Determining smart meter lifetimes to support investment decisions for new metering infrastructures
The desire to create a greener, more sustainable future is driving the adoption of new forms of energy generation (e.g. renewables) and consumption (such as electrical vehicles). Balancing the new patterns of supply and demand these technologies requires increased ability to monitor and control the flow of electricity within power networks, which is leading to the greater exploitation of digital technologies within grids.
In line with this digitalization of electrical power networks, utilities are investing billions of dollars in rolling out advanced metering infrastructures to replace electromechanical meters, bringing a wide range of benefits to both stakeholders and consumers. In this paper, we explore some key questions around the lifetime of these meters that underpin the decision to invest in them.
Defining and proving a meter’s lifetime
Traditional electromechanical energy meters have been a familiar installation for both corporate and personal consumers. Increasingly, they are now being replaced with smart energy meters, automatic metering reading (AMR) and advanced metering infrastructures (AMI). While electromechanical meters typically lasted 30 years or more, utilities are basing their decisions and cost-benefit analyses for smart metering on the assumption that these new meters will have a lifetime of at least 15 years. To verify that assumption two questions need to be considered: what is meant by ‘lifetime requirement’, and how do we prove that a certain type of meter meets it?
Firstly, when defining a lifetime requirement, several factors must be taken into account. How many failures are allowed before saying the device doesn’t meet the requirement? What types of failures are considered when calculating the lifetime? What is the required confidence level of the lifetime estimation? Without these parameters, a lifetime requirement has no value.
Secondly, proving a device meets the lifetime requirement is achieved by accelerating its lifetime. This can be done in all sorts of ways, each based on its own mathematical-statistical approach. One method is through accelerated reliability testing, as currently described for electricity metering equipment by the IEC 62059-31-1 standard. The standard uses a stable, elevated humidity and temperature to accelerate the device’s lifetime.
Standards-based accelerated reliability testing protocol
DNV GL’s KEMA Laboratories has developed a test protocol for reliability testing based on the IEC 62059-31-1 standard. Testing is a key step towards proving ‘less than F% of meters have failed within Y years with a confidence level of CL%’ is achieved. The values F, Y, and CL are specified by the customer and define the duration of the lifetime estimation test.
For the test, stresses are applied based on IEC 62059-31-1. The meters are placed in climate chambers. Each meter is connected to nominal voltage, and 10% maximum (Imax) current (direct connected meter). The meters are compared to a reference meter at reference conditions. Five kinds of temperature (T) and humidity (RH) stress combinations are applied to the devices under test (DUTs), as follows:
- Tmax & RHmax
- Tmax & RHmed
- Tmax & RHmin
- Tmed & RHmax
- Tmin & RHmax
Each combination is applied to a batch of 30 meters, meaning a total of 151 individual meters are needed to perform a lifetime test (150 for testing in the climate chamber and 1 reference). The IEC standard does allow for a lower sample size. However, this gives less data and less reliable results. Therefore, DNV GL always uses 30 meters per combination which, in addition to increasing the quality of the results, also decreases the time required for testing.
Minimum testing times
To find the minimum testing time, we first calculate the acceleration factor (AF) of these combinations using Peck’s model:
- RHu is the relative humidity in typical use conditions;
- RHmax is the relative humidity at maximum stress level;
- Tu is the temperature in typical use conditions;
- Tmax is the temperature at maximum stress level;
- k is the Boltzmann constant, ;
- n and Ea are empirically determined constants and set as n = 3 and Ea = 0.9.
From this, a minimum testing time can be calculated:
- Y and F are parameters of the lifetime characteristics to be checked:
- Y is the required lifetime
- F is the allowed failure rate
- UCL1 is the unreliability estimate for a given confidence level and order number 1
- UCL depends on the sample size, and is pre-calculated in annex D of IEC 62059-31-1
- AFmax is the acceleration factor at maximum stress level
- The IEC 62059-31-1 recommends accepting fault modes ß between 0.5 and 5, with a contribution (C) less than 15%
To make this more concrete, let’s consider a meter that will be used at 35 °C and 30% relative humidity and is rated for conditions up to 85 °C and 95% relative humidity. The maximum acceleration factor in this case is then:
To calculate the minimum time this meter must be tested, we require the lifetime specifications from the customer. For this example, let’s assume the customer has specified that no more than 5% of meters are allowed to fail within 15 years with a confidence level of 50%. Hence the minimum test time for this particular combination is:
The full set of testing conditions and times for this example are:
|Temperature & Humidity||AF||Dmin (days)|
|85 °C & 95%||3619.74||14|
|85 °C & 85%||2592.77||20|
|85 °C & 75%||1781.11||29|
|75 °C & 95%||1565.27||33|
|70 °C & 95%||1010.61||51|
Based on the above assumptions the total testing time of the 5 combinations of ambient temperature and humidity is 147 days (21 weeks). Increasing the confidence levels or lowering the maximum temperature for testing will drastically increase the total testing time.
During testing, faults are assigned to one of the following categories:
- Display – display functionality is verified once a day
- Accuracy – meter accuracy is verified once a day
- Data exchange – verified once a day to check the meter is available and that it is possible to establish a connection and exchange data with all meters
Determining the lifetime
The combinations are tested for at least the calculated minimum test time. If a fault occurs during the test time the tests continue until five faults in each independent fault mode are detected, with a maximum duration of twice the minimum test time.
After testing, the different independent faults are categorized and prepared for a Weibull distribution. From this distribution, it is possible to prove statistically, to a confidence level of CL%, that the target of less than F% failures in Y years has been achieved.
Is 15 years a sensible target for smart meter lifetimes?
The above discussions and calculations are based on the 15-year lifetime requirement that is common across the industry. While this figure is somewhat shorter than lifetimes expected from traditional meters, it is worth asking whether such a long lifetime requirement is justified for smart meters. After all, such meters:
- Are subject to significant technological changes, making it difficult to maintain hardware and software for the first-generation meters, which do not have the advanced functions of newer models
- Have complex features, such as radio communications and digital displays, which are subject to higher malfunction and failure rates
- Are similar to other types of information technology, computer equipment and electronic devices in that they are backed by short warranty periods and require significant upgrades or more frequent replacements as the technology matures, and
- Will likely be obsolete by the time they are re-verified as required by many regulators every 5 to 10 years
The question of what is a suitable lifetime requirement for smart meters is thus one that is still open to discussion within the industry.
To justify large investments in new generation metering technology, utilities are using meter lifetimes of 15 years and more in their cost-benefit analysis. To determine and verify this lifetime, DNV GL’s KEMA Laboratories have developed a protocol for accelerated reliability testing according to IEC 62059-31-1 standard.
Investors should consider how to specify Life expectancy. Off course testing is an option, but I expect parameter from Energy companies will soon point out that testing is too time consuming and therefore too expensive. For instance: when we choose a longer life time, or higher temperature and humidity conditions during operation, the test duration will increase substantively.
Talk to us at European Utility Week!
Are you interested in energy meter lifetime and do you want to discuss this topic with our experts? Then visit us in our booth B.j1 at the European Utility Week 2018 in Vienna, Austria from 06 to 08 November.
DNV GL – KEMA Laboratories is one of the leading independent testing laboratories for certification, type testing and calibration of energy meters according international and national standards. Our services include testing of direct and CT-connected electromechanical (Ferraris) energy meters, static energy meters for 4-quadrant active (Wh) and reactive (varh) power, pre-payment static energy meters with load control switch and source control switch and time switches. The test programs included metrological accuracy testing, influence testing, electromagnetic compatibility testing, product safety testing, mechanical testing, control switch testing, time keeping tests, reliability and durability testing, etc. The DNV GL – KEMA Laboratories operates under the ISO 17025 accreditation for testing and calibration laboratories and is Notified Body for the Measuring Instrument Directive MID 2014/32/EU with notification 2290.