Making Decisions With Data by Chris Trayhorn, Publisher of mThink Blue Book, May 15, 2006 How well does a utility perform its distribution operations? The answer to that question depends in part on what you are trying to achieve. Most utility stakeholders (customers, regulatory bodies, local governments, staff and owners) formulate an answer based on some combination of reliability, cost, power quality and safety. Lets review how these factors can be evaluated as they relate to the performance measuring, monitoring and reporting challenges facing todays electric distribution company. From Data to Decisions Critical to the evaluation of the success or failure of any business process is accurate knowledge of how the process performed in the past, how it is performing now and how it is likely to perform in the future. That knowledge can arise only from data. But amassing this data, transforming it into knowledge and communicating it to stakeholders can be a daunting task. This raw data must be captured and transformed into meaningful information that can be understood by the stakeholders. The stakeholders need to then process the information in the context of their existing knowledge, building a broader knowledge base from which to make decisions about tactical corrections and strategic modifications. Once this knowledge has been cultivated, cause-and-effect relationships can be investigated to determine the drivers behind absolute performance and performance trends. The electric distribution company is no exception to this proven method of achieving organizational excellence. The Complexity of Data Assembly and Analysis Measuring In the area of electric distribution operations, key performance indicators (KPIs) have traditionally taken the form of engineering calculations that provide an indication of how well the delivery system itself (the wires) succeeds at keeping the lights on and how well operations responds when the inevitable failure does occur (restoration). These measures are commonly referred to as reliability indices and are intended to measure performance from the electric customers perspective (i.e., customerbased measures). Distribution operations typically perform in one of two modes of operation normal or abnormal. Abnormal operation stems from the fact that electric distribution systems operate in harsh environments that are beyond the control of the utility. These systems are inherently exposed to natures elements, including animals, vegetation and weather, as well as geological and human incidents. Since it is cost prohibitive to build a system that operates flawlessly in this environment, each utility has adopted design, construction and maintenance standards based on an acceptable trade-off between cost and performance as determined by their regulators and/or customers served. Incorporated in the development of these standards are limits related to the harshness of the environment that loosely define normal operating conditions. The lights are expected to stay on while operating within these limits. The environmental harshness will occasionally exceed these limits (termed major event), changing the operating mode to abnormal and altering performance expectations. Standard Practices Consistent measurement practices with defined methodologies and terminology are critical to meaningful tracking and analysis of electric distribution performance. Not only required in support of internal distribution company decision making, analysts, investors, regulators, owners and large commercial/industrial electric customers regularly compare performance between different companies in support of their own internal decision making. The Institute of Electrical and Electronics Engineers (IEEE) is the commonly accepted authority on standardization in this area and has published the IEEE Guide for Electric Power Distribution Reliability Indices IEEE Std. 1366. Industry surveys conducted by the IEEE/PES Distribution Subcommittee Working Group on System Design have set the most commonly used customer-based indices (see Figure 1) and in order of frequency of use. These calculations transform raw interruption data typically logged during the day-to-day operation of the electric distribution system into meaningful information. Transformation of this information into a knowledge base requires further augmentation with operational data related to the cause of each interruption, operating mode at the time of each interruption, type of isolating devices involved in each interruption, and type of any electric system component that may have failed. A recently submitted white paper by the IEEE/PES Distribution Subcommittee Working Group on System Design, titled, Collecting and Categorizing Information Related to Electric Power Distribution Interruption Events: Data Consistency and Categorization for Benchmarking Surveys, further defines a minimum set of data collection categories required for benchmarking (see Figure 2). The resulting collection of knowledge provides a set of powerful information on which to base operational, engineering and financial decisions across the entire electric distribution enterprise. Monitoring and Reporting Simply measuring performance and building a base of knowledge does nothing, in and of itself, to affect performance. Performance must be monitored over time for the measures selected to be useful in steering processes in a direction that will result in organizational success. Winning decisions require knowledge of where we are as well as where we are heading. Performance monitoring and reporting methods vary depending upon the business needs of the particular function or process that is to benefit from the available knowledge. Required periodicities of access and information granularity are important factors in determining the best methods for the accumulation, structure and dissemination of knowledge. Static annual reports of KPIs may suffice for some planning functions but operations management functions typically demand dynamic access to current information; not only to the KPIs, but also to the raw data that surrounds and impacts the KPIs. A diverse range of performance monitoring and reporting needs exist within the electric distribution enterprise: Predefined reports of aggregate information about performance over relatively long periods of time (quarter-to-quarter, year-to-year) typically meet the needs of performance-based regulation and regulatory compliance reporting; System planning and design functions need similar reports of like information but also benefit from more interactive reporting methods; The ability to view information from various system perspectives is necessary to support reliability planning efforts; Maintenance functions typically require more granular information about performance over somewhat shorter periods of time (week to week, month to month) and need even more interaction with the raw data; The ability to view information from various equipment perspectives is necessary to support reliability-centered maintenance efforts; and Operation planning functions require very granular performance information over even shorter periods of time (hour to hour, day to day) and need maximum flexibility in reporting. All of these functions can benefit from, if not require, the ability to filter specific abnormal occurrences or accepted normal occurrences from the KPIs and other information reported. For example, comparison of the performance of a particular system design against related standards could be skewed if data from events occurring outside of the predefined limits within the standard (e.g., major events or scheduled interruptions) are not excluded. This could result in unnecessary and costly upgrading and overbuilding of the distribution infrastructure. Obviously, the definitions of excluded occurrences are critical to accurately evaluating and comparing KPIs and related information. When a major event occurs on the electric system resulting in customer interruptions, the required periodicity of access to information increases to less than hour to hour, the required granularity of information increases significantly and the number of interested parties increases by orders of magnitude. Operations management and support personnel need current information, including information down to the individual electric customer level, on which to base the minute-to-minute decisions necessary to assist operators/dispatchers in quickly and safely restoring electric service; Corporate communication personnel need similar information for dissemination to the media and response to direct public contact; Planning, design, maintenance and other personnel may be called upon to lend assistance in developing restoration plans and communicating with external entities; and Depending on severity of the event, regulatory and emergency management organizations may require access to current information. Technology Challenges Measuring, monitoring and reporting performance presents several technology challenges that cannot be adequately addressed by traditional operational IS approaches. Although additional users could be added to the operational system to gain access to the operational data, each additional hit to the operational database degrades performance, however minimally, of the application programs that depend on the operational database. Users interested in the data, as opposed to the application program results, will typically generate many, many more hits per unit of time than the application programs for which the database was designed. Furthermore, these additional hits will tend to occur during the same time as maximum usage of the application programs (normal working hours). Neglecting IS performance, direct cost and physical implications, opening access to the operational data on a broad scale through the operational system presents additional security risks that could directly and negatively impact the operational system itself. Preventing breaches would significantly add to the challenges of the system administrators. Consideration must be given to the functionality of the data analysis and reporting tools available. Custom development and maintenance of these tools is very costly. Without creative solutions, these challenges lead to severe limitations and ultimately abandonment of access to operational data the exact opposite of what the organization The Data Mart Helpful in understanding the data mart solution is a comparison of operational system data (see Figure 3) and business management environments (see Figure 4) to that of the data Fundamental to the proper design of a data mart solution is a thorough understanding of the requirement issues dictated by the specific implementation under consideration. Requirements typically fall into one of three categories: The business function and scope requirements definition asks questions about which specific business problems in what part of the enterprise are to be addressed; The data requirements definition asks questions about characteristics of source data and the needs for its extraction, refinement and re-engineering; and The access and usage requirements definition asks questions about who will use the solution, when will they use it, what will they use and how will they use it. Business Function and Scope Requirements These various functional areas that can benefit from a distribution operations data mart have varying needs. These needs affect data mart design with respect to dimensional analysis, granularity of information and temporal analysis. Dimensional analysis (see Figure 5) involves determining how to best examine or slice and dice the information captured to best address the business problems at hand in a particular area of the enterprise. Dimensions relate to things such as time, space, frequency, etc. Figure 6 illustrates some possible examples of time and space dimensions in the realm of electric distribution operations. Other possible dimensions are: number of interruptions, number of customers out, etc. Multidimensional analysis (see Figure 7) entails slicing and dicing by a combination of multiple dimensions. Some examples are: outage duration and reliability index by outage cause; daily trouble by branch by region by company; reliability indices by year by feeder by substation by branch; momentary events by feeder by substation by branch, etc. Granularity analysis (see Figure 8) refers to determining how much detail is required in the data to meet the business requirements of the data mart. Different business functions typically require different levels of granularity with varying levels of summarization and/or aggregation. The importance of each of these requirements must be balanced against the cost of the IS necessary to provide them. The granularity of the time unit against which data is required to be captured and analyzed must be scrutinized (e.g., this year, last year, this quarter, last quarter, today, etc.). Temporal distortions can occur due to the differences in the rate of change of the various dimensions. For example, changes to normal circuit topology and customer information typically occur at a slower rate than changes in abnormal device states. The data of interest from electric distribution operations ranges from the transactional level (customer call) to the aggregate level (outages by district) and from the historical (last year) to the current (now), each having different impacts on the data mart design. Access and Usage Requirements Different users require varying levels of access to, and views of, the information contained within the data mart. Management is typically interested in easy retrieval of predefined information summaries at multiple levels to support decision making. Other business users are typically interested in the added ability to massage the information and vary the views, as well as access to detailed, achieved data. A powerful tool for the business user that is enabled by a properly designed data mart is Online Analytical Processing (OLAP). This tool provides for multidimensional analysis using drill-down and roll-up techniques as well as iterative analysis of data by changing the order of dimensions. Summary A comprehensive analysis of the combined set of issues leads to the conclusion that the application of data warehousing concepts is fundamental to the development of a robust solution. These concepts include warehousing data offline from the operational database, organizing data for efficient data analysis and reporting, opening access to operational data on a broad scale at minimal cost and isolating business users from the operational system. In addition, the popularity of data warehouse solutions has spawned the development of powerful, readily available, cost-effective data analysis and reporting tools. Filed under: White Papers Tagged under: Utilities About the Author Chris Trayhorn, Publisher of mThink Blue Book Chris Trayhorn is the Chairman of the Performance Marketing Industry Blue Ribbon Panel and the CEO of mThink.com, a leading online and content marketing agency. He has founded four successful marketing companies in London and San Francisco in the last 15 years, and is currently the founder and publisher of Revenue+Performance magazine, the magazine of the performance marketing industry since 2002.