OpenEEmeter working group discusses API improvements in 4.0, CalTRACK 2.0 model challenges with overfitting and solar PV, and a proposed shift from CVRMSE to PNRMSE and static to rolling cross-validation.
OpenEEmeter working group announces the public alpha release of 4.0 daily model, and discusses approaches to improving the hourly model for solar PV customers including elastic nets, neural nets, and rolling cross-validation.
OpenEEmeter working group announces the deprecation of the CalTRACK name in favor of OpenEEmeter, previews the upcoming 2.1 daily model release, and discusses limitations of the 2.0 hourly TOWT model and goals for 3.0.
OpenEEmeter working group wraps up the CalTRACK 2.1 daily model with up to 100x speed improvements, and kicks off a literature review for CalTRACK 3.0 hourly methods covering machine learning, probabilistic models, and solar integration.
OpenEEmeter working group reviews CalTRACK 2.1 testing results showing dramatic reductions in seasonal and weekend/weekday bias, discusses an adaptive loss function, and previews plans for CalTRACK 3.0.
OpenEEmeter working group discusses relaxing data requirements for delivered fuels like propane and heating oil, and reviews CalTRACK 2.1 results showing 74% reduction in seasonal bias and up to 100x speed improvement over 2.0.
OpenEEmeter working group discusses CalTRACK data sufficiency barriers for delivered fuels, and previews CalTRACK 2.1 results showing dramatic bias reductions and a 20–120x improvement in computational speed.
OpenEEmeter working group reviews goals for CalTRACK 2.1 daily model bias improvements, discusses international implementation challenges from Carbon Co-op, and debates tradeoffs between daily and hourly model investment.
OpenEEmeter working group opens with a tribute to Shuli Goodman of LF Energy, then discusses sources of bias in CalTRACK 2.0 and the challenge of balancing additional parameters against overfitting risk.
Announcement for the February 7, 2023 CalTRACK working group meeting, with early results on CalTRACK 2.1 daily methods showing progress on reducing seasonal and weekend/weekday bias.