Failures in medium-voltage power cables and their components cause a large proportion of annual power service interruptions, especially in high-density urban areas. Locating and repairing weak areas in cables at an early stage can improve the reliability of the energy supply considerably.
Analytical methods for asset management that depend on cable-specific failure models work best when the condition of the cable is known. Partial discharge testing is a proven condition assessment/test methodology. A partial discharge (PD) is a short pulse that originates from weak spots along a cable.
In its current form, PD testing occurs offline and involves some drawbacks in terms of cost, risk of damage to the cable, safety, and operational impact. However, a new measuring technique has been developed that can diagnose and locate PD activity on a medium-voltage power cable while the cable remains in service. The patented system, which was created by KEMA in partnership with the Eindhoven University of Technology in the Netherlands, uses inductive sensors to detect, diagnose, and locate partial discharge activity along a medium-voltage power cable while the cable remains online.
The PD Detection Online With Localization (PD-OL) system provides significant benefits over the current offline test systems. While the offline systems require disconnecting a cable from the grid and energizing it with a separate power supply to generate PDs and are performed on an occasional basis, PD-OL enables continuous on-line cable monitoring, allowing utilities the benefits of:
• Maintaining continuous power supply during measurement;
• Testing the cable system under exact operating conditions, including over voltages and load variations;
• Cost-effective operation with minimal personnel effort required after installation;
• Continuous registering of data that captures time-related information, such as variation of PD activity in time, the effects of sudden over-voltages, and cable temperature variations; and
• Detecting PDs occurring shortly before failure.
Many of North America’s critical customer areas are served through underground distribution systems, at voltages between 5 kV and 35 kV, using paper insulated, lead covered or solid dielectric (polymeric) cables. Although these systems have given good service in the past, wear and tear at several utilities are driving up failure rates, operating problems, and service interruptions as the systems age. Repair and replacement costs, as well as service quality problems due to circuit and equipment outages, are expected to escalate in the future unless preventative action is taken.
An asset-management strategy that includes some form of proactive cable replacement, life extension, condition assessment, and continuous monitoring to manage cable failures is essential to avoid these future problems. In some cases, load reduction, switching, dynamic ratings, and temporary overloads also can be used as strategies to reduce failure rates and mitigate reliability impacts of aging cables. However, an overall cable-management strategy is essential for long-term improvement in cable-system operation and electric-system reliability to the customers. For example, a KEMA study at a major U.S. investor-owned utility predicted that the number of cable failures on its system dramatically would increase in the coming years unless certain preventative measures are taken. Currently, this utility experiences about 600 cable failures each year on an installed base of approximately 25,000 circuit miles of several different types and vintages of cables. Based on a statistical analysis of the data, this number could increase by a factor of 10 over the next 30 years. An average customer served by this utility currently experiences about 167 interruption minutes per year, with 13 interruption minutes being caused by cable outages. The number of interruption minutes is likely to increase to more than 280 minutes if only a reactive cable repair policy is employed, thereby significantly reducing the reliability of electric service to its customer base.
The PD-OL methodology has been tested and developed by KEMA in Europe and is entering the pilot stage in the United States. The system locates the origins of PDs by using two inductive sensors, each at one cable end. A patented solution is used for the time synchronization of the data intake at both cable ends and for the online calibration. One computer at a remote location collects data from various PD-OL measuring systems for final interpretation and presentation, which can be made visible on a secured Web site for the network owner.
PD measurements can help to diagnose medium-voltage cables. The offline test systems applied today can measure and locate PDs by placing one sensor at one cable end and using PD pulse reflection. This implies that the cable has to be disconnected from the grid and energized with a separate power supply to generate the PDs. The primary reason PDs currently are not measured and located online is that under online conditions, PD pulses barely will reflect at the cable ends, making localization with one sensor practically impossible. The reflections become too small to detect because of the non-infinite impedance of the equipment connected to the cable under test at the transformer house or substation. Furthermore, measuring with one sensor at one cable end, it is impossible to distinguish pulses from the cable under test and pulses originating from other adjacent equipment.
For online PD detection and localization, the only effective solution is placing sensors at both cable ends illustrated in Fig. 1. While this solution may look simple, achieving accurate time synchronization of the digitizers at both cable ends is problematic. In the case of a required PD localization accuracy of about 1 percent of the cable length, the related time synchronization of the digitizers should be in the order of 100 ns for a 2 km long. For this synchronization, KEMA has a known and patented solution with GPS, which is used for offline PD measurements on long (up to 10 km) and branched cable networks. However, this is an expensive and impractical solution for long-time monitoring. A possible alternative, as suggested by many, is the use of atom clocks, but these either are not stable enough or even more expensive.
The PD-OL measuring system uses pulse injection for synchronization. Pulses are injected at one cable end on a regular basis and measured at both cable ends. These pulses are injected with an additional coil along the measuring sensors at the cable termination. This method of synchronization, a crucial and patented element in PD-OL, is both inexpensive and very effective.
PD-OL also uses inductive sensors, which have the advantage of being cost-effective and easy to install. Such inductive sensors have certain disadvantages. In general, inductive sensors placed in a ground lead will pick up more noise/interference than do capacitive sensors. In addition, inductive sensors only can be applied on terminations where the metal screen at the cable end has a separate ground lead (a path for the PD pulse) to the general grounding system. In a termination where the metal enclosure of, for instance, the switchgear is fully bonded with the metal screen of the power cable, such an ideal location for an inductive sensor cannot be found. Since PD-OL currently is based on inductive sensors, this measuring system is optimized for terminations without full metal enclosures.
Before a PD can be measured, it is important to calibrate the system. In PD-OL a smart solution is applied. At each cable end, steep pulses are injected, and measured separately at both cable ends. Magnitude and shape of resulting signals measured with the inductive sensors depend on the impedances of the power cable under test and the connected equipment at the transformer house or substation. In this way, the transfer impedance for a pulse coming from the power cable and entering a sensor is calculated. This information is used for the actual calibration of PD-OL. Again, the pulse injection method is crucial here.
PD data acquisition is carried out at both cable ends. There the first step of the data analysis is done to extract possible PD signals from the noisy data, remove interfering signals and thereby reduce the size of the data to be communicated. The resulting data is sent by LAN, modem, and telephone connection or GPRS to a server. Here, the final PD map is made, based on PDs and their time of arrival.
The costs of the PD-OL equipment together with the interpretation of the results are only a fraction of the costs of replacing a cable connection. The costs that will be reduced are the costs related to outages. These costs usually are only partly direct costs for the utility itself. Furthermore, if the knowledge of the condition of the cable network improves (which happens when using PD-OL) it will help the utility with replacement strategies. If utilities can postpone cable replacement investments, benefits are huge compared with the costs of PD-OL.
PD-OL has been demonstrated successfully on several sections of differing types and lengths of cable in the Netherlands, and is now at the first production stage, with approximately 70 systems on order for several utilities in Europe. The activities foreseen in the second half of 2006 are directed at bringing PD-OL to the market, combined with extensive testing in the field. In addition, further work on the PD-OL methodology focuses on the application of PD-OL in difficult networks (branched, cross-bonded, many RMUs, etc.) and to the further enhancement of knowledge rules.
The advantages make PD-OL attractive in a much wider environment of cable diagnostics. Especially because (1) a control room can monitor all cables connected to PD-OL; and (2) not only long-term defects, but also so-called “short term” defects can be traced prior to breakdown.