Struggling with weather normalization for a retail client in Minneapolis on Xcel's A-15 schedule. This winter was brutal - we had 7,200 heating degree days compared to the 10-year average of 6,800. Client's usage spiked 35% over benchmark but I'm not sure how much to attribute to weather versus operational inefficiencies. Anyone have good methodology for weather-normalizing commercial accounts in extreme weather years? The standard regression models seem inadequate for this level of variance.
Xcel Energy commercial benchmarks - heating degree days throwing off analysis
Christine, I feel your pain here in San Jose. We don't get your winter extremes but had similar issues with cooling degree days last summer. For weather normalization, I use a three-year rolling average rather than ten-year to account for climate trends. Also break out baseload versus weather-sensitive load using change-point analysis. What's their actual heating system - electric resistance, heat pumps, or mixed?
Up here in Huntsville, we deal with Tennessee Valley Authority and they actually provide weather-adjusted benchmarks in their commercial reports. Christine, does Xcel offer anything similar? Also consider the thermal lag effect - extreme cold means buildings take longer to recover temperature, so you get compounding usage even after the weather moderates.
Paul, they're running electric heat pumps with resistance backup, so definitely weather-sensitive. Albert, Xcel doesn't provide weather-adjusted benchmarks unfortunately - that would make this so much easier! The thermal lag point is interesting. Looking at their hourly data, I can see the recovery periods lasting 6-8 hours after cold snaps versus 2-3 hours in normal conditions.
Christine, try the PRISM method (Princeton Scorekeeping Method) for weather normalization. It's specifically designed for this type of analysis. You can download the software free from Princeton. I've used it successfully with Ameren Missouri accounts when we had that polar vortex winter. It handles extreme weather much better than simple linear regression.
Elmer's right about PRISM - great tool. Also check if the building envelope degraded over winter. Extreme cold can cause seal failures, increase infiltration. Had a Memphis client where weather stripping failed during an ice storm, doubled their heating load for weeks until they figured it out. Sometimes the "inefficiency" is actually physical damage that needs repair.
These are all excellent suggestions. Elmer, downloading PRISM now - that should help with the statistical analysis. Amir, good point about envelope integrity. Client mentioned some ice damming issues, so infiltration could definitely be a factor. Going to recommend a blower door test to quantify any air leakage.
Christine, one more thought - check their thermostat programming. Extreme weather sometimes causes facilities managers to override setbacks out of fear the system won't recover. I've seen 24/7 heating during cold snaps that should have been setback periods. That alone can add 20-30% to usage during extreme weather events.
Jim raises an excellent point about setback overrides. Human behavior during extreme weather is often the wild card that breaks benchmarking models. Facility managers get scared and throw efficiency out the window to ensure comfort. Document everything for your weather normalization - it's as much about operations as meteorology.