Development of a fault location algorithm based on distributed neutral-to-ground current sensor measurements.

摘要

Faults in power systems cause power interruption, thermal damage of equipment and even complete collapse of the power system if they are not detected and isolated from the healthy parts of the power system. Therefore, location and isolation of faults in power systems is very important not only for service continuity, but also for quality of power delivered. In recent days, power quality problems in power system are one of the major issues and area of interest for research. Line-to-ground (L-G) faults are the most common type of fault in power distribution systems. Therefore, many techniques for L-G fault location have been proposed and implemented in transmission and distribution systems. In recent developments of techniques for fault location in distribution systems, Cannon Technologies (Cooper Power Systems) has proposed a new method based on real-time monitoring of current levels in the neutral-to-ground paths at different key locations along a distribution feeder. The proposed method was implemented in a real system incorporated with a communication system to retrieve data recorded by sensors placed at key locations, back to the substation whenever there is a L-G fault in the distribution feeder. The field tests were conducted on Oak Park distribution feeder, Minneapolis. The data obtained from field tests indicated that there is a potential useful relationship between the current through the neutral-to-ground paths and the fault location. Therefore, Cooper Power requested SDSU to perform further studies to help confirm their findings and establish the viability of the proposed method. The multi-grounded Oak Park distribution system, with and without ground-mesh at T sections of distribution feeders, has been modeled in EMTP (Electromagnetic transient program). The study showed that the results are influenced by the change of geographical layout of distribution feeder. The current monitored in neutral-to-ground path can provide useful information to determine the fault position in distribution feeder. Therefore, based on this information, an algorithm for L-G fault location, using the least number of sensors throughout the distribution system, is proposed. The proposed algorithm for fault location is the method of triangulation based on simple differential neutral-to-ground current measurements. In this method, the data, recorded by sensors placed at key locations along the primary feeder, are employed in an exponential curve fitting algorithm, which leads to a predicted fault location. The intersection point of two exponential lines is the predicted fault position. As the position of sensor pairs changes both in location and pair separation, the exponential fit changes accordingly, resulting in changes of expected fault position in the feeder. When the sensors at key locations are uniformly distributed by a distance of 36 poles and the method of triangulation with exponential approximation is implemented, the fault location accuracy is achieved within a maximum error of 6 poles for a pole-to-pole distance of 40 meters, which is close to the desired fault-location target of +/- 5 poles.;Since the profile of current through the neutral-to-ground paths (R ng) is affected by the change in value of resistance of modeled earth-ground (Rg), a sensitivity test is performed to observe the dependency of current through Rng due to change in Rg from 0 to 1000 O/km. The sensitivity test showed that the nature of profile of current through Rng remained the same when the value of R g increased, but an increase in value of Rg results in an increased distribution of current through Rng near to the fault location. Similarly, the validity test of the proposed algorithm for fault location is performed by changed profile of current through Rng due to change in Rg from 0 to 1000 O/km. The proposed algorithm for fault location worked for all the test cases. However, the number of sensors and spacing between the sensors increased for large value of Rg. When value of Rg is 100 O/km, the fault location algorithm, with 4 sensors uniformly distributed by 36 poles, can locate fault within a maximum error of 6 poles. However, for Rg of 250 O/km, the fault location algorithm, with 4 sensors uniformly distributed by 36 poles, can locate fault within a maximum error of 28 poles, but for the same value of Rg (= 250O/km), if 6 sensors are uniformly distributed by 21 poles apart, the fault location algorithm can locate the fault with a maximum error of 5 poles. This indicates that if high level of accuracy is desired, greater will be the number of sensors used for fault location, which consequently increased the cost for fault location.