Verifying your energy savings
Vilnis Vesma explains how to verify that an energy-saving measure has worked, and to estimate early on what it will save you long term.
If you are trying to prove the savings you have achieved through an energy project, the question you have to answer is simply this: how much am I now using, compared with what I would have used in the absence of the project? Except, of course, that it is not quite that simple, especially if you need an answer within weeks and the figures are affected by factors like the weather or production throughput which themselves cause consumption to vary.
In energy-intensive manufacturing industry the default method is to track and report the consumption per unit of output. Unfortunately, in the vast majority of cases, there is an element of constant ‘base load’ consumption, which means that the ratio of energy to throughput changes with throughput, falling when throughput is high and rising when it is low. These specific energy ratios are therefore fairly useless because you cannot easily disentangle the production-output effect from the energy efficiencies.
When assessing the performance of heating systems in buildings, where the weather is the major factor driving variation, the traditional approach to evaluating savings has been to apply weather adjustment using ‘degree-day’ figures. These are regional indices of how cold the weather has been: the higher the number, the colder the month was. Weather-adjustment of heating fuel consumption has tended to be a bit more sophisticated than specific energy ratios in manufacturing, as the existence of base load consumption is recognised and only the weather-related portion of annual fuel consumption is adjusted. What this method gives you is an estimate of what the building’s consumption would have been at standard weather conditions. This is fine for comparing one year with another, and quite valid, but you have to wait a whole year before you can say what the savings have been.
There is a better way, which gives an answer without having to wait a year, and which addresses the needs of buildings and processes alike. The clue was in the question I first posed: we need a way of calculating what consumption would have been in the absence of the project.
We start by analysing past performance on a weekly or monthly basis and determining how consumption varied with the relevant driving factor. We might for example discover that a bakery oven consumed a constant 1,200 kWh a week of gas, plus 200 kWh per tonne of bread baked. We could establish this simply by plotting weekly consumption against output on an x-y scatter diagram and finding the best-fit straight line through the points. Once you have such a model of how consumption relates to production, any improvement in performance becomes clearly visible regardless of variations in throughput. Even just one week after your project is completed, you could have both a figure for actual consumption and an estimate of the kWh consumption you would previously have expected for the week (200 times the tonnage produced, plus 1,200 for the fixed weekly element). Going back to my opening paragraph, you will see that was all you needed.
As the weeks go by and you collect more data, you can re-evaluate the relationship between consumption and its driving factors to see how the model has changed, and from that you can project your whole-year savings. For example Yassen Roussev at Tyneside Cyrenians (the charity providing accommodation and diverse support for homeless people on Tyneside) initiated a boiler replacement programme in their largest residential hostel, with condensing units controlled by a building management system. Its old heating characteristic was 9,867 kWh per week plus 107 kWh per degree day. Now, thanks to the improvements, it is 2,480 kWh per week and 154 kWh per degree day. The old and new characteristics are compared in figure 1; not only is consumption generally lower under all weather conditions, the reduction is greater in mild weather thanks to improved part-load control. As the typical annual degree-day count in their region is 2,300 their annual gas saving can readily be predicted from the difference between the old and new characteristics: it is (9,867-2,480) x 52 weeks plus (107-154) x 2,300 degree days = 276,024 kWh
Yassen was fortunate that he had been diligently collecting weekly meter readings before and after. Not everyone does this, and a lack of reliable consumption history makes it very difficult to prove savings, which inhibits energy users from making the investment in the first place. One typical scenario is where a building’s heating boilers are retrofitted with, say, an anti-cycling control subject to a savings guarantee and the customer wants rapid proof that it has worked. In the absence of past consumption data, ‘day on, day off’ tests are sometimes used in this situation whereby the add-on devices are disabled on alternate days and a daily record is kept of the number of hours that each boiler fires, along with the local degree-day measurement (which is a proxy for the heat output from the boilers). This method will be fair if the burners operate in an on/off mode at fixed rates. The daily run hours can then be plotted against daily degree days on a scatter diagram, and separate best-fit lines drawn for the two ‘families’ of points. This is similar to figure 1, but with daily rather than weekly data, and alternating days rather than before-and after sets of data. Similar arithmetic to that used above for the weekly before-and-after characteristics can then be used to project from daily enabled/disabled characteristics to the annual reduction in running hours and hence gas consumption.
Having reliable consumption data, associated with relevant ‘driving factor’ data by means (usually) of a straight-line targeting model, is also the basis of effective energy monitoring and targeting. Just as the method enables you to verify savings, it also enables you to spot unexpected excess consumption caused by hidden random faults, and paradoxically the more consumption varies with variations in its driving factor, the easier it is to detect adverse changes.
About the author : Vilnis Vesma is a specialist on energy monitoring and targeting and author of the Carbon Trust’s guide to the subject, CTG008. Vilnis has validated Sabien's measurement and monitoring package for validating M2G's energy performance.