IT ALL STARTED WITH A SIMPLE PING TO READ A METER. During a routine review of outage causes, PECO (Philadelphia, Pennsylvania, U.S.) personnel noticed that repair technicians had closed out several outages with the indication: Lights OK on arrival. The outage recording system was resulting in unnecessary trips by repair technicians. Measures such as outage verification calls to customers were already in place to minimize the potential for wasted crew dispatch. However, there continued to be a number of such events, so a better solution was needed.

PECO turned to its internal AMR Strategies Group to see if there was a way to leverage its new automated meter-reading (AMR) system to mitigate this issue.

PECO's decision to deploy an AMR system in 1999 was based on several factors, including improved meter-reading accuracy, cost savings and the ability to provide a broader range of customer services. While this decision anticipated there would be some potential benefit in using the AMR system for outage management purposes, there were no hard benefits or savings identified. PECO contracted with Cellnet (Alpharetta, Georgia, U.S.) for its AMR system.

PECO officially completed the deployment of 1.7 million electric meters and 470,000 gas meters in late 2004. Since then, PECO's Meter Reading Technology team has been focusing on new ways to leverage the AMR system and create new value from the data the system delivers. This effort has led to the successful development of a new suite of applications that link PECO's Cellnet AMR system with its Intergraph Public Safety Outage Management System (OMS).


The Cellnet AMR system is based on a radio-frequency fixed network. The network consists of nearly 100 CellMasters and over 8500 MicroCell Controllers (MCCs). Each electric meter has either an internal AMR module or, in the case of electromechanical meters, a retrofitted AMR module. All of the AMR meter modules at PECO are transmit only. Under normal circumstances, the electric meters are expected to be heard by the network every 5 minutes.

The MCCs are responsible for listening for the transmitted data packets from the AMR-enabled meters. Each received packet is processed and the meter-reading information is stored according to the service level of the meter. The CellMasters are responsible for polling and consolidating the meter-reading data from the MCCs. Each night, a process is initiated to retrieve all meter readings and meter data.

The network is also designed to receive and process special meter information packets, such as “last-gasp” outage messages and “power-up” restoration messages. This type of information is processed immediately and passed to the utility via Cellnet's network operation center.


The first application designed by PECO was a tool to pass on-demand meter-reading requests to Cellnet. Such requests are known as meter pings. This application design takes advantage of the expectation that each electric meter transmits data every 5 minutes. In this case, if a meter is heard from in the past 20 minutes, it is concluded that power is available to the meter. Conversely, if a meter is not heard from for at least 20 minutes, it is concluded that power is off. This logic has been validated via extensive testing and over 250,000 pings during the lifetime of the application.

Time Outage Notification Without AM R With AMR
Customers Affected OMS System Response Customers Affected OMS System Response
11:27 Call 1 1 Single Customer Outage 1 Single Customer Outage
11:30 Last Gasps 1, 2 and 3 12 Transformer Event 1
11:31 Call 2 2 Single Customer Outage 13 Single Customer Outage
11:34 Call 3 3 Single Customer Outage 14 Single Customer Outage
11:34 Call 4 4 Single Customer Outage 86 Transformer Event 2 and Fuse Outage
11:38 Calls 5 and 6 6 Single Customer Outage
11:43 Call 7 18 Transformer Event 1
11:49 Calls 8 and 9 86 Transformer Event 2 and Fuse Outage
This table details outage notifications for a sample fuse outage. The first data set illustrates that it would have taken 22 minutes to identify the cause of the outage, while the second data set shows that the AMR last-gasp data reduced the identification time to 7 minutes.

The first true test of the pinging application was during Hurricane Isabel in 2003, when a special team of engineers and analysts was set up to assess storm-damage reports and customer-outage calls via the new pinging tool. This team completed more than 2400 manual pings. Those pings led to the cancellation of more than 900 outage reports that would have otherwise been fielded. Additionally, nearly 100 single customer outages were escalated into primary events.

Subsequently, the single customer pinging application has been adopted for use during routine outages to assist in verifying outages. The dispatch office values this tool during late-night hours, because they use it to verify power status rather than calling customers. More than 100,000 pings are completed annually.


The success of the pinging tool led to the need for more robust tools. The next tool, labeled transformer analysis (TA), is designed to quickly assess whether any customers on a particular transformer have power available at the meter. If customers indicate they have power, it is concluded that the transformer is energized.

This is significant because PECO employs two types of crews to respond to outages. The first type is an energy technician (ET) skilled at secondary or low-voltage problems. The other is a trouble man trained and equipped to work on primary or distribution voltage issues. Prior to the pinging and TA tools, there would be occasions when an ET was dispatched, only to find out he could not correct the problem because it was on the primary side of the transformer. This would result in a wasted trip by the ET. The TA tool helps identify the true nature of an outage so a properly skilled crew can be sent the first time.

Other tools built included a circuit analysis application and batch pinging tools. The circuit analysis tool offers a quick assessment of a random sample of customers on a given circuit. If any pinged meter indicates power on, it can be reliably concluded that the circuit breaker at the substation is closed. The batch pinging tools provide the means to ping any group of meters, including all customers that have currently reported an outage. The batch tools helped to speed up outage assessment and gave a clearer indication of the current status of the distribution system.


While the manual pinging processes and tools proved to be effective in verifying outages and helping the dispatch office send the appropriately skilled crews to an outage, it quickly became evident that the manual processes themselves were inefficient. The next task was to automate the manual processes. In this case, all single customer outage calls not otherwise validated are automatically submitted to the batch pinging process every 20 minutes. The ping results are then fed into the analysis tools to determine if the outage is actually part of a larger unreported event. If the initial ping indicates power off, a new outage record is sent to the OMS. Conversely, if the ping indicates power on, the outage is automatically cancelled and an automated follow-up call to the customer is placed to notify them PECO has determined that power is available at the meter and would not be sending a repair crew.

This process takes advantage of the internal logic of the OMS. The system is designed to predict the true extent of a reported outage. When two customers on the same transformer report an outage, the single customer calls are converted into a primary transformer event. Similarly, if two transformers behind a fuse indicate an outage, they are rolled into a fuse event. This logic continues throughout the distribution system until the substation circuit breaker is reached.

The automated process is labeled reactive automation (RA) because it was initiated in response to an actual customer call.

Like the manual pinging processes, the RA process has been a resounding success. For 2005, PECO has evaluated over 31,000 single customer outages with the RA process. Over 7500 of those outages were cancelled because the meter indicated power on. Over 2750 outage reports were escalated into primary events. This equates to more effective dispatch for over 10,000 jobs.


The next step in the process of linking the AMR and OMS systems was to evaluate and implement methods to leverage the outage notification capabilities of the AMR system. It was noted that only about 20% to 40% of the customers who experienced an outage actually reported it to PECO. Many customers assume their neighbor will make the call.

Each electric meter in the Cellnet AMR system generates a last-gasp power-outage message when power is lost. An extensive analysis of the last-gasp messages was performed to ensure that quality outage information would be passed into the OMS. This analysis indicated that automatic outage notification could potentially shave minutes off outage durations by reducing the amount of time required to identify the true extent of an outage. However, it also was determined that several filters would need to be created to eliminate nuisance messages and system noise.

With a goal of reducing outage durations, PECO moved forward to implement an AMR-based outage notification tool. It was named “AMR Push” because the messages would be pushed to PECO from Cellnet and its AMR network. The AMR Push application is designed to work in concert with the previously deployed reactive automation processes. When a last-gasp is received from the AMR network, it is qualified and then passed to the OMS system in the same manner as any customer call. The net result is an effective increase in the amount of trouble reports to the OMS system. The additional information speeds the outage identification process and crew dispatch. Analysis times for typical fuse and transformer outages have been reduced by more than 15 minutes.

The AMR Push application was first enabled in August 2004. The application was rolled out with a challenge to the dispatchers to identify the first outage and have it completely restored before there were any customer calls to PECO. This occurred approximately three weeks later in September. There was a small outage on a transformer serving a residential cul-de-sac. The outage occurred mid-morning, though, apparently no customers were home to report the outage. A crew was dispatched and found that a transformer fuse had been blown because of an animal. In this case, the actual outage was not avoided, but the customers' impact was minimized and the restoration did not require customer input.

In another case, there was a weekend outage at a school facility. Without the outage notification, the outage would have gone unreported until Monday morning when the school's staff reported to work. Classes would potentially have been cancelled and the day significantly disrupted. The AMR system identified the outage when it occurred. A crew was dispatched and service was promptly restored.

Since the deployment of this tool, there have been thousands of outage messages sent to the OMS system. There have been many additional success stories that clearly demonstrate the proactive nature of PECO's outage management process.


The most recently deployed tools are focused on the power-up restoration messages. When power is restored, the AMR meter sends a restoration message that is passed on to PECO personnel. This message was found to be useful for many purposes, including a spot verification that power for all customers affected by an outage had been restored and confirmation of the actual outage restoration time.

Several reports were developed to present the restoration information data in convenient and useful formats. Two reports have shown particular value. The Push Data History Report shows all of the last-gasp outage messages and the power-up restoration messages associated with any outage. The Power-Up Grouping Report compares the recorded restoration time from the OMS system to the time reported by the AMR system. Through both a review of the past year's storm days and daily use of these two reports, PECO has been able to accurately correct restoration times, which in turn has lowered system CAIDI by several minutes.

With the power-up message reports in place, the project to link the AMR and OMS systems is now complete.


Even though the formal project is complete, new uses and applications of the AMR Push data continue to be found. Underground crews in Philadelphia have used the batch pinging tools to help identify the extent of failed secondary mains. During storms, back-office personnel have begun to use the tools to verify restoration work packages before they are dispatched to the field.

Future opportunities include the possibility of using the last-gasp and power-up messages to predict where failures are about to occur. It has been demonstrated that a number of what were once thought to be nuisance reports may actually be a series of precursor messages indicating that some device is flickering and/or failing. Ongoing study into these phenomena is occurring.

Another area of interest is using the AMR outage data to identify potential theft scenarios. By matching unexplained outages to changes in consumption and other AMR alarm data, revenue protection crews can be dispatched more effectively.

Throughout this journey, from simple manual meter pinging to automated outage notification and power-up reporting, PECO has created tools to continually improve customer service and its response to outages. The tools have proven to be effective and successfully demonstrate PECO's ability to know when power is lost via its AMR network.

Glenn A. Pritchard is the project manager responsible for linking PECO's AMR and OMS systems, and is currently part of Exelon's Meter Reading Technology team. He has been with Exelon and PECO for 14 years and his experience ranges from distribution automation to reliability engineering. Pritchard holds a BSEE degree from Clemson University, and is a registered professional engineer in Pennsylvania and a member of the IEEE Power Engineering Society.