Bridging the Knowledge Gap to Identify Energy Theftby Michael Madrazo, CEO, Detectent
The majority of theft occurs in the residential sector, but the majority of revenues lost to theft — estimated at between one and three percent of total distribution revenues — occur in the commercial account sector. A utility with $1 billion in revenues potentially loses between $10 million and $30 million each year to theft, and the majority of the lost revenue is to a relatively small number of commercial accounts.
Traditionally, the utility has relied on meter readers and service personnel for tips on energy theft. Meter readers, visiting each meter every month to collect a reading, made for a very effective line of defense against theft. With the advent of Automatic Meter Reading (AMR) and Advanced Meter Infrastructure (AMI), the manual task of meter reading is being phased out and replaced by electronic readings so the utility has lost its eyes and ears in the field. No one is keeping an eye on the utility’s cash register.
Automated Meter Reading systems provide tamper flags that are designed to identify energy theft when the tampering occurs at the meter, and while they can alert the utility to tampering, they also trigger far too many false alarms. According to a Chartwell research report, “AMR is not a perfect system for detecting theft and many customers have discovered this, coming up with clever ways to bypass or tamper with the meter without triggering a tamper flag. Plus, some of the utilities consulted for this report say tamper flags do not always indicate theft, but are a product of oversensitivity and benign outside forces.”
Unfortunately, tamper codes are often triggered by valid field activities or external forces and so valid flags are lost in the myriad of false ones. As a real-life example, a utility with 128,000 electric meters on a fixed network AMR system received over 1,600 tamper codes daily. Clearly, there is a need for a way to use the valuable information supplied by the automated meter reading systems in an efficient and intelligent way. But harnessing the flow of tamper flags is just part of the solution. A broader more encompassing solution is required if the utility is going to safeguard itself against all forms of energy theft. The complete solution requires the utility to understand how each of their customers uses energy and focus on those that deviate from expected usage patterns.
Know Thy Customer
Knowing the customer is vital. Typically, utility data consists of the account holder’s name, phone number, address and for commercial accounts, type of business and maybe the name of the business. The utility may know a client account is a restaurant, for example, but probably doesn’t know how large the premise is or whether it’s a sit-down or take-out establishment. For commercial accounts, additional information such as the number of employees, chain affiliation and other metered services on the account need to be a part of any intelligent usage analysis. For residential accounts, knowledge of the size of premise and heating and cooling sources is also very valuable.
Understanding the characteristics of an account provides valuable insight into the customer’s use of energy. With this insight, energy usage can be compared to a group of “like peers” with similar characteristics and outliers can be flagged for further analysis and investigation.
The process of detecting theft analytically can only truly be accomplished once multitudes of characteristics of an account are identified and proper peer grouping is done. Useful data for analysis includes:
• Usage from other meters for the same service type
on the account
• Usage from other metered services
• Correct business codes
• Square footage
• Business information such as number of employees and hours of operation
Case in Point
Detroit Edison has been addressing energy theft issues since the 1930s - the difference today is the magnitude of the problem. With DTE Energy serving over 3.4 million customers, deploying the right technologies and processes to combat theft is crucial. “First and foremost addressing energy theft is important to DTE Energy due to the potential safety hazard it creates. This illegal activity can create safety hazards for both the residents at the location engaged in the theft and those in the surrounding neighborhood should an explosion or fire result,” said Mark C. Johnson, revenue protection manager for DTE Energy. “Product losses can also impact DTE Energy’s earnings and the rates paid by our customers.”
As part of stepping up their efforts to combat theft, the utility is now utilizing sophisticated Customer Intelligence Solutions from California-based Detectent, Inc. By complementing DTE Energy’s revenue protection processes with Detectent’s peer analysis and characteristic analysis, DTE Energy is able to identify a wider spectrum of theft cases.
Using information from many sources, advanced analytics and proven processes, utility resources can focus on cases that have the highest likelihood of uncovering theft and the maximum potential for revenue recovery. This repeatable process allows DTE Energy to collect more revenue and obtain more intelligence about their customers, year over year. The result is a more efficient use of back-office and field resources and increased revenue.
Utility customer data can be significantly enhanced and improved by acquiring third-party data and integrating it using sophisticated pattern matching algorithms. Correct business codes as well as a myriad of additional premise and operational information can be gathered and thereby dramatically extend the known customer information for each account.
Energy theft detection models generally fall into two categories: peer comparison and characteristic analysis. Peer comparison models contrast all available information about residential and commercial customers to similar homes and businesses within similar geographical and environmental settings. Deviation from expected usage can indicate that not all the energy used by a customer is being metered correctly. Characteristic analysis looks for anomalies in a customer’s consumption pattern that might be indicative of un-metered equipment in an account.
For example, you can expect a Laundromat to have a relationship between electricity consumed by washing machines and gas consumed by dryers. If one service is not metered to the expected ration of the other, then it may be indicative of one of the services not being metered correctly or having been tampered with. When combined, these two types of theft detection models can monitor for adherence to peer usage and micro-analyze energy usage for expected characteristics.
Proper Analysis is Key
Analysis on its own does not totally replace the need for the common sense and intuition that people bring to an equation. So even with all of the known data captured and analyzed, a review of all available information needs to be done in order to confirm that the indications of energy theft are not in fact simply the work of other outside forces. For example, one needs to consider if demographically, the area has declined, or if the business is going through renovation or other legitimate changes that might lead to a deviation from the expected normal usage.
Resources can expend energy where the payback is likely to be the highest. Unlike simpler query tools, the energy theft detection solutions that have emerged in recent years have the capability for organizing cases in order of value and probability so that both back office and field resources are used most efficiently. For commercial accounts, this might be accomplished by determining:
• Average consumption compared to the capacity of the installed meter(s)
• Ratio of one service’s consumption to that of another
• Degree of deviation from expected normal values
Using many data sources and a combination of models that look for independent features in a customer’s consumption profile have transformed theft detection into a viable and cost-effective solution for utilities. Previous attempts to analytically identify energy theft resulted in marginal improvements over past practices and typically were not cost effective. The newer solutions, which have emerged only in the past three to five years, have significantly increased in cost effectiveness.
To illustrate this point, a recent case featured on Fox-TV in North Carolina involved a large restaurant that was paying for about 1000 kilowatt hours of electricity on average, per month. In reality, the restaurant was consuming in the range of 10,000 to 12,000 kWh per month. The average monthly bill that the customer was paying was in the range of $150 to $180 while the amount should have been at least $1300 to $1500. This had been the case for more than eleven years.
The cause of the problem was a wiring fault and not an actual theft. But did the customer know that he was underpaying? Did the proprietor know that something was amiss after the monthly electricity bill miraculously dropped by 88% in one month and then stayed at this much lower level?
One can assume that the proprietor did in fact know that the bill was far too low, but chose not to draw attention to it. The case came to light in the summer of 2009 after the utility deployed an analytic theft detection solution from Detectent and began making use of peer comparison and characteristic analysis. The account, once other external data was gathered and matched to the information from the utility’s Customer Intelligence System, was compared to a group of similar restaurants and immediately jumped to the forefront.
The utility corrected the problem in the field, recognized the loss of over $171,000 and billed the cost for 12 previous months of unmetered usage to the amount of $15,869. Although significant revenue was unfortunately lost and ultimately passed through to the ratepayer base, the utility has avoided future losses by identifying and correcting this situation.
Compare, Contrast and Learn
With an analytical approach to energy theft, it’s important to know that no one analytical model stands alone. Numerous models need to run in parallel in order to evaluate an account’s energy usage from a variety of angles and flag anomalies based on different forms of assessment. Today’s energy theft detection systems do just that and therefore go well beyond the utility’s traditional high/low and zero-consumption reports.
Deploying an energy theft detection solution is not only important to the future of a utility with an AMI or AMR deployment; it is also critical to traditional utility operations. Economic conditions have caused utility customers – both residential and commercial – to act in ways and do things that they have not done in the past. All indications are that energy theft in both the residential and commercial sectors is rising rapidly across the country. As well, with revenues dropping from reductions in industrial energy usage, remediating theft conditions and other conditions causing revenue loss can be a valuable stream of revenue to the utility.
As with the restaurant example above, only by truly understanding the customer and how they use energy can utilities expect to detect cases of theft and other forms of revenue loss. Utilizing sophisticated customer intelligence tools, today’s utilities can identify and reduce future revenue losses by analytically and proactively protecting their power delivery networks from tampering, malfunctions and theft.