Schedule a meeting
The power of Automation and Accuracy
Comparison between traditional M&V and the automated solution of Eneos
1 or 2 variables
Time consuming and costly
Quality of model is acceptable
Low accuracy according EVO-benchmark
Frequently via Excel
Difficult to detect NRE
Manual calculation of NRA
AI based contextualized models
Hourly, daily or monthly data
Statistical validation of > 10 variables
Fixed price and automated
Optimization of model Quality
Best software according EVO-benchmark
Daily or monthly updates
Integrated in EMS or own platform
Automatic detection of potential NRE
Automatic calculation of NRA
Let’s make your optimization projects transparent together!
"Through the automation of M&V, you can follow up the energy performance of your assets much closer and cheaper. This enables a lot of possibilities."
Measurement & Verification (M&V)
Measurement and Verification (M&V) is the term given to the process for quantifying savings delivered by an Energy Conservation Measure (ECM). Also the energy performance of a building can be followed up by M&V.
Calculation of energy savings
It's impossible to measure energy savings, due to the fact that you cannot measure the absence of something. Where it may look easy to calculate energy savings (energy use before - energy use now), often it's not that simple. A lot of variables may have an impact on the energy use and thus the energy savings:
- Weather variables
- Occupation of the building
- Production of an industrial site
- Comfort levels
The impact of the significant variables have to be excluded (or normalized) from the energy data, so they don't influence the result.
In M&V there are always two periods, sometimes three:
- Baseline Period: This is the period you want to compare your energy use with. Often this period covers a whole year. It's important that the whole range of each of the impacting variables are included in this period.
- Sometimes there is also an energy conservation measure (ECM) integration period: An energy conservation measure is installed, where the energy savings are not calculated.
- Reporting Period: This is the period you want to calculate the energy savings for. The length of this period can be one time interval to a whole year.
Some buildings have a variable occupation rate. When this occupation rate has a statistical significant impact on the energy use, occupation is a key impact variable for the energy use of this building. The occupation should be taken into account when calculating the energy savings. In this way an increase of the occupation of the building does not necessarily lead to a decrease of the energy savings, and the building performance can be followed up more accurate.
IPMVP (International Performance Measurement and Verification Protocol) is the protocol, managed by EVO, that creates a quality framework around M&V which is worldwide acknowledged. The use of IPMVP in M&V build trust and certainty about the used methods and documentation of M&V. All the results of Eneos are IPMVP-coherent if possible.
Advantages of M&V
- Quantify energy - and water savings in a standardized and statistically approved way
- Be able to communicate about energy savings
- Be able to exclude impacting effects to follow up building performance and strategic targets
- Detect deflections and anomalies of the building performance, making it possible to intervene
In ISO 50.001 or any other energy management system, there is a 'Plan–Do–Check–Act' cycle. The continuous monitoring of the energy use and the performance is an important part of this cycle (check - part). Integrating M&V makes it possible to objectively evaluate the effectiveness of an 'Energy Conservation Measure'. In ISO 50.001 the terms Energy Performance Indicators (EnPI) and Energy Baselines (EnB) are used to address M&V.
Energy performance contracts (EPC's)
In any EPC, there is a defined way to guarantee the energy performance during the contract. Mostly this happens with M&V (IPMVP: option C - Whole building). The energy performance is linked to a bonus-malus system, where the ESCO gets a bonus when there is more energy saved than required.
Sometimes legislation requires the use of M&V, like PLAGE in Brussels.
Best practice for follow-up energy performance
It is hard to measure the energy performance of a building or site. The combination of M&V and benchmarking makes it possible to compare different locations or buildings with each other and follow up a group of buildings in a structured way. This makes it possible to set clear targets for each building.
Quality of regression model
During the baseline period, a baseline (regression) model is created. This model links the energy use to external (outside temperature, humidity,...) or internal (occupation rate building, production volume of an industrial site) variables with one sole purpose: to explain the variability of the energy/water use. With this baseline model it is possible to exclude these effects and calculate the real energy savings.
The better the regression model, the more accurate the results can be calculated. There is more certainty about the calculated energy savings or less risk linked to these models.
How has the quality of the model been measured?
The general way to verify the quality of a regression model in M&V is via two statistical parameters: CV(RMSE) and NMBE.
CV(RMSE) (Coefficient of Variation of the Root Mean Square Error) measures the variability of the errors between measured and simulated values. It gives an indication of the model’s ability to predict the overall load shape that is reflected in the data.
NMBE (Normalized Mean Bias Error) is a good indicator of the overall behavior of the simulated data with regards to the regression line of the sample. NMBE is subject to cancellation errors; consequently, the use of this index alone is not recommended.
In Formula 1 and Formula 2, following parameters were used:
- m gives the measured values
- s gives the simulated values
- n are the number of data points
- p is suggested to be one
Importance of these parameters
These two statistical parameters give an idea about the quality of the model. The lower these error-indicators are, the better the model is and the more accurate it will work.
There are some pitfalls:
- In the traditional linear regression methods, these parameters were calculated with values used to build the model. This works for monthly data and simple linear regression models, but for more complex models a different strategy is needed to verify the statistical relevance. The strategy often used is the calculation of these metrics with data points what where excluded when building/training the model.
- Interpretation of these parameters: some buildings (called 'bad buildings') cannot be modeled due to their randomness in energy use. These buildings cannot have a (really) good model. Luckily, this hardly happens.
Minimal statistical criteria for simple linear regression models
Worldwide there are three main organizations which stated statistical criteria for energy model: ASHRAE (Guideline 14), FEMP and EVO (IPMVP). In the following graph the statistical criteria are shown. Note the difference between monthly models and hourly models. There is a higher allowed variability using high-frequency data.
Advanced M&V or M&V2.0
M&V is not new. For decades the 'whole building' approach has been applied to normalize energy use for outside temperature via linear regressions. These models were created with monthly energy data, resulting in a 12-point model considering a whole year. The introduction of automated, smart or digital metering systems changed everything: suddenly there was 15 min data available.
The traditional linear regressions are inadequate to deal with this granularity of data, so more complex AI-models come into play. With these complex models, it's possible to apply M&V on an hourly or daily frequency. It also makes it possible to detect NRE's and even correct for them after validation. The downside of such a model is the complexity. A linear regression model is easy to explain and understand, where more complex models don't have that advantage.
Which benefits does AM&V have?
There are multiple benefits to Advanced M&V:
- Enable to follow up energy saving with a higher frequency
- Automating Advanced M&V makes it cheaper, which can democratize the technology
- Be able to set priorities for the energy manager of a group of buildings
- Detection when energy performance of energy conservation measure (ECM) decreases on an hourly or daily basis
- Gaining insights about energy saving: when do they occur and how much impact do they have
- Automate the detection of NRE's (non-routine events) and the execution of NRA's (non-routine adjustments)
How can it benefit to automate NRE's?
Advanced M&V makes it possible to detect NRE's fast and also provides a solution for calculating the NRA itself. Using monthly updated linear regressions, this would be hard and time intensive. There is a new EVO Application guide which describes the execution of a NRA (EVO 10400 1:2020: IPMVP Application guide on non-routine events and adjustments).
How can these features be enabled?
The documentation of the API of Eneos provides all details about the integration of Advanced M&V. Contact us to get more details about our solution.
How to asses the quality of Advanced M&V-software?
Due to its complexity and the linkage to the building, it is hard to interpret the traditional statistical parameters in order to evaluate AM&V-software. This is why EVO developed an independent platform to perform benchmarks of Advanced M&V-software. In this way it is possible to compare them. This platform is called EVO’s Advanced M&V Testing Portal. Every software developer can do the benchmark and publish the results online. This is an independent way to verify and compare different softwares.
The lower the statistical parameters in the platform, the better the software performs. Eneos has one of the best performing M&V-softwares based on this benchmark!
Energy Use - baseline period
The start of any AM&V-project lays in the collection of the energy data of a certain location or site. The data quality is very important. AM&V requires granular (15 min, hourly, or daily) data, to make it possible to extract more insights. In this example, year 2019 has been selected as the baseline period. All the data of the energy use and the variables are from 2019 on an hourly basis.
Not only the energy use should be taken into account, but also the impacting variables. Eneos collects automatically all weather variables and checks the statistical relevance of every of them. The following graphs gives the variables in an hourly frequency for the year 2019 (baseline period).
Building baseline model
With all the available data, a baseline model has to be built. This is the core of the AM&V-software. The software selects the necessary variables and trains (builds) the AI-model. After the model is built, it can be applied to follow up the energy performance.
Applying baseline model
After the statistically best model has been trained for the baseline period (2019), it has to be applied on the reporting period, which was 2020. All the data necessary to use the model (energy use, impacting variables, timestamps,..) have to be collected and processed. Then the avoided energy use (energy savings according baseline model) can be calculated per hour. This makes it possible to describe the energy savings:
- When are the energy-savings being realized (during day, night, weekend, winter,...)?
- What's the value of the total savings being realized? And how does it compare against the expected savings?
- When do we see changes in the accumulated savings (CUSUM) graph?
Trends in energy savings patterns
In the last graph the average energy savings per hour of the week for 2020 are expressed in a heat map. Here it's possible to see when there are savings and when there is a higher energy use.
What is Measurement and Verification (M&V)?
Idea behind M&V
It's impossible to measure energy savings, due to the fact that you cannot measure the absence of something. Where it may look easy to calculate energy savings (energy use before - energy use now), often it's not that simple. A lot of variables have an impact on the energy use and thus the energy savings:
- Weather variables
- Occupation of the building
- Production of an industrial site
- Comfort levels
The impact of the significant variables have to be excluded (or normalized) from the data, so they don't influence the result.
Some buildings have a variable occupation rate. When this occupation rate has a statistical significant impact on the energy use, occupation is a key impact variable for this building. The occupation should be taken into account when calculation the energy savings. In this way an increase of the occupation of the building does not necessarily lead to a decrease of the energy savings.
IPMVP (International Performance Measurement and Verification Protocol) is the protocol, managed by EVO, that creates a quality framework around M&V which is worldwide acknowledged. The use of IPMVP in M&V build trust and certainty about the used methodes and documentation of M&V. All the results of Eneos are IPMVP-proof via statistical validation.
Why is M&V important?
Make it possible to quantify energy- or water savings
Energy savings are impossible to measure. M&V provides a framework to calculate them in a statistically approved way
Taking variables into account
Correct the energy use for all the influencing variables like outside temperature and usage of the building
If the activity rate of a building is increasing and it has an impact on the energy use, it can be compensated.
Follow up strategic energy targets
Verify if the energy saving targets are being reached
Detect when energy performance decreases
Trough M&V it's possible the follow up the energy performance of a building. This is the reason it can be used to detect changes in the buildings behaviour.
Where is M&V being used?
How is the quality of the model assesed?
Why is the quality of the model important?
How to compare M&V-software?
What does Advanced Measurement and Verification (AM&V) means?