March 29, 2024

Non-Operational Data Can Provide Valuable Benefits to Utilities That Exploit It
Georgia Power Kicks Off Pilot to Investigate Non-Op Data Automation greater efficiencies

by Mitch Cowan
Like money left on the table, too many utilities are failing to retrieve and analyze all of the valuable data collected in their substations. With SCADA doing an adequate job of reporting substation events, utilities have not been proactive in attempting to mine digital fault records, transformer gas measurements, lightning strike archives and other critical data sets that can help them fully understand why these events have occurred.

In recent years, interest in these long-ignored data sets has slowly built to the point where the term ‘non-operational’ data has been adopted. Although some variations in definitions still exist, non-operational data consists of the records of power fluctuations, current loads, voltage levels, fault events, breaker positions, transformer health and environmental conditions. These are typically stored in non-point formats, such as oscillating wave forms, not designed for transmission via SCADA protocols, which means they must be retrieved manually.

Conversely, operational data generally refers to the instantaneous measurements of volts, currents and breaker status transmitted in near real-time by the SCADA system to the control center. These data sets are often linked to alarms or automatic control devices and stream continuously from substation.

The irony surrounding the non use of nonoperational data is the fact that most utilities have already built vast storehouses without realizing it. Non-operational data is routinely recorded in digital fault recorders, intelligent electronic devices, protective relaying devices and a host of other automated equipment installed in many substations. Some utilities have a decade worth of historical non-operational data in their substations, capable of providing valuable insight into equipment performance and reliability.

A handful of utilities, including Georgia Power, tap into these applications manually when fault events occur. But the lack of an automated technique to easily access, integrate and analyze these data sets has prevented most utilities from using them on a regular basis. This tradition of non use, however, is about to change.

In early 2003, Georgia Power participated in a study with Kreiss Johnson Technologies of San Diego that revealed these data sets can indeed be retrieved and analyzed in an automated fashion. As a result of this study, we are now moving ahead with a substation pilot to further demonstrate how the data can be accessed, interleaved and delivered as useful information to multiple departments within the utility.

Based on our experiences, Georgia Power believes these data sets – especially when analyzed together – have great potential in identifying fault sources more quickly and recognizing the early warning signs of system weakness. By tapping the full power of this data, utilities can potentially restore outages faster, plan maintenance more accurately and replace malfunctioning equipment before a major event occurs.

Tapping Non-Op Data
Georgia Power, a subsidiary of Southern Company based in Birmingham, Ala., provides electric power to 2 million customers spread across the entire state of Georgia. For nearly 20 years, Georgia Power has investigated the practical uses and potential benefits of non-operational data. We are among a small group of utilities that actually makes regular use of this data, but believe much more can be done with it.

Our two primary sources of non-operational data are digital fault recorders (DFR) and a lightning strike database. For DFRs, we have chosen products from Utility Systems, Inc. a division of Magnetic Instrumentation, Inc. because they provide virtual channel capabilities allowing us to monitor a larger number of lines at a lower cost. The lightning data comes from Vaisala – GAI, Inc. of Tucson, Ariz., as part of a subscription service.


In the past, Georgia Power relied solely on smart relays to collect fault data, but we have found these devices do not perform as well as DFRs. The primary difference is the sample rate. The smart relays on system sample anywhere from 4 to 32 samples per cycle, while DFRs run at 80 samples per cycle. Much more information is captured at the higher sample rate. As a result, we have installed DFRs at all switching stations rated at 500 kV or higher and all 230 kV plants and other switching stations considered critical to observe a system disturbance.

Georgia Power has also invested in event recorders for installation in the same switching stations as the DFRs. The event recorders monitor breaker status, primary and secondary relay outputs and other miscellaneous substation alarms. Today these devices serve as back-ups to the DFRs. They were installed several years ago as supplements to DFRs when earlier versions of fault recorders were incapable of capturing all of the necessary data points.

It is important to note that we have not abandoned SCADA. Georgia Power has implemented a SCADA that scans each substation every six seconds. This interval is sufficient to notify the control center that a fault has occurred, but it cannot define the series of events that led to the fault. Many things happen in six seconds, and operational SCADA data cannot pinpoint which breaker tripped first, where it tripped or precisely when it tripped.

Our primary goal in harnessing fault and lightning data, therefore, is to fill in the blanks in SCADA data to quickly and accurately identify the fault event – was it an over-current or undervoltage condition, for example – and determine where it occurred. This basic fault information assists the control center in determining how to handle the situation. It helps them decide whether to clear the fault remotely or send a field crew to perform equipment maintenance. And if the field crew must be dispatched, the control center knows where to send them.

By collecting and integrating this nonoperational data for analysis, we are turning a substation fault event into what Kreiss Johnson Technology’s call a Value Event – a situation where the additional data analysis enables us to restore service quickly, better maintain equipment and prevent small system problems from cascading. These value events positively impact the bottom line.

Analyzing Data Manually
By installing this network of DFRs and subscribing to the lightning strike service, Georgia Power has created a manual system for accessing and analyzing non-operational fault data. In a typical fault scenario, this system is activated when the SCADA detects an event at one of the substations. This triggers an alarm at the control center, where a Transmission Operator then places a call to me or another of our transmission specialists asking where the fault occurred.

The first step for the transmission specialist is to check weather reports and determine if storms are in the area. If weather is present, we then access the live Vaisala database to see if lightning may be the cause. This is accomplished by correlating the timing of the fault with those of lightning hits. If a correlation is established, we can use the lightning data to pinpoint the fault location to within 500 meters. This information is relayed to the control center so that a truck can be dispatched to inspect this section of the line.

In approximately 40 per cent of fault events, lightning is quickly ruled out. In that case, our job is to call the DFR at the substation in question and download the fault data. We can examine certain characteristics of the fault by viewing its wave form on screen. We try to interpret the following:

  • Fault inception time – By pinning the time down to a millisecond, we can more accurately correlate with lightning strikes.

  • Fault amplitude – Our protection group reviews these measurements and verifies that our transmission system model is performing correctly and protective devices are properly set.

  • Voltage dips – Georgia Power provides this information to large electricity customers to determine if the dip impacted their activities.

  • Fault duration – By analyzing amplitude and duration of faults, we can determine whether a breaker operated properly. Duration can also indicate a low- or high-impedance event.

For immediate action, fault location is the most important characteristic for us to ascertain. The newer DFRs can automatically calculate which line the fault occurred on and its distance from the substation. The transmission specialist takes the distance information and pulls the transmission network map up on a computer screen. Using the distance as a guide, we can usually measure the fault location to a precise transmission structure, which is identified by number on the map. This information is relayed to the control center transmission operators so they can decide whether to clear the fault remotely or dispatch a crew to the scene.

Depending on the DFR download time, this retrieval and analysis process requires 30 minutes to two hours for completion.

Georgia Power’s use of non-operational data has accelerated restoration times, but we see several other value events coming from this data as better technology becomes available. If the entire retrieval and analysis process can be automated, the fault location information could be fed continuously, perhaps by email or a web page, directly to the control centers. Instead of waiting hours for manual interpretation of the DFR data, transmission operators could act on it as soon as the event occurs.


Automating the Process
With so few utilities pursuing the nonoperational data issue, Georgia Power assumed we would have to develop the technology required to break down the barriers to data access and integration. In early 2003, however, we were introduced to Kreiss Johnson (KJT), a technology software company that has been working on a solution to allow utilities to leverage this data. We agreed to conduct a joint study using the KJT software.

Georgia Power provided a database of historical fault records downloaded from a Georgia Power DFR. This data related to faults whose data we had already manually deciphered. We asked the software developer to run the fault data through its automated analysis routine to determine fault inception time, phase involvement, voltage dip, clearing times, clearing results, restrike occurrence, and fault current.

We were not expecting the results we received. After two weeks of analysis, KJT provided us with a report of each fault event, complete with oscillography, answering the questions we had posed. We were impressed with the results, which convinced us that automated analysis of DFR data was possible. Immediate plans were made for a field pilot.

For the study purposes, we had intentionally limited the focus to DFR data only, but data integration is clearly where the greatest potential for non-operational data lies. So the pilot will introduce a second data set – lightning. We have selected a very long line in South Georgia that gets numerous lightning strikes. In cooperation with USI, Georgia Power will place a DFR at either end of the line to record events that occur.

Instead of communicating with these DFRs through dial-up connections, the utility will establish a live high-speed link that will allow them to quickly feed data back to our Atlanta headquarters. A direct lightning strike feed from Vaisala will also be established. KJT will install its nonoperational analysis software on Windows computer at headquarters that will be receiving the DFR and lightning data streams.

In this automated pilot, we fully expect the software to receive the fault data, correlate it with lightning strikes and determine the same fault characteristics as we now do manually. Once this aspect of the technology has been field tested, we expect that nonoperational data use will quickly move into the mainstream at Georgia Power and other utilities.

In the near future as we envision it, information of much greater value will be extracted from the DFR and lightning data when these data sets are integrated with power quality, dips, and flicker data from line panel meters, and with dissolved gas, temperatures and vibration reports from transformer monitors. Combined analysis of these and other non-operational data sets will allow control room personnel to narrow a fault down to its precise cause within a piece of equipment and determine if it requires maintenance.

Exploitation of non-operational data in these and other applications will directly impact a utility’s bottom line by accelerating restoration time, keeping equipment more well maintained, and reducing the field time of crews. Most important of all, early identification of minor system or equipment problems can avoid major outages such as the blackout of 2003.