Does AMI need a robust MDM system to reach its full potential?


Does AMI need a robust MDM system to reach its full potential?

Over the past several years, the number of utilities looking to upgrade their mass-market metering, meter reading, and meter data processing infrastructure has grown and is accelerating. I

n past years, these utilities would have been offered simple one-way drive-by, radio, or power line carrier automatic meter reading (AMR) technologies that provided either lowcost monthly cumulative meter readings or more expensive daily reads. Hourly readings could be obtained, but at a much higher operational cost, which restricted their use to load research and customer usage complaint resolution. The primary business case behind the AMR of old was reduction in meter reading labour costs, missed reads because of inaccessible meters, and billing errors due to incorrect meter readings.

Metering diagram.JPG

 Today, utilities have the choice to go with much richer twoway AMR technologies. These systems, which use a wide range of communication mediums, are marketed to provide outage detection, remote connect/disconnect capabilities, on-demand reads, tamper detection and – most importantly – revenue-grade hourly or shorter interval energy readings that can be used for market-based rates or demand response programmes. All this new functionality has given birth to a new term for AMR: automatic metering infrastructure (AMI).

As was the problem with the older AMR technologies, no one AMI system can meet all the meter reading needs of today’s utilities. For example, one technology may be the most cost-effective solution for high-density distribution areas, but is cost prohibitive for low-density areas. Moreover, the new technologies are designed for single-phase mass market residential and small commercial and industrial (C&I) customers, but have not been proven effective for multiphase medium and large C&I sites, which still rely on legacy modem-based telephone dial-out/dial-in systems.

Given the above AMI realities, there has been an emergence of meter data management (MDM) systems designed specifically to rationalise disparate AMI technologies onto one unifying platform. Basic MDM functionalities include connectivity tools for system integration; flexible validation, estimation and editing algorithms; ability to transform interval meter data into simple or complex billing determinants; flexible aggregation algorithms for load research, forecasting and distribution planning; and a web interface for endcustomer data access.

More advanced systems also provide support for enhanced theft detection, outage management, AMI asset management, workflow management, AMI installation support, and direct connectivity to AMI concentrators, gateways or metering end points.

The big question in the marketplace today is: “Can all this functionality exist in one MDM system? And if it can, what are the characteristics that separate a poor MDM system from one that can deliver its advertised value proposition?” Without a doubt, the new AMI technologies are able to produce the raw data elements that serve as primary inputs to the MDM.


One of the biggest challenges of any MDM system is to ensure that the interval data being generated by the various AMI systems is accurate. In the past, cumulative monthly kWh or maximum demand meter readings were validated, estimated and edited (VEE) using simple algorithms to compare historical values with more sophisticated systems incorporating weather normalisation. Now, to get the same billing results using interval data, more than 700 hourly values per customer will need to be passed through the MDM’s VEE processes.

Inherently the VEE process is nothing new to any utility business that has been collecting fifteen minute interval data for large C&I customer billing and load research, typically using interval metering technology that has seen little change over the last 20 years. The most common VEE issues today for these systems are gaps in data owing to communication or meter errors, and erroneous data owing to incorrect meter multipliers or clock errors.

Although utilities have used automatic validation for years to identify incorrect data, for the most part all meter data estimation and editing for interval data is done using some form of human intervention. Typically, most utilities today have specialised departments or groups responsible for processing interval data, ranging in size from a minimum of one person for small systems to ten to 15 personnel for systems processing 10,000 to 15,000 interval meters.

Do all the above meter data issues disappear with AMI, or will utilities need a large team of VEE experts to keep the data clean? One thing is clear – each AMI technology has its relevant strengths and weaknesses that will become more apparent as utilities move from using standard cumulative meter reads for customer billing to actually using the interval data in the billing process. Issues like timestamps, gaps, and meter data resolution will stand out clearly when the first customer calls in with a usage complaint.

Although an MDM system cannot correct any inherent weakness in any particular AMI technology, it can certainly address the VEE processing issues. Today, professional interval data collection agencies use automated knowledgebased VEE processes to minimise the amount of human intervention in the editing process. A large data collector in the UK recently specified and implemented an MDM system that was not only capable of automated validation, but also capable of rule-based automated estimation based on a strict hierarchy of estimation algorithms. The system only rejects data for manual editing if it cannot find an appropriate rule that addresses the data issue. Additionally, automatic editing only makes sense if the MDM system has robust versioning capabilities that track why and how the data was edited.

With this in mind, utilities planning to implement AMI should look for three fundamental characteristics while evaluating MDM systems – performance, flexibility, and traceability.


First, AMI vendors now promote the capability to generate revenue-grade interval data with their respective technologies. This means that simple cumulative monthly kWh values used for customer billing will require the storage and VEE processing of over 720 hourly values. More complex TOU, market-based, real time or critical peak rates will require additional processing requirements.

It is paramount that any MDM system has a demonstrated performance capability to handle these additional volumes. During their evaluation process, utilities should insist on seeing their specific processing requirements being met, either by asking the MDM vendor to provide a reference that is processing similar volumes of data, or to set up a test environment that will demonstrate that capability.

Second, utilities should look for an MDM system that allows a high degree of configuration flexibility around the utility’s unique business logic requirements and existing legacy systems. Systems that dictate fixed or inflexible business logic should be avoided.


Operational diagram.JPG

This will become particularly important when MDM is being implemented where investments have already been made for new outage management, workflow management, and asset management systems. In these instances, which are typically the norm at most utilities, the MDM will act more as a message broker between the AMI system(s) and the other existing systems, and will require a high degree of flexibility given the number and type of those systems.

One other key flexibility requirement will be the MDM system’s capability to be used by all the groups intending to access AMI data. Simply put, the functional views and screens used by the billing department will be fundamentally different to those of the distribution planning department, or even the group responsible for the VEE process. Trying to accommodate every group’s business requirements into a common view will not be possible. A better scenario is that each group is able to configure its particular business logic into the system without interfering with another group’s requirements or the underlying physical data.

Third, the MDM system must provide the required data versioning and user activity logging to meet regulatory mandates such as Sarbanes-Oxley in the US. Given the scale and number of users of these new MDM systems, it will be paramount that both management and auditing agencies have the necessary tools in the system to track and report on all the changes that happen during the normal course of business.


Although sometimes at odds with each other, these three characteristics are available with MDM systems built on performance-based relational databases using modern programming languages. Selecting the right MDM system will ensure that AMI potentials are fully realised.

In conclusion, the author recommends that utilities intending to implement mass market AMI technology in the near future should first look at selecting and implementing an MDM technology that can address current interval data processing, distribution, and integration issues as mentioned in this article for C&I customers. In a sense, any new AMI technology is just a modern extension of meter and communication technologies that have been used for C&I customers for years. Getting it right for this customer base, which would not require any additional field metering equipment, would give the utility the opportunity to fully integrate the MDM with the required legacy systems using a smaller customer base.

If the full AMI data value proposition cannot be achieved with a utility’s largest customers, then there would be little hope – outside of reduced meter-reading costs – that a full-blown AMI system would add extra value. If the value is real, then adding additional AMI hardware will not be such a daunting task, as most of the MDM integration work would already be in place.