Plenary Lecture

Cable Asset Management Using Failure Forecasting and Statistical Assessment of Diagnostic Testing

Professor Miroslav Begovic
School of ECE
Georgia Institute of Technology
Atlanta, GA, USA
miroslav@ece.gatech.edu

Abstract: Utilities the world over and especially in North America are facing a significant future challenge to maintain and renew their underground assets. Their aging assets are leading to ever increasing failures while at the same time more energy is being delivered. Immediate replacement is prohibitive in terms of cost and availability of correct quality components. Thus asset management strategies are increasing being used to help address the issue.
Methods like long-term simulation, multi-criteria decision-making under uncertainty, critical asset identification, condition assessment, and advanced statistics for the extrapolation of condition assessments of representative samples of assets should be applied. By using these methodologies in a smart and integrated way, costs and performance can be kept at an acceptable level.
The problem of resource management has long been recognized as one of the burning issues in electric utilities. Knowing how much to invest in creating a reliable and successfully performing resource pool (i.e. distribution cable network), when to repair or replace, and what human and financial resources are needed from year to year in order for such a network to operate successfully, the answers to those questions may represent substantial savings for the utility. Among the most acute problems that utilities are facing is the problem of accurate logging of system past performance and failure rate. As far as cables go, very little or no information is available to support such an activity.
Failure forecasting based on historical failure performance is proposed as a basis to extract the parameters of the statistical distribution, which is assumed to describe the failure rate performance of the entire cable population. We have expanded and modified that approach to include multiple parameters identification and nonlinear models, and tested it using field data which was obtained from actual field data. In addition, the have expanded the methodology to include Monte Carlo simulations of the failure rates in order to produce the estimates of distributions of failures rather than the most likely estimates. By doing so, we associate confidence ranges with estimates of failure and replacement rates that are forecasted in the short time horizon into the future. By doing so, planning can be associated with the desired level of confidence, which provides better quality information for a cost-conscious utility planner. It should be noted that the accuracy of the proposed methodology strongly depends on the quality and quantity of the input data, and it is envisioned that it could be enhanced in the future by combining chronological failure rate information with some form of condition monitoring, which can be coupled with the failure model that may sharpen the accuracy of the failure forecasts needed for a precise planning.
It is therefore clear that if we will also have to place a great deal of importance on Diagnostic Information, then we need to be fairly certain that this information is relevant. It would also be advantageous to determine the accuracy by looking at network performance. In this area most practical engineers are not searching for perfection but they do want to be sure that when they place their funds into Diagnostic Testing they are winning more than they loose and that they are doing much better than if they relied on chance. To this end we have been examining a number of ways in which it is possible to test and validate Diagnostic Information against the true system performance. As there are, what seems to be, a bewildering variety of Diagnostic Techniques at a Utilities disposal we have further concentrated on the methods that are ‘Technique Independent”: applicable to all.
The presentation will look at methods to assess how well the diagnostic information relates to a specific system. Primarily this means comparing the predictions from the diagnostic information with subsequent failure performance in real life, both before and after the diagnosis.

Brief Biography of the Speaker:
Miroslav M. Begovic is a Professor in the School of Electrical and Computer Engineering at Georgia Tech. He received a Ph.D.E.E. from Virginia Tech. Dr. Begovic’s work on monitoring, analysis, and control of power system stability, asset management in power system equipment and application of diagnostic and condition monitoring techniques, as well as applications of wide area monitoring and protection technologies in solving power system problems, has resulted in a number of projects for government and industrial sponsors, as well as publications and awards. Dr. Begovic has actively contributed to the IEEE Power Engineering Society, where he is involved in several technical and administrative activities. Dr. Begovic is a Fellow of IEEE and a member of Sigma Xi, Eta Kappa Nu, Phi Kappa Phi and Tau Beta Pi.
Professor Begovic coordinated projects with a total funding of over $10 million. (This does include the new project on the cable diagnostics focused initiative, co-funded by the US Department of Energy (DOE) and NEETRAC with a budget of $3.5 million.) Prof. Begovic authored or co-authored over 80 technical papers, one book section, one patent, two IEEE special publications, and several seminars, short courses and invited presentations. He was part of the design team which developed the first working prototype of the GPS-synchronized phasor measurement system for power system real-time monitoring, later commercialized by several manufacturers and currently the object of widespread implementation across the US eastern interconnection.