Historical temperature data are available from various sources, including the National Weather Service and Weather Underground. The National Weather Service reports the high and low temperatures and number of degree days for each day. Weather Underground does not report degree days, but has data for more locations.

Analysis Tips and Techniques

(1) Analyzing gas and electricity use usually is fairly straightforward because most buildings have monthly metered data. Fuel oil often is delivered in drops or at intervals based on estimated usage. Those intervals could be several months during summer, but only a few weeks during winter. With fuel oil, it sometimes helps to normalize usage and degree-day data to 30-day periods. For example, suppose our office building used 260 gal. of oil over a 22-day period with 847 HDD. The data point for entry into a regression analysis would be:


30 ÷ 22 × 260 = 354.5 gal. of oil

30 ÷ 22 × 847 = 1,155 degree days


Normalizing to a 30-day period works because base use is an important factor in most buildings. Remember that the intercept of the regression line is in units of energy per month (or other selected analysis period).

(2) Because the fuel-use data is energy input, the load determined from a fuel-use analysis is expressed as energy input. Building heat-loss calculations express results as energy output. The difference is boiler or furnace efficiency. While calculated heating loads are divided by boiler efficiency to determine required boiler size, the value of the slope of the regression line is based on energy input and, thus, includes the effect of boiler-efficiency losses.

(3) A high correlation coefficient says only that a system operates consistently and in response to outdoor temperature. It says nothing about equipment or system efficiency. It also says nothing about whether a building is overheated or underheated. Despite such a limitation, fuel-use analysis can provide useful insight. For example, it can sort out the effects of price escalation and weather changes in comparisons of dollar costs from one year to those of another. Also, if the analyst has an idea of what the heating load should be, fuel-use analysis might provide a clue concerning where to look for excessive energy use.

(4) When buildings such as offices and multifamily residences have a substantial base use, that load most often is for domestic water heating. The base use from a regression analysis is the average over the course of a month, including over night, when there is little or no domestic-hot-water use. Unless a system has a lot of hot-water storage, a boiler needs to deliver much more heat than the average “base” load on a call for domestic water heating. Few phenomena in nature have a peak-to-average ratio (crest factor) greater than 3:1 (a sine wave, which is typical of many phenomena, has a peak-to-average ratio of √2:1, or 1.414:1). Therefore, when sizing a replacement boiler based on energy-use data, a factor of three times the base-use intercept often is a good allowance for domestic-water-heating capacity.

(5) Table 2 contains a column labeled “Hours use of demand.” This is the kilowatt-hours in a period divided by the peak demand (kilowatts) for the period. It represents the demand needed for the system to operate continuously for an entire billing period (usually a month) to produce billed kilowatt-hour consumption.

Hours use of demand is not used in degree-day analysis, but provides useful insight into electricity use. There are 730 hr in a month, so for bills rendered monthly, hours use of demand has to be between 1 and 730. Typical office buildings in the northeast United States range from 300 to 350 hr use of demand. Higher hours use of demand, especially during winter, suggests an electrically heated building. Otherwise, hours use of demand over 400 or 450 raises a question of whether unnecessary loads operate during unoccupied hours. An oversized cooling system with few steps of control might increase demand enough to push hours use down to 100 or less. That low an hours use of demand suggests looking into some type of control to reduce peak demand.

Motor starting inrush current often is not a reason for excessive demand. Motor starting power factor typically is in the range of 20 percent (compared with 85 percent to 90 percent for an operating motor). Most of the high current draw during motor start-up goes to magnetizing current to develop torque. Magnetizing current does not develop real power that the kilowatt-hour meter measures. More importantly, metered demand is the highest number of kilowatt-hours measured during a demand-period window, not a momentary reading. Sometimes, the demand window is a half-hour; sometimes, it is 15 min; sometimes, it is the highest three consecutive 5-min periods. Motor starting current draw lasts only a few seconds, so even if a motor draws a lot of power during start-up, that high power draw is divided over the demand-period window to determine peak demand. Averaged over a 15-min demand window, starting a motor does not “spike” demand. Running it for several minutes, however, contributes to demand.


The ability to analyze utility bills is another skill in an engineer’s arsenal of tools to understand HVAC energy use and find ways to use energy effectively in buildings. This article described a concept and how to apply it.

Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.


A longtime member of HPAC Engineering’s Editorial Advisory Board, Kenneth M. Elovitz is an engineer and the in-house counsel for Energy Economics Inc. His knowledge of and experience with HVAC, electrical, and life-safety systems allow him to understand system function and performance, including interactions among disciplines. He is an adjunct instructor in the Architectural Engineering program at Worcester Polytechnic Institute.


Did you find this article useful? Send comments and suggestions to Executive Editor Scott Arnold at scott.arnold@penton.com.