out a flexible system design which can be integrated with Hadoop clusters in existing relational database to handle workload of heterogeneous big data analytic. In here to deal with big data they have used various sources to gather data such as smart meters to measure customer data, Supervisory control and data acquisition (SCADA) system to gather distribution and transmission data and specially they have used external factors as well including prices, weather, publicly collected   census and text for analytics purposes. Finally they have tackled privacy and protection issues of data in distributed systems related to big data applications. Dong Zhu and et. all [7] in their research of creating an analyzer for characteristics of Electricity behavior  based on big data, have proposed a method of three levels such as data preprocessing layer, basic analysis layer and advanced application layer which is totally aligned with big data techniques. In here they analyzed such as load classification method, load characteristics, power consumption behaviors analysis and abnormal electrical behaviors. Finally they presented, that this method consider multiple factors affecting due to behavior of users using big data analysis point of view which outcomes scientific theoretical support for power system operation. 2.2 Implementation of Analysis & Prediction   In this section it was described how the existing researches implemented data analysis and few analysis algorithms used. 2.2.1 Decision Tree Algorithm Decision Tree is one of a popular technique for prediction. Most of researchers have used this technique  because of its simplicity and comprehensibility to uncover small or large data structure and predict next failure [5]. Quadri (2010) used decision tree to technique to choose the best prediction and analysis student data for academic performance [8]. They used Gender, Attendance, Scholarship, Working, Firstchild, Dropout as variables. Hongqin Fan used regression trees models used to validation and forecasting Time between Failures[6]. The algorithm searches over the data space and recursively partitions it into subspace, where more pure information or promising relations can be found. Based on that, they predict forecast next failure. Teng (2015) proposed decision tree algorithm on the subject of regional voltage stability problem. They used Matlab and CART software to process the test data, the test case were simulated [7]. The training set is used to build the decision tree algorithm model. The method used is Gini index. Linear combinations option is used in the analysis. Panigrahi (2015) proposed decision tree algorithm for stock market forecasting, during the data selection they used 12 technical indicators. The transformed data set they run using decision tree J48 splitting   M A B Madurapperuma et al, International Journal of Research in Information Technology, Volume 1, Issue 2, Dec 2017, Pg: 65-73 ISSN (Online): 2001-5569 UGC Approved Journal 68 ©Copyrights IJRIT, www.ijrit.net  algorithm. Based on that they propose text based decision tree model. And they prove that text based model is more accuracy than normal tree [1]. 2.2.2 Naive Bayes Algorithm The Classification technique is a significant data mining technique with large applications. And it is used to categorize each item in a set of data into one of predefined set of groups or classes. R.Nithya [2015] used to analyze heart disease dataset using Bayes algorithms. Bayes classifiers algorithms namely Bayes Net, Naive Bayes, Naive Bayes Multinomial Text. They found Naive Bayes algorithm performs better than other algorithms. Hence we decided to analyze probabilities based on  Naive Bayes.   3. Implementation Methodology Outage Management System (OMS) of Ceylon Electricity Board (CEB) records data of all outages and interrupts data for whole island since 2011 and in parallel the outage management data as well. This system record all outage related data in all levels such as grid wise, sub grid wise, area wise, depot wise and transformer wise. For this research it was used power outage data from 2013:01 to 2017:11 for Areas of Gampaha, Colombo north and Weligama. In Sri Lanka, outages are mainly depending on the weather, Landscapes and vegetation according to the experts. Therefore the selection of data set was done by covering the above factors. So selection of Colombo city data is based on, identifying the difference of frequency of outages and efficiency of outage managements as an urban underground network, Gampaha as a semi urban area with slight vegetation. Finally, Weligama as an area where lot of vegetation and various landscapes. There are four deports for an area in average in Ceylon Electricity Board that means in selected three areas, there were all together 12 depots. But when considering the total number of outages reported for last five years and it was dropped 7 deports which was not having 5000 outages reported, in data cleansing part because it was very little amount for analysis when considering 150000 outage records. So the research was done by using 5 deports across the three areas. 3.1 Methodology The process starts with the existing outage management systems’  Informix database. Then the data was extracted to Excel flat data file. Then the extracted data was transformed and loaded into Microsoft SQL Server by the use of SQL Integration Services. Accessing SQL Integration Services was done by SQL Data tools plugging inside Visual Studio 2015. Then it was created mining structures for analysis algorithms and Analysis Server Data Base. This is also handled through SQL Server Data Tools. Inside Analysis Server Data base all the possible test cases were created and data was trained for predictions. The final part is new proposed outage prediction application which is developed base on proved analysis algorithm.
Search
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x