By default, I tend to favour a weekly assessment interval for routine exception reporting and associated analysis. Monthly is too long for all except the smallest users (although it may be appropriate for passive top-management reports for users of all sizes) and months are also too inconsistent in terms both both of duration and number of working days.

In some applications, daily analysis may be viable, for example:

• in buildings such as data centres which operate seven days a week and respond rapidly to changing weather conditions; or
• in energy-intensive manufacturing processes.

More frequent near-real-time assessment is sometimes attempted but this brings complications that tend to outweigh the benefits. Firstly, there will be error induced by short-term effects such as transients, lags, latencies, and factors which are not practical to take into account but whose random influences would have cancelled out over a longer time interval. Secondly, the cash values of excess consumptions over a short interval are very small. Thus with too-frequent reporting the user is continually bombarded with trivial alerts which often prove fleeting. Not the best recipe for engagement.

Having said that, where fine-grained data are being collected they can be an invaluable diagnostic aid; but the best reporting tactic is to review performance at, say, a daily or weekly interval and use the real-time record for diagnosis by exception.

For the purposes of estimating the annual savings related to building insulation and ventilation-rate reduction, two tables of annual degree-day values have been published, one for heating and another for cooling. Both tables provide regional values to a range of useful base temperatures.

When people ask me for advice on how to verify energy savings, it is usually because their analysis is not giving the results they expected. Often they have left it too late, developing a methodology after the event or even making it up as they go along. So if you are contemplating an energy-saving project the first plea I would make is this: agree a measurement and verification plan between the interested parties before the project starts. That way, everyone is forced to think about the calculation methodology and (just as importantly) focus on what data will be needed, who will collect it, and even how much uncertainty there is likely to be in the conclusions. It also pays to think about what non-routine changes might occur (patterns of occupation, extensions, demolitions, etc) and agree how those will be factored in if they occur.

Sometimes, fortunately, it is possible to rescue the verification of a project where the “shoot first, ask questions later” approach has been used. To achieve a resolution one needs two things: first a willingness on both sides to accept a retrospective definition of procedure; and secondly, at least some accurate prior consumption data. That consumption data can, however, be sparse, so the presence of a lot of estimates (a common situation) need not necessarily be a problem. The analysis in such circumstances is done using a technique called “back-casting”.

Recall that in a normal evaluation, accurate and complete pre-project baseline data are needed so as to establish the prior relationship between consumption and relevant independent driving factors (such as degree days, hours of darkness, production and so on). A formula is derived, typically using regression analysis, for predicting consumption from those driving factors. After the energy conservation measure (ECM) has been installed, that same baseline formula can be fed with driving-factor data and will yield an estimate of what consumption would have been in the absence of the ECM. The spread between this estimate and actual consumption is a measure of the ‘avoided’ energy consumption.

The back-casting method is different. It turns that logic on its head. Using post-ECM rather than pre-ECM measurements, a formula is developed which relates consumption to driving factors for the improved installation (rather than its original performance). Thus you can say that the analysis “baseline” period follows, rather than precedes, the ECM, which some people find odd. In this scenario, pre-ECM actual consumptions can be compared with what they would have been if the ECM had been active all along, and one would expect those actual consumptions to be higher than the model’s predictions (the opposite of the conventional approach where post-ECM consumptions turn out lower than the baseline model predicts).

Back-casting is no less valid as a method, but it enjoys one big advantage in that you only need two firm meter readings predating the ECM. They should be as far apart in time as possible, and you need to be able to retrieve driving-factor data spanning exactly the entire period between the meter readings, but if those conditions are met, your model formula can tell you what the expected consumption of the installation would have been over that entire period if the ECM had already been in place, and hence how much more was actually used in the absence of the ECM. This back-to-front approach is attractive because regular meter readings are generally easier to assure after the project than before.

Link: Energy management training