On September 1, the Northern California electric utility PG&E posted a 414-page “Metering Assessment Report” of an independent study of the performance of its smart meters on its Web site. The California Public Utilities Commission (CPUC) commissioned the study, which was carried out by Structure Consulting Group LLC of Houston, Texas.
The report was in response to numerous customer complaints about perceived large increases in electric bills that correlated with the installation of smart meters at various locations around California. It details the testing and customer data analysis that Structure conducted and concludes, in a diplomatic way, that the meters are accurate, but PG&E’s customer relations were flawed. That’s the short version. The devil, as they say, is in the details.
As I wrote in June, electricity consumers in some communities, particularly in Fresno County, were gathering pitchforks and torches and preparing to march on PG&E’s castle (see “Smart-Meter Rollouts And Standards Stir Controversy”). Statewide, PG&E had already installed 5.5 million electrical smart meters and was adding around 10,000 a day. The San Jose Mercury News reported that the number of customers with smart meters who received inaccurate charges “may be as high as 23,000.” As one citizen complained to ABC affiliate KFSN at a town-hall meeting to protest higher bills, “How did my bill go from $300 to almost $1100?”
The citizen went on to say, “It didn’t make sense,” but perhaps it did. The bigger increases corresponded to the end of summer in 2009, and that was an especially hot summer in the San Joaquin Valley. At the same time, CPUC had allowed the utility, which charges based on a five-level rate system, to radically increase the rates at the top levels, a change that was implemented with very little public awareness.
Under the new rates, at level 1, Fresno customers paid only 11 cents per kilowatt-hour used, but at level 5, they were shelling out 44 cents. On hot days in the San Joaquin Valley, with the air conditioning on, usage could quickly escalate into the higher tiers. That was the situation the Structure engineers walked into.
The Report
Structure broke down its assignment into five parts related to the complaints: testing samples of meters in the lab and under field conditions, testing the entire system (meters and radios) end-to-end, analyzing the actual situations of residential customers who had complained to the CPUC and/or PG&E about high bills, and analyzing the utilities conformance with industry best practices with respect to smart meters. In addition, they analyzed the security framework for the communications link from the meters to the utility.
Lab And Field Testing
The tests were performed at an outside laboratory. The original sample comprised 156 smart meters selected at random from five of PG&E’s warehouses. Some of the meters were General Electric I-210+ units. Others were Landis+Gyr Focus AXR-SD units. Both used Silver Spring network interface cards.
After the ambient temperature testing, six meters were further tested at high (50°C) and low (-20°C) temperature. At ambient temperature, testing at full load, light load, and with a 0.5 power factor, 100% of the sample was within an accuracy range of 99.81% to 100.15%, with an average accuracy of 100.01% and a standard deviation of 0.0408%. The CPUC accuracy requirement for all meters is ±2.0%.
Continue to next page
After “soaking” at high and low temperatures, all six meters in the environmental-test sample still met the +2.0% standard, but one of them flunked its manufacturer’s more stringent 0.5% accuracy standard by 0.07% at high load and high temperature. They all passed under other test conditions.
Those were the lab-test results. The field tests were conducted on customer meters, both smart meters and conventional reluctance meters. Structure used dual sockets, with a measurement instrument in one and the test sample in the other. The sample size was 760 units, 147 old-fashioned reluctance meters, and 613 smart meters. (I can’t make my totals agree with the table on page 25 of the report, so I’m using the numbers in the narrative here.)
Of the smart meters, one was way out of whack, and it’s possible that the lab tech broke the other. (There’s an art to reading test reports. In my early days as an engineer, I used to write them.) The remaining 611 smart meters met the standard. Of the reluctance meters, 141 passed, and six had readings that were slightly out-of-spec.
Just to make sure that the customers who had complained about high bills were fairly represented, 36 of the field tests were performed at residences where customers had made those complaints. These tests were simply smart-meter performance checks using a field test set, but at another 19 high-bill complaint customers, Structure conducted “shadow-meter” tests in which an Elster Digital Shadow Meter was installed side by side with the smart meter, and the test was run at full load, at light load, and with a 0.5 power factor. The CPUC’s pass/fail criterion was the 2.0% accuracy standard, and the smart meters passed.
End-To-End Testing
When I was writing test reports, the bright spots were the parts where something had been bungled and the writer had to explain it without making it sound like an excursion to Dilbert’s world. In those cases, I found it useful to insert a good deal of formal language and an even larger helping of minute detail. Thus, as a former practitioner of the art, I enjoyed Structure’s section on end-to-end testing.
What was supposed to happen was that, as part of field testing, Structure would set up six dummy accounts for certain customers who had complained of high bills. Then, while Standard “shadowed” the customers’ smart meters with its Elster Shadow Meters, another RF access point would send its own data back to PG&E’s monitoring and billing systems. The dummy accounts included dummy “billing addresses” to make the process “end-to-end.”
What went wrong was that the test was set up so there was also a central address where the physical bills were supposed to be sent. Unfortunately, the PG&E billing system wasn’t up to the subterfuge and sent the bills to the nonexistent dummy addresses. Thus, while there was data in the system to analyze, the last step, the actual ink-and-paper bills, ended up in Postal Service limbo.
That was disappointing. But apart from that, as I might have written myself, the smart meter data correlated with the test data all the way back to the billing office. There was still no sign of a smoking gun.
Continue to next page
Complaints Analysis
It did not look as if the smart meters were over-reading, nor did it look like the old reluctance meters were under-reading And apart from the billing foul-up, it looked like the data links from the meters to the advanced metering infrastructure (AMI) head-end and to the meter data management system (MDMS) and the customer care and billing (CC&B) systems were not faulty. What’s could be the problem?
To answer that, Structure analyzed high-bill complaints, some 1066 records, “for usage sensitivity to weather, unusual spikes, meter problems, manual or system based issues, meter reading issues, rate impacts, and service issues and identified factors that contributed to high bill complaints during late 2009 and early 2010.” Here are some selected quotes:
- “Meter deployment schedules coincided with increased energy usage caused by a heat wave.”
- Specifically, “Customers experienced increased kilowatt hour (kWh) usage after installation of Smart Meters due to weather. The comparison utilized the 2006 and 2009 years with similar summer profiles and determined that in 86% of the 2009 complaints, the average daily usage was less than the 2006 summer although the 2006 summer months were hotter. Structure verified that the weather in the same July/August period for 2007 and 2008 was 2 to 3 degrees cooler than in 2009.”
- “Some Customers experienced load changes that were reflective of changes in personal circumstances. Examples included room additions, pool additions, and equipment malfunctions.”
- “Rate increases compounded the financial impact of the additional weather-related usage.”
- “PG&E utilized field meter readers for an average of 131 days after Smart Meters were installed, resulting in similar meter reading errors as electromechanical meters.”
Where the Structure document did take PG&E to task was for a number of customer relations failures—which, in fairness, it later notes are being or have been rectified.
The Baby Monitor Conundrum
What next? An RF interference issue may be brewing. As noted above, both the GE and Landis+Gyr smart meters use Silver Spring Networks radios, which PG&E says operate in the 900-MHz and 2.4-GHz bands. The PG&E Access Points operate both in the 900-MHz band and/or via cell service providers.
Here’s the interesting part. There is an ongoing thread in the consumer ombudsman column of The San Jose Mercury News about bursts of interference on baby monitors, portable phones, and wireless speakers that newspaper readers are associating with the advent of smart meters in their neighborhoods.
According to the Structure report, the utility says that (a) the average smart meter transmits only about 45 seconds each day, (b) maximum output power is limited to 1 W, and (c) at a distance of 25 feet, the smart-meter power density is 0.005 μw/cm2, while the FCC’s limit is 601 μw/cm2. (That’s what the utility told Structure, according to the report. I’m not seeing similar information in the newspaper ombudsman’s columns.)
It’s hard to say whether this baby monitor issue is a red herring or not. The reports in the newspaper do appear to describe bursts of data. On the other hand, the essence of the industrial/scientific/medical (ISM) bands is that there are no guarantees of interference-free reception, as long as the interfering device meets the radiated emissions standards for the appropriate band and operating mode.
Perhaps a bigger potential problem is this, from the Structure report: “PG&E has determined that certain models of Ground Fault Interrupter (GFI) breakers (such as those used on hot tubs) may be impacted if they are in close proximity to the meter. PG&E has also engaged smart meter manufacturers to develop low power transmitter solutions to the GFI interference issue, and has trained the installation contractors to listen for GFI tripping upon installation of a new meter.” In hip California, false-tripping GFIs for hot-tub heaters could cause almost as much consternation as quintupling electric bills.