Papers discussing DFA Technology
Automated Power System Waveform Analytics for Improved Visibility, Situational Awareness, and Operational Efficiency
The proliferation of “smart” devices on distribution feeders in the past decade has resulted in a deluge of data. Most utilities recognize that recorded waveforms and other data contain information that could enable them to operate more effectively. In practice, however, most utilities find themselves confronted with an intimidating mountain of data, without tools or expertise to differentiate the important data from the pedestrian. Much data analysis continues to be performed manually, off-line, in response to specific perceived problems. That can provide forensic value, if the utility has the requisite expertise and the time necessary to identify and interpret the relevant data, but it provides little real-time system visibility or situational awareness that would enable operational improvements.
For multiple years, supported by EPRI and EPRI-member utility companies, Texas A&M researchers have used sensitive, high-fidelity waveform recorders to collect data from scores of feeders, using technology readily achievable with modern electronics. This has created the most comprehensive extant database of waveforms of incipient failures and feeder events. Based on that database and experience, they developed sophisticated waveform analytics and reporting methods. Dubbed distribution fault anticipation (DFA) technology, the system acquires high-fidelity waveforms from conventional CTs and PTs and then uses automated processes to apply analytics to those waveforms and thus report events and conditions. This provides personnel with real-time visibility of feeder events and conditions, including incipient failures. This newfound visibility, or awareness, enables improved reliability, improved operational efficiency, and true condition-based maintenance.
This paper explains general DFA concepts and then uses selected case studies to illustrate concrete operational benefits to utility companies.
Advanced Monitoring of Low-voltage Secondary Networks for the Detection and Mitigation of Arcing Faults
Low-voltage secondary networks represent a particular power-system topology deployed in load-dense urban environments requiring ultra-high reliability. These grid networks have a high degree of interconnectivity that provides multiple, redundant paths to loads and, consequently, also to faults. For nearly a century, arcing faults on grid networks have been a well-known problem, creating smoke, fires, and explosions, collectively known as manhole events, as well as disrupting normal service. Network arcing faults draw intermittent current of relatively low level that seldom operates conventional protection devices. They generally remain undetected until manhole events or other operational problems cause their discovery. Though discussion of arcing faults appears in secondary network literature as early as the 1920’s, as late as the early 2000’s, most experts have considered them “an industry problem that presently has no available solution.”
Texas A&M University has partnered with the Consolidated Edison Company of New York (Con Edison) on a project to detect, locate, and ultimately mitigate arcing faults on secondary networks. Researchers instrumented a single secondary network with thirty high-speed, high-fidelity data collection devices, as well as one functionally identical device on a primary feeder serving that secondary network. Project results show that arcing faults 1) occur more frequently than previously understood, 2) can persist for long periods, 3) can recur multiple times over a period of days or weeks, interspersed with quiescent periods of hours, days, or longer, and 4) can be detected by direct network monitoring and also by monitoring medium-voltage primary feeders serving the secondary network.
This paper presents selected examples observed on operational secondary networks during a nominal two-year period. During this period, thousands of distinct arcing events were recorded and analyzed, with many events simultaneously recorded at multiple network points and the primary-feeder point. Preliminary results suggest multi-point monitoring of secondary networks may enable identification and location of incipient arcing conditions before they rise to the level of a public safety hazard.
Automated Waveform-Based Analytics for Enhanced Reliability, Power Quality, and Operational Efficiency
The past decade has witnessed numerous technological innovations to improve various aspects of operational efficiency and reliability. Upon the occurrence of an outage on a distribution feeder, for example, automated schemes can perform switching to localize the outage and restore service to as many customers as possible, often within seconds to minutes. Other systems use data-mining analytics to discover patterns and problems identifiable from AMI (advanced metering infrastructure) databases.
Research at Texas A&M University applies analytical algorithms at a finer level of detail than do AMI-based data-mining analytics. Specifically this research recognizes that electrical waveforms, as measured from conventional current and potential transformers (CTs and PTs), reflect load and other activity occurring on the power system and that, therefore, these waveforms hold the potential for providing the utility with heightened awareness of feeder operations and conditions.
Texas A&M’s efforts in this area began in the late 1990’s, at which time they focused specifically on anticipating, or predicting, future faults. Long-term instrumentation of 70 feeders at 11 utility companies created a large database of waveforms representing a wide variety of activity on feeders. The process also demonstrated that waveforms reflect a wide variety of valuable feeder information, not just incipient failures. As an example, as has been recognized for decades for protective relaying purposes, a conventional overcurrent fault produces specific variations in measured currents and voltages, and a switched capacitor bank produces specific variations that differ markedly from those of the overcurrent fault. Other project findings document, for the first time, that anomalies such as failing in-line switches or clamps also produce specific variations in waveforms. Algorithms capable of detecting and differentiating specific types of variations provide valuable situational awareness and intelligence that can help utilities proactively avoid certain faults and respond more intelligently and efficiently to faults and other system anomalies.
Certain smart grid technologies can reduce the number of customers affected by prolonged outages, and thereby increase reliability through automated switching to restore service. Such technologies are useful, but reactive in nature, performing their function only after a fault occurs and an outage has been detected. They must presume that nonfaulted feeder sections and alternative feeders are healthy and capable of carrying increased power flow. Research at Texas A&M University has demonstrated that sophisticated, automated real-time analysis of feeder electrical waveforms can be used to predict failures and assess the health of distribution lines and line apparatus. Reliability can be substantially improved by detecting, locating, and repairing incipient failures before catastrophic failure, often before an outage occurs. Requirements for data and computation are substantially greater than for devices like digital relays and power-quality meters, but feasible with modern electronics. This paper provides selected examples of failures that have been predicted by intelligent distribution fault anticipation (DFA) algorithms. The data requirements and processing analysis to detect these failures are discussed. The problems related to full-scale deployment of the proposed system in a utility-wide application are presented. The authors use experience gained from their long-term research to propose concepts for overcoming these impediments.
DFA Technology Puts End to Three-Month Problem
In September 2004, lights went out at several residences fed from a single-phase 50-kVA pole-top transformer on a 13-kV circuit in Staten Island, New York. Con Ed field personnel found that the secondary breaker of the CSP transformer at this location had tripped. At this location, secondary service cable from the transformer went down the pole and into a buried connection box at the base of the pole. A "crab" in the buried box connected this supply cable to three direct-buried service cables, each of which fed a customer meter. Personnel found no evidence of a problem, and reset the transformer breaker. The unit remained closed with all affected customers in service and no reports of flickering lights or partial service.
High-impedance, arcing faults (HiZ faults) are a perennial problem for distribution systems. They typically occur when overhead conductors break and fall, but fail to achieve a sufficiently low-impedance path to draw significant fault current. As a result, conventional protection cannot clear them, resulting in situations that are hazardous both to personnel and to property.
Texas A&M researchers spent two decades characterizing HiZ faults and developing and testing algorithms for detecting them. In the mid 1990's, General Electric commercialized the algorithms in a relay for detecting a large percentage of these faults, while maintaining security against false operations.
In an effort to mitigate problems associated with these faults, Potomac Electric Power Company (Pepco) installed the HiZ relays. They evaluated the performance of these relays on 280 feeders over a period of two years and gained significant operational experience with them. Being the first utility to apply high impedance fault detection technology on such a widespread basis makes Pepco's experience valuable to other utilities that are struggling with decisions regarding their own response to the problem of high-impedance faults.