CalTRACK 2.0 Standing Meeting - Shared screen with speaker view
FYI - These tie to ASHRAE Guideline 14
Could you explain how the Portfolio Uncertainty is calculated?
Can you translate the CVRMSE to +/- at 90% confidence. I'm having a hard time comparing.
Thanks for those questions - we will tackle them next
If anyone else has questions feel free to put them here
Note actual ASHRAE CV-RMSE threshold built up from monthly modeling (IMT, PRISM) and has some additional language...
the baseline model shall have a maximum CV(RMSE) of20% for energy use and 30% for demand quantities whenless than 12 months’ worth of postretrofit data are availablefor computing savings. These requirements are 25%and 35%, respectively, when 12 to 60 months of data willbe used in computing savings. When more than 60 monthsof data will be available, these requirements are 30% and40%, respectively;
^^Blog McGee is talking about
These are the three major forms of measurement for different use cases… there is no perfect form of measurement, we have to use the right approach for each use case.
MattG: Sorry if I wasn’t clear; I’m not arguing for specific changes to the methods (yet!). I’m promoting the idea of tests for accuracy/precision of results so we keep improving. I believe such tests are possible.
We are conducting and will continue to use an out of sample approach to testing accuracy of the models. Completely agree we need to ground truth our models as often as possible.
Can you reliably find non-performners before the customer participates - afterwards is a lost cost.