It seems like the peak reading method will be more conservative than code, and I'm really hoping for something that doesn't add in extra demand beyond the code requirement. I do work on whole-house electrification, and the best case for speed, cost, etc, is if if the calcs can show that no service upgrade is required.

Meanwhile, I got my hands on a large chunk of PG&E data. Out of 2251 non-solar homes with more than 6 months of 15-minute interval data (00:00-00:15, 00:15-00:30...), this 15-minute average demand was larger than the hourly average demand by a factor of 1.30x, with a median of 1.26x and a standard deviation of 0.2x. This means that 95% of homes had 15-minute interval average demand that was less than 1.7 times the hourly average.

15 of the 2251 entries had 15-minute averages that were more than 2.0x the hourly average. The highest was 2.75x the hourly average, and *almost but not all* the ones with the highest factors were all very low-demand readings, so I suspect that we were looking at situations where adding a small load (AC or heat turning on) was able to bump the factor up above 2x quickly if the heat came on for only the last 15 minutes of the hour, for example.

This may not be a way to calc to code, but I'm going to find it super useful for understanding if service upgrades are actually necessary, since all PG&E clients have hourly data at the very least. Using 2.75x the hourly average seems like it would be the most conservative way to interpret this data, but 2x is probably good for most typical SFR houses.

**Anyone know anyone who's working on the next NEC code cycle? With the big push for electrification, this is going to be an important section of code to take a close look at.**