Why NMEC is Simply Sublime: Normalized Metered Energy Consumption

What’s cool about NMEC and M&V 2.0

 

There’s a lot of buzz in our state right now around the concept of M&V 2.0 and Normalized Metered Energy Consumption (NMEC). The term was created as part of AB 802 in 2015 with the inclusion of the following proposed simplified means of counting energy efficiency savings:

(the commission shall…) authorize electrical corporations or gas corporations to provide financial incentives, rebates, technical assistance, and support to their customers to increase the energy efficiency of existing buildings based on all estimated energy savings and energy usage reductions, taking into consideration the overall reduction in normalized metered energy consumption as a measure of energy savings.

With that stroke of the pen, an old approach to energy savings and baseline treatment has been renewed. I would have dubbed it ORNMEC so that we don’t forget that it is Overall Reduction . We should be stubborn and ‘ornery’ about that.

Using metered data to estimate savings means those savings must be significant, otherwise we don’t “see” them at the meter. So NMEC projects by nature must achieve real savings. As an engineer, it’s very appealing to spend more of my time finding, implementing, and commissioning projects that save our customers money, reduce their reliance on fossil fuels, and create markets for innovative technologies. The requirement to quantify “all savings” is also very attractive for our clients, because it lines up with how they think about energy efficiency projects – “how much did I save this month on my energy bills?”

What is NMEC?

Savings based on NMEC is similar to the Option C Whole Facility Approach of the industry standard measurement and verification (M&V) guidebook, IPMVP (International Performance Measurement and Verification Protocol). This approach uses empirical modeling, such as statistical regression, to develop relationships between energy use and independent variables. These relationships are needed to “normalize” energy use.

“Normalized” is a term with many possible interpretations. But, generally speaking, it’s been interpreted to mean a statistically valid method to adjust pre- and post- project energy use to a common set of conditions, so that savings can be properly estimated. Typically, energy use is normalized based on weather data, though you could also use other variables shown to influence energy consumption. For instance, in industrial plants, production levels (e.g. number of widgets or pounds of product) are likely to be correlated with energy use. In office buildings, operation schedules are a major influence. The key is that we can identify and develop a reliable relationship between energy use and its influencing factors.

I like that the legislation described “metered” energy use and did not use limiting language such as whole-building in its description. With the availability of cheaper sensors, improved connectivity, and a greater prevalence of system-level data, we can apply NMEC methods at a more granular scale; building subsystems. This approach has several advantages; savings will be a larger portion of the measured energy use, savings are measured where they are achieved, and owners obtain more direct and actionable information about the performance of their systems. The axiom “you can’t manage what you don’t measure” is foundational with NMEC methods.

What is M&V 2.0?

Whole building approaches are nothing new. Energy service companies (ESCOs) have been using regressions with monthly billing data for decades. But what’s changed is that now, more buildings have smart meters that can provide average consumption or demand data every 15 minutes. Compared to linear regressions using twelve points of billing data, we can now develop energy models with thousands more data points per year.

However, simple regression methods are inadequate for developing energy models with so much data. Advanced algorithms and machine learning approaches have been developed, with impressive names such as ”vector machines,” ”neural networks,” and ”random forests”. These algorithms can result in more accurate energy models, but the drawback is their complexity and unfamiliarity to the layperson. In most cases, the relationships cannot be expressed as a simple equation in an M&V Plan.

Advanced algorithms do lend themselves to automation, and many companies now provide automated M&V methods as a part of their software. M&V 2.0 is a term coined to refer to these next generation applications based on M&V 2.0 approaches have demonstrated the ability to improve model accuracy, add greater statistical validity, and, most importantly, to provide faster feedback. This real-time feedback helps catch operational issues as they occur.

M&V 2.0 approaches provide faster feedback to help catch operational issues as they occur.

M&V 1.0 (N = 12)

M&V 2.0 (N=35,040)

 

Sublime? That’s a Strong Sentiment…

This approach is very compelling for many reasons:

  • Simplicity
    Meter-based savings are easy to explain to customers – they get it. The approach relies on relatively few data streams. Since software applications can automate aspects of M&V, it frees up time to focus on other aspects of projects to make them successful.
  • Pay for Performance
    Meter-based methods demonstrate the real savings that result from efficiency and conservation projects. If program implementers only get paid when their customers save energy on the bill, they are motivated. This is, frankly, always what customers have expected anyway, rather than the historical practice of being paid on estimated savings on other buildings (deemed) or customized calculations.
  • Commissioning and Persistence
    Feedback from energy measurements facilitates better management of a building’s performance. Tracking savings on the bill motivates parties to ensure that efficiency measures work and their savings persist, and promotes better commissioning and energy management practices.
  • Results Driven
    The emphasis of the entire process is on results, not calculations. That helps us to spend more time achieving savings and less time estimating them using models and spreadsheets.

 

With simplicity comes some complications as well:

  • Not All Buildings Work
    The best application of whole building methods is for “predictable” buildings and systems. Predictable means an accurate energy model can be developed for a building’s energy use.

Data of a “Bad Building”

  • Non-Routine Events

Non-routine events (NREs) are the greatest risk in using whole building meter-based methods. Changes in building energy use can be caused by factors unrelated to efficiency measures and, if ignored, can result in biased savings estimates. Methods to identify NREs and quantify their impacts are only starting to emerge. One way to avoid NREs is to focus on metered building subsystems.

  • Black Box Models
    There are energy management software vendors who provide M&V services using proprietary modeling algorithms (so called “black boxes”). Because the algorithms are not transparent, some may doubt the validity of their savings estimates, especially when payments are based on them. Work at LBNL has produced test methods to validate vendors’ proprietary models (Granderson, et al) without divulging their algorithms, and implemented these tests on many such proprietary algorithms over hundreds of data sets. These test methods may be used to validate savings estimates in efficiency projects.
  • Measure Attribution
    With a whole-building approach, you lose savings attribution to individual measures. However, if savings on the bill is our goal (carbon savings really), we should lose our myopic focus on measures. We can achieve measure attribution in ways that make more sense than estimating individual measure savings. We can use functional testing and commissioning processes for each measure to assure they are properly installed and operating, and use this documentation for attribution. Owners can repeat the functional testing to assure the measures persist. Non-performing measures has been cited as a major reason some efficiency programs have not achieved their expected savings. The NMEC approach promotes best practices for sustainability as well.

 

Potholes in the road ahead?

Currently, all the IOU’s in California, and many third parties, are implementing or considering various combinations of NMEC approaches, through High Opportunity Programs and Projects (HOPPs), whole building pilots, and as potential components of new pay for performance programs. Given the limited experience with using these methods for program savings claims, we see some potential problems.

  1. Suddenly, Everyone’s On the Road
    Simultaneously, utilities are just getting HOPPs and NMEC programs off the ground and building experience with them. There is a high likelihood that there will be as many interpretations of the NMEC methodology as there are program implementers, which could lead to confusion and potential waste of funding. For site-specific NMEC applications, SoCal Edison is attempting to head off this divergence by publishing the NMEC Procedures Manual and PG&E is developing a Meter-Based Savings Platform Rulebook (last released on the CAEECC website).
  2. The New Kids Have to Prove Themselves
    The new kids on the block always seem to have to overcome an extra burden before joining the local crowd. The same is true for NMEC and M&V 2.0 to be accepted as viable methods in California’s regulated efficiency program portfolios, which consist of deemed and calculated methods. There is a lot of infrastructure built up around these traditional methods, including extensive rulebooks, documentation requirements, and evaluation methodologies. There are many drivers behind this complicated infrastructure, primarily to do with the rules governing measure eligibility, baseline definition, cost-effectiveness, and demand forecasting analysis. Efficiency programs are meant to achieve ‘additional’ savings, and that’s been interpreted as “measures must exceed code.” State code requirements change approximately every five years or so, causing confusion for customers (“I got a rebate last year?”), and requiring frequent updates to the rulebooks.
    NMEC methods enter this world by legislative act. We still need to determine how these programs and their savings factor into cost-effectiveness and load forecasting calculations, and how these programs can be evaluated without overly encumbering them with rules that demotivate participation.
  3. We Don’t Yet Know How to Integrate EM&V
    We don’t yet have a clear understanding of how NMEC fits in the overall portfolio of energy efficiency programs. We’ve not fully defined what data must be collected, and we don’t know how NMEC programs will be evaluated. There is much talk about ‘early EM&V’ which may provide timely results on successful programs and practices. Early EM&V can mean provision of data and information to evaluators at each step from customer engagement though final incentive payment. However, the necessary information must be identified soon so that we don’t launch programs without a clear vision for how these programs will be evaluated, and what data we need to collect.
  4. Measure Execution
    One of the great advantages of whole building and metered approaches is shifting the emphasis from attribution, to execution. We can get more savings, more cost effectively, if we’re spending our time on making sure buildings work, rather than spending time doing simulations and calculations that only fulfill regulatory requirements. In fact, another bill passed in 2015, SB 350, doubled California’s energy efficiency goals. However, we need efficiency program regulations to align better with market best practices to make this shift. If we must approach every project two ways; measure-by-measure calculations and whole building savings, we’re going to lose the cost-effective benefits of this approach.

Here at kW we’re excited by these possibilities and looking forward to working through the issues as they come up. Overall, we believe NMEC helps us achieve our core mission; to save our customers energy, money, and help reduce carbon emissions in the process.

Want a Deeper Dive into NMEC?

Southern California Edison will soon release an NMEC Savings Procedures Manual that describes how to conduct the savings estimation process on a site-by-site basis. It describes the typical issues encountered, and how to manage them. It will be available on the California Emerging Technology Coordinating Council’s website, www.etcc-ca.com.

 

1Comment
  • Anatoli Naoumov
    Posted at 20:19h, 16 March Reply

    Solid article, thank you.

    To comment on your comparison of M&V 1.0 and M&V 2.0: a baseline built on 12 points is likely to have a very high margin of error, just because 12 points is not enough to create an accurate AND statistically solid baseline. It is quite likely that error properly adjusted to degrees of freedom will overflow savings.
    On the other hand, while analysis of 8760 data points for a building may result in a more accurate baseline, this is not necessarily the case when this approach is used in manufacturing, If length of production cycle is comparable with 1h, then allocation of production to the proper hour becomes an issue. Also, in many manufacturing environments, production is recorded at times convenient to accounting/shipping/warehousing clerks. Therefore, I a compromise I found is to stop granulation of manufacturing data at daily (shift) level.

Post A Comment

How can we help you save money and lower your carbon footprint?