The 5 Golden Rules for energy data analysis
September 16, 2020 By Anatoli Naoumov
I’ve been in the energy management field for over a decade, and I am delighted to see end users increasingly relying on metered data to manage their energy use. After all, you cannot manage what you don’t measure, and metering energy consumption is a great start.
But it is just the start. Installing an army of meters is ultimately meaningless unless someone actually does something with all that data. Proper action is the key to efficiency and profit. Without proper action, all metered data is just costly digital dust. So how do we chart a profitable course of action from our metered data?
To help put everyone on the right path toward proactive energy management, I’ve distilled my experiences into these 5 Golden Rules for Energy Data Analysis.
Rule 1: In metering we trust
During my time in the energy management discipline, I’ve come across a variety of beliefs trying to pass themselves off as data. Here are just a few examples from various sites:
• I came across a compressor labelled “30 hp” (45 kW), and the new maintenance manager honestly believed the compressor to be a 30-hp piece of equipment. After a little digging, it was revealed the compressor’s motor has been replaced with a bigger one 10 years prior, and nobody cared to update the plate or any other record.
• Three (3) supposedly identical 5-hp fans controlled by the same VFD controller drew three (3) different currents, consistently, in unison. Turns out, two of those fans had damaged blades on their inner propellers. (Replacing those faulty blades effectively doubled condenser fan capacity and reduced the energy load on compressors.)
• A supposedly sequenced refrigeration plant compressor would randomly come online for no reason. (This was caused by an incorrect suction pressure setting.)
I’ve learned the hard way that claims such as “We always keep this operational schedule”, “Compressor/boiler/lighting is Off on weekends” or “Current never exceeds X amps” cannot be relied upon for energy use analysis until they are validated by direct metering.
Data can only serve as a reliable basis for investment decisions when it has been collected through routine automatic metering, with validated accuracy and consistency. Data collected manually is guaranteed to have typos and incorrect time stamps.
Beliefs are great for keeping conversation going. Business decisions require data.
Rule 2: Data has no meaning until it is analyzed
Energy metering is extremely valuable for performance monitoring and analysis. However, even meticulously recorded data, accurately measured by calibrated meters, has no meaning until it has been analyzed.
I came across a huge manufacturing company that meticulously collected electricity, water and natural gas consumption only to calculate “per ton” ratios every month. That is literally all they did with the information. They collected it. Nothing else.
So, when these per-ton ratios varied by as much as 100% from one month to the next, nothing happened. No one did anything, because no one knew why they were collecting this information in the first place, nor what the ratios were supposed to be.
Rule 3: Analysis is a comparison of actual to expected
The absolute value of metered numbers has no meaning, even when expressed in tables, graphs or ratios. Value comes from analysis; that is, by comparing what was expected against what is. Graphs, reports and dashboards (a.k.a. data visualization) do not constitute analysis. They just show the same raw data in another way.
For example, if I tell you that your flight has landed at 8:45 am, has it arrived on time? Well, that depends. What does your schedule say i.e. what was your expectation?
The same is true for energy data; whether derived from formal models or informal experience, we can only obtain business value from metered data when we compare it against our expectations.
Expectation infuses meaning into metered data. Consider this example:
The statement “Power draw was 565 kW at noon on Wednesday” has no practical meaning in and of itself. Is 565 kW good, bad or ugly? Does the fact that this value was measured at noon on a Wednesday have any deeper meaning? Your guess is as good as mine.
However, when we add “while peak demand at this plant is 1000 kW” to the first statement, we can conclude the plant’s machinery was not running at full capacity. If, instead of “Wednesday”, we say “July 1” (a statutory holiday), we can conclude that this is either a 24/7/365 operation or something is wrong. On the other hand, by adding “while total installed capacity of machines is 450 kW”, we can reasonably conclude that the meter is malfunctioning.
As you can see, metered data by itself has no meaning until it is compared with expectations; that is, analyzed. Depending on the importance of whatever is being measured to the business, the expectation for its value may come from a complex engineering calculation, a data-based regression model, or simply from a quick estimate.
Rule 4: Watch for sustained deviation from expectation
No model is perfect, no data is perfect; therefore, small, random variances between metered and expected values are normal. However, a small but sustained deviation constitutes a trend, which signals a change in operational conditions—energy savings or energy waste.
The best way to determine whether a small variation constitutes a trend is to use the CUSUM method to calculate the cumulative sum of differences between actual and expected values.
Again, do not sweat small deviations but watch for sustained trends.
Due to its cumulative nature, CUSUM averages out random fluctuations, but accumulates sustained ones. Small deviations do not matter, unless they are sustained over time.
Consider: small, random steering wheel turns do not radically alter your driving direction, unless for every two adjustments to the left you make only one adjustment to the right. This is the kind of trend that CUSUM picks up.
Rule 5: Check math against common sense
Mathematical models and calculations are tricky beasts; math does not care about common sense, nor about input errors or typos. I once saw a measurement-based estimate of achieved energy savings that exceeded the building’s overall consumption.
Use your common sense when checking mathematical/engineering calculations before arriving at any final conclusions.
Energy submetering is the foundation of energy performance analysis, but all the hard work can be lost due to trivial typos in data preparation or coding of formulas.
Avoid known pitfalls for profitable results
So, why bother with these rules? Well, if the profitability of your manufacturing process does not matter to you, then you need not bother. However, when you seek profitable results through energy management with 21st-Century accuracy, then these 5 Golden Rules are, well, golden.
Quite simply, good data allows better analysis, which leads to intelligent business decisions when it comes to planning energy retrofits, new equipment purchases, facility maintenance, and the like. In the worlds of energy management and business, ignorance is not bliss.
Anatoli Naoumov, MBA, CEM, CMVP, is managing partner at GreenQ Partners, which helps clients cut energy costs by identifying profitable energy projects, assessing their impact on the business, preparing business cases for investments, overseeing implementation and reporting the results. Contact Anatoli at firstname.lastname@example.org, and visit greenq.ca.
Print this page