Comparing energy savings from monthly billed data can be misleading without a weather adjustment. For example, comparing a hot summer to a cold summer will give the appearance of energy savings when it was only due to reduced cooling. But when can weather adjustments be made? Enter the world of witchcraft.
Normalizing the un-normal
The basic premise behind weather normalization is to separate the weather-dependent energy consumption from the weather-independent consumption, and then normalize everything to a consistent baseline.
Let’s consider a home’s natural gas usage. The weather-independent energy consumption – like your home’s hot water use for showers, dishes, and hand washing – doesn’t vary much when the weather changes (unless you’re a crazy person who takes cold showers in the summer).
Weather dependent sources – like your home’s heating system usage – vary based on 1) how cold it gets and 2) how long it stays that cold. The opposite is true for cooling systems.
Weather normalization then statistically splits the weather independent and weather dependent energy consumption and compares the two based on typical weather patterns. This removes the weather factor and allows an apples-to-apples comparison.
To normalize, or not to normalize?
Comparing annual before and after energy usage is only useful on buildings where expected energy savings are at least roughly 10% of the total energy usage and where heating and cooling are not provided by the same fuel source. Systems that have the same heating and cooling fuel, like heat pump systems, can be done, but that takes some wicked black magic.
Even meeting these requirements, weather normalization still requires careful assumptions. Generally, these assumptions are fairly safe for natural gas usage. However, electricity usage is different. There are two major problems:
- The statistical analysis requires identification of a “balance point” temperature, or the temperature when no heating or cooling is required. This is easy for small, simple buildings, but for large complex facilities, this is more challenging to identify.
- Consider a building that uses 80% of its natural gas for heating and 20% of its electrical consumption for cooling. The electrical consumption has a lot more noise (other sources of electric consumption) where the signal (weather-dependent cooling) may be lost.
To demonstrate this, consider a 250,000 square foot hospital with natural gas boilers and an electric chilled water system located in Minneapolis, MN. Estimate its energy consumption with an energy model, and then compare it (leaving everything else constant) to its energy consumption in Seattle, WA. These two climates are noticeably different, yet while the natural gas consumption changes by over 8.0%, the electrical consumption changes by less than 0.1%.
Because of these issues, teasing out the weather-dependent energy consumption can sometimes be a high art that requires some extra milk bloodwort, jawbone of an ibis, and the whisker of a virgin lioness. You can’t get those at your local supermarket…