The following technical papers supporting this presentation are available at www.selinc.com: Trip and Restore Distribution Circuits at Transmission Speeds by Jeff Roberts and Karl Zimmerman International Drive Distribution Automation and Protection by James R. Fairman, Karl Zimmerman, Jeff W. Gregory, and James K. Niemira Distribution Single-Phase Tripping and Reclosing: Overcoming Obstacles With Programmable Recloser Controls by Robert M. Cheney, John T. Thorne, and Greg Hataway. Implementation of a High-Speed Distribution Network Reconfiguration Scheme by Greg Hataway, Ted Warren, and Chris Stephens. 1
2
In this discussion of distribution power quality, we will: Review distribution service reliability measurements. Compare some common tripping and restoration schemes and show that faster is indeed better for most installations. 3
4
Distribution automation (DA) has been around for many years, and it simply means the automation of the distribution system. This automation may involve coordinating protection between two or more points on the system, automatic reconfiguration of the distribution system, or even something as simple as remote engineering access, control, and automatic collection of events. 5
In order to justify a DA system, a utility must have a reasonable return on investment. Generally, this return is most easily realized from improved reliability metrics. The primary goal of a DA system is to reduce customer outage times by maintaining high system reliability. There are several metrics used to track outages, including standard IEEE reliability indices. A DA system can reduce the opportunity for human error by performing the necessary decisions, verifications, and switching operations during a fault condition, a time when operations staff will be experiencing higher stress levels. By automating identification, isolation, and restoration operations in response to an outage, operations personnel can focus on other important tasks, such as dispatching crews to the known trouble spot to repair the faulted section. The crews will spend much less windshield time because they do not have to examine miles of distribution lines to find the faulted section. 6
There are several different philosophies when it comes to DA. They can generally be divided into two groups: centralized control, either at the substation or a control center, and distributed control. There are many variations of these two options, including the highspeed peer-to-peer system and the close-before-open (CBO) system that both fall under the distributed control type of DA system. Because different utilities have different requirements and other considerations on their systems, each utility will need to consider the benefits of each system type. A system that works well in a residential setting may not be the best choice for commercial loads. 7
With so many options in DA systems, utilities must consider a number of factors when choosing a system, including the following: Criticality of load. If a load cannot tolerate interruptions, then the best option may be a high-speed peer-to-peer system. Cost. Centralized control systems are generally more expensive because of the need for additional communications equipment, engineering services work, or other equipment. Distance between devices. Less distance between devices provides for better sectionalizing, results in fewer customers impacted by an interruption, but does add cost. Time to restore load. Reduced restoration time improves reliability indices but may require higher levels of automation or additional communications equipment. Type of switches applied. SEL equipment with MIRRORED BITS communications is well-suited for peer-to-peer applications. If equipment from other manufacturers is being used, it may be more practical to use a centralized control system that offers more flexibility in communications options. 8
9
The SEL Distribution Network Automation (DNA ) system is a perfect example of a centralized control DA system. The control may be located in a substation or placed in a central control center. Some utilities are using hybrid arrangements that have individual DA controllers in substations and a distributed DA system that incorporates into recloser controls in the event that communications are lost. 10
A DA control system collects data from many devices located throughout the distribution network. The DA controller can act as a remote terminal unit (RTU) in the supervisory control and data acquisition (SCADA) system. One extra node on the SCADA system can provide a wealth of information about multiple feeders to the distribution operations center. 11
Communications infrastructure is a major cost related to any automation system that services a wide geographic area. The application of automatic isolation and restoration is one of several wide-area applications that can provide value to the electric utility. Other applications include capacitor bank control, load balancing, volt/var optimization, demand management, and automatic meter reading. One of these applications alone may not provide enough benefit to the utility to justify the cost of the necessary communications infrastructure. However, by evaluating multiple applications including DA, the benefits become significant and outweigh the costs. 12
The distribution system has a wide range of applications for DA solutions. A simple application may include only three or four switching devices and two sources. However, there are situations that demand a more complex solution. Some applications will include four, five, or six sources and dozens of switching devices that must work together to minimize the impact of faults in the system. It is possible to design DA control systems that can be applied to simple as well as complex distribution networks. 13
There are constraints that must be respected by the DA control system. When the control system reconfigures the distribution network, loads are automatically picked up by adjacent feeders to return service to customers. The control system must avoid actions that result in overloading equipment or conductors. In larger systems, there are often multiple alternate feeds from which to choose. The control system must intelligently select the alternate feed that ensures the best operating margin. Modern distribution networks utilize pole-mounted reclosers to provide better selectivity in clearing faults. It is important to recognize that once the distribution network is reconfigured, the protection devices located throughout the network may not be coordinated in the new network configuration. The DA control system must attempt to preserve the coordination of protection devices, such as feeder breaker relays and recloser controls. Any automated system must hold the safety of personnel as a fundamental principle guiding its actions. Many line crew personnel may not be aware that a wide-area DA control system is in operation. Therefore, the DA control system must monitor typical indications of line crew activity, such as hot-line tag indication and nonreclose status from the recloser controls. Also, automatic actions at one switching device affect other areas in the distribution network. The DA control system must be selective to ensure the safety of personnel. 14
In order to meet the objectives of a DA control system, an accurate representation of the operating state of the distribution network is necessary. Fault current detectors allow the control system to determine which sections of the feeder are faulted. Fault detectors used in conjunction with lockout indications from upstream breakers and reclosers distinguish temporary faults from permanent faults. Voltage indications allow the control system to detect broken jumper conditions. Also, voltage indications provide verification that potential alternate feeds are energized and capable of picking up additional load. Load measurements allow the control system to calculate the amount of load that is deenergized and requires an alternate feed. Also, load measurements are needed to calculate the present available capacity for all potential alternate feeds. These data allow the control system to select the best alternate feed to re-energize customers. Abnormal condition indications, such as hot-line tags, nonreclose, diagnostic failures, and others, allow the control system to avoid undesirable actions. Status and control of switching devices allow the control system to reconfigure the distribution network. Status and control of settings groups allow the control system to preserve the coordination of protection devices after the distribution network is reconfigured. 15
The SEL DNA system responds to common failures on distribution networks and manages the situation in an intelligent manner to minimize customer outages. Permanent fault detection is based on fault current and lockout indications from switching devices on the feeder. The feeder is evaluated zone by zone, starting at the point farthest from the source. The fault current indications are used to identify the switching devices that have experienced fault current. When a fault occurs, many fault current indications may acknowledge the fault. The fault indication farthest from the source will identify the affected zone. Broken jumper detection is based on loss-of-voltage indications from switching devices on the feeder. If a fault current indication is present anywhere on the feeder, then the DA controller does not attempt to evaluate for broken jumpers. If no fault current indications are present, then the feeder is evaluated zone by zone, starting at the substation feeder breaker. Similarly, loss-of-source detection is also based on voltage indications from switching devices on the feeder. The control system considers operating conditions, including protection system miscoordination, hot-line tags, and load levels, to determine the most effective and safe course of action. 16
The most common service reliability and quality indices are the following: SAIDI (System Average Interruption Duration Index). This is the total time the average customer does not have power in one year. It is measured in minutes per year. SAIDI = (Sum of all Customers Interruption Durations)/(Total Number of Customers Served) SAIFI (System Average Interruption Frequency Index). This tells how often the average customer s service is down, measured in interruptions per year. SAIFI = (Total Number of Customer Interruptions)/(Total Number of Customers Served) CAIDI (Customer Average Interruption Duration Index). This measures the average time it takes to restore power for a given group of customers, measured in minutes per interruption. CAIDI = SAIDI/SAIFI ASAI (Average Service Availability Index). This measures the percentage of time that service is available on average. ASAI = (8760 SAIDI)/8760 MAIFI (Momentary Average Interruption Frequency Index). This is the number of momentary interruptions experienced by the average customer per year, measured in momentary interruptions per year. This should be applied at the customer level, especially for industrial customers with sensitive equipment. SAIDI, SAIFI, CAIDI and AISI measure long time interruptions such as 5 minutes or longer. MAIFI is related to short time interruptions. 17
The table on this slide shows data taken from a 1990 survey. SAIDI is a measure of service availability duration. It states that the average utility customer in the United States has been without electric service for an average of 96 minutes per year. SAIFI is a measure of service outage frequency. It states that customers have an average of 1.18 outages per year. CAIDI is SAIDI divided by SAIFI. The average outage lasts 77 minutes. ASAI shows an overall service availability (i.e., per-unit value of service availability). Based on the numbers on this slide, electric service reliability appears excellent. However, the survey indicates that most utilities did not consider outages shorter than five minutes. It is these shorter outages that contribute to common power quality problems, such as voltage sags. No matter how high we believe an existing service reliability to be, the increasing sensitivity of loads to voltage sag should motivate us to review new methods of improving service reliability. These new methods must reduce fault duration and minimize voltage sags on unfaulted circuits. 18
To compare the reliability of distribution service, we consider two cases. Case 1 shows a system with two breakers, 1 and 3, and three manually operated switches, 2, 4, and 5. SW5 is shown as normally open. In this case, if there is a fault on any of the line sections, tripping and reclosing are performed by the breakers. For example, if a permanent fault occurs on Line 1, Breaker 1 trips and proceeds through its reclose cycle to lockout. An operator must manually open SW2 and close SW5 to restore power to the customers on Line 2. 19
Case 2 shows a system with breakers and/or reclosers at positions 1 through 5. Each switch is capable of interrupting fault current. Further, we add a relay, or control, to each switch and a communications line between each relay. This allows SW2 to be opened automatically and SW5 to be closed automatically, thereby eliminating the additional time to restore load if there is a permanent fault on Line 1. Using intelligent devices that communicate with each other allows the devices to quickly determine how to reconfigure the system to isolate the faulted line and restore power to unfaulted lines. 20
For Case 1, we assume that it takes 0.5 hours to operate each switch manually. The time includes travel time and any time needed for conversations with system operators. Thus, for the permanent fault on Line 1, we assume a 1-hour restoration time. For Case 2, we assume a 5-second delay to send a trip signal to SW2 and a 1-second delay to send a close signal to SW5, for a 6-second restoration time. These times can be faster depending on the communications and switch equipment. The 5 second and 1 second time delays used are simply an example. Next, we compare the reliability of these two schemes. 21
To compare the reliability of the two schemes, we calculate the unavailability of load served to Line 2 customers for permanent faults on Line 1 or Line 2 using fault tree analysis. The unavailability of service is defined as the mean time to repair divided by the mean time between failures. For our example, we made the following assumptions: 0.2 permanent line faults per year 3 hours to repair the line 0.01 breaker or switch failures per year 1 hour to repair a breaker or switch We used the IEEE Gold Book for the reliability data. If the utility has more accurate data, substitute those numbers in the fault tree calculations, which can be found in Appendix I of the paper Trip and Restore Distribution Circuits at Transmission Speeds. To illustrate this equation, we assume it takes 3 hours to repair a line after a permanent fault, and 0.2 permanent faults occur per year (MTBF = 1/0.2 = 5). Thus the unavailability is 3 hours divided by 5 years. 22
The diagram on this slide is the fault tree for Line 2 load not served because of permanent faults on Lines 1 or 2. All of the unavailability numbers are multiplied by 10 06. The OR gate simply sums all of the unavailability values: Line 1 fault with switches OK, SW2 bad, and SW5 bad and Line 2 fault with switches OK and SW2 bad. The value 140 is not an absolute reliability but is useful when making comparisons to other schemes to show relative reliability. 23
For the same conditions as on the previous slide, the automatic scheme has an unavailability of 86, which is better than the 140 obtained with the manual scheme. This improvement results from the time saved by using intelligent devices and communications to trip and close switches instead of relying on manual operation. 24
25
The figure on this slide is a power acceptability curve. The shaded region above and below the lines defines the unacceptable voltage variation region (UVVR). For example, a load defined by this curve can withstand an 80 percent voltage for 0.5 seconds. These power acceptability curves are also referred to as CBEMA (Computer and Business Electrical Manufacturers Association) and FIPS (Federal Information Processing Standard) curves. There is no universal standard for power acceptability curves. Reasons for this include the following: Different loads have different tolerances-to-voltage variations. This means that we cannot use a standard curve. Power acceptability curves do not consider multiple voltage variations that occur in rapid succession. A single voltage variation may be tolerable but a second voltage variation that occurs very close to the first may not be tolerable. Power acceptability curves also do not account for load recovery time. 26
In the IEEE Gold Book, a survey shows that 25 percent of industrial plants must completely restart production if service is interrupted for greater than 10 cycles. The same survey shows that the average restart time is 17 hours. This should motivate us to find ways to trip and restore load faster. 27
The diagram on this slide shows a system one-line diagram for a rural distribution feeder. To meet the coordination requirements, Relay 1 needs to coordinate with the slowest downstream device, which in this case is Relay 2. Likewise, Relay 2 must coordinate with its slowest downstream device, which is a 50T fuse. 28
The chart on this slide shows time current characteristics for the system. Note that Relay 1 coordinates with Relay 2 and the 50T fuse. However, for faults on the Line 1, Relay 1 provides slower clearing than would be possible if the relay only needed to coordinate with the 25T fuse. As shown, Relay 1 trips in about one second for a 2000 A fault. 29
Communications between Relays 1 and 2 reduce tripping times for Line 1 faults. For example, Relay 1 no longer must time-coordinate with Relay 2 if communications are present. Instead, Relay 1 now must only coordinate with fuses tapped off of Line 1. In this scheme, Relay 1 uses two time-overcurrent elements one that coordinates with Relay 2 if the communications channel is not in service and another that coordinates with the 25T fuse when the channel is available. The communications channel and the supporting logic in Relays 1 and 2 allow Relay 1 to discern when a fault is downstream from Recloser R. For faults downstream from Recloser R, Relay 2 senses the fault and instructs Relay 1 not to trip using its fast timeovercurrent element. If Relay 2 does not sense a fault in the forward direction while Relay 1 does, then the fault must be on Line 1 or on a Line 1 lateral. For such faults, Relay 1 does not receive a block signal and is permitted to trip using its fast time-overcurrent element. With this scheme, the Relay 1 fast time-overcurrent element is only required to time-coordinate with the 25T fuse. To handle contingencies such as a communications failure between Relay 1 and Relay 2 or failure of the downstream recloser, some level of backup protection may be desired. For these reasons, the time-delayed curve for Relay 1 can also be left enabled as backup protection. 30
The chart on this slide shows the improved time-current coordination. Note that the trip time is reduced to about 0.2 seconds for a 2000 A fault. This is a significant reduction from the 1-second trip time without communications. It also reduces voltage sag on this feeder and adjacent feeders. 31
Next, we examine the implementation of the two schemes: A conventional scheme using relays or controls with standard voltage transformers and some associated logic. A communications-enhanced scheme that allows communication between each of the protection devices. 32
Here is a one-line system diagram. SW1, SW2, SW3, and SW4 are normally closed. SW5 is normally open. 33
The table on this slide shows the switching operations required to restore load for a particular faulted line section. 34
One method of reducing restoration times and directly improving traditional reliability data (SAIDI, CAIDI) is to use conventional microprocessor-based relays and/or recloser controls at each switch location with voltage signals supplied from voltage transformers on each side of the switch. This enables us to detect hot/dead voltage conditions, thereby allowing automatic tripping and restoration of switches and improving speed and reliability. 35
The equipment requirements and capabilities for the conventional scheme are described on this slide. 36
In this example, each relay uses voltage elements to declare dead or hot voltage (e.g., DL indicates dead voltage on Line 2 and HL4 indicates hot voltage on Line 4). The top portion of the figure on this slide ensures that Line 2 load is restored after a Line 1 fault. This occurs when: Voltage was initially hot on both sides of SW2 (HL1 HL2). Voltage goes dead (DL1 DL2). SW2 is closed. A 50 element at SW2 is not picked up (a security check for bolted three-phase faults). Conditions 1 through 4 are true for tt2 time (which is greater than the maximum total reclosing time for SW1, about 90 to 120 seconds, depending on reclosing delays). The lower portion is to restore Line 1 for Line 2 faults. If the 50 element at SW2 picked up and dropped out twice and conditions 1 through 4 are true for tt2a time (set less than the third reclosing interval at SW1, about 5 to 10 seconds), then trip SW2. If the fault is on Line 1, we trip SW2 after SW1 has completed its reclosing sequence. If the fault is on Line 2, we trip SW2 before SW1 advances to lockout. 37
The closing logic for SW5 requires that the following conditions be true for tc5 time: Voltage is initially hot on both sides of SW5 (HL4 HL2). Voltage goes dead on Line 2 (HL4 DL2). SW5 is open. SW5 did not trip after a restoration close attempt. Set tc5 greater than tt2. 38
Finally, if the fault on Line 2 is permanent, we need to avoid another reclosing sequence from SW3 after SW5 has been closed. Thus, if the voltages are dead (DL2 DL4) after a close attempt by SW5, the logic trips SW5 before SW3 recloses. We also need to put a short time delay for the first reclosing attempts at SW1 and SW3 to allow time for a SW5 trip after a close attempt. The disadvantage of this scheme is that it is possible to close into a permanent fault on Line 2 and momentarily disrupt service to Lines 3 and 4 when we attempt to restore the load on Line 2. 39
The diagram on this slide shows a communications-enhanced restoration scheme. 40
41
Fast tripping is enabled for SW1 if the communications circuit is healthy and SW2 is open. If SW2 is closed, the instantaneous element of SW2 will block the fast tripping of SW1 for a fault on Line 2. 42
For permanent faults on Line 1, it is desirable to open SW2 and close SW5 in order to restore power to the customers on Line 2. This is accomplished by noting that the instantaneous element of SW1 was picked up via the rising edge, but is now not asserted, and SW1 is open. This also requires the communications circuit to be healthy. 43
After SW2 is open, SW5 is closed to restore power to Line 2. This is accomplished by checking that SW2 is open, verifying there is no current flowing through SW2, and that the SW1 instantaneous element has operated. As before, the communications circuit must be healthy because it is the means by which the information is passed from device to device. 44
For a fault on Line 1, we wish to restore the critical load on Line 2 as quickly as possible. When Relay 1 senses the fault, it sends a signal to Relay 2, which in turn sends a signal to Relay 5. Relay 2 trips SW2 and Relay 5 closes SW5 as quickly as possible to restore power to the critical load. Meanwhile, Relay 1 can trip via its own fast curve to clear the fault. 45
The diagram on this slide shows the timing diagram for the fast trip and restoration. How fast can we trip and restore load to Line 2 after a fault on Line 1? Can we meet the 10-cycle service interruption requirement. Relay 1 overcurrent elements can be set sensitively to operate in 0.25 to 0.5 cycles. If we apply optical fiber, the communications delay to send the signal to Relay 2 (Comm 1-2) is negligible. Assuming it takes 0.25 cycles (not shown) for Relay 2 to process the signal, then, the Comm 2-5 time is virtually zero. It takes Relay 5 another 0.25 cycles (not shown) to process the signal, then it issues a close signal to SW5. Thus, it takes about 1 cycle (0.5 + 0.25 + 0.25) from fault inception to close assertion at SW5. This leaves 9 cycles for SW5 to close to meet the 10-cycle criterion. If we use spread-spectrum radios instead, we must allow about 20 milliseconds each for times Comm 1-2 and Comm 2-5. Thus, it takes about 56 milliseconds (20 ms + 20 ms + 1 cycle) or about 3 1/2 cycles from fault inception to SW5 close assertion. 46
Observation of the restoration times for the conventional and communications schemes, shows that: For a fault on Line 1, the conventional scheme can improve the traditional indices, SAIDI and CAIDI, for load served to Line 2. SW1 trips and recloses three times to lockout, then timer tt2 times out to trip SW2, and tc5 times to close SW5. Depending on the reclose open interval times, we can usually restore load in 1 to 2 minutes. For the communications scheme, SW1 also trips and recloses three times to lockout. However, we can significantly improve power quality by reducing the restoration times of Line 2 load to delays close to 10 cycles (SW2 trip, SW5 close). 47
If the communications link fails, we resume normal coordination. If the system changes (for example, SW2 is normally open), we can change setting groups to accommodate this arrangement. 48
Conclusions: Using communications-assisted protection and control schemes for distribution circuits significantly reduces trip and load transfer times. Traditional performance indices do not consider the reduction in service reliability caused by fault-induced voltage sags. Considering the effects of these sags on customer loads in the immediate vicinity of a fault, we conclude that we must also consider new protection and control methods that reduce sag duration to cycles instead of seconds. Fault tree analysis shows that upgrading breaker and recloser controls with communications scheme logic realized a 40 percent improvement in service unavailability (compared with traditional distribution protection and control). Communications-assisted trip logic simplifies difficult time-coordination applications by limiting the number of devices requiring coordination. This simplification also decreases tripping time for main-line faults. Without a communications channel, we can apply relay logic that combines voltage elements, switch status, and other logic to improve service reliability for unfaulted feeders served by a faulted source. 49