After two years of collaborative development, OpenEEmeter working group wraps up with new hourly model showing 10x faster processing, improved accuracy, and transition to OpenDSM under Linux Foundation Energy.
Latest OpenEEmeter developments optimize handling of solar and non-solar data with elastic net approach, incorporating solar irradiance for better accuracy and computational efficiency.
OpenEEmeter working group discusses completed documentation, 24-hour hourly model approach, bias correction strategies, and new daily usage pattern clustering module.
OpenEEmeter working group proposes new umbrella naming convention, discusses autocorrelation interpolation, population-level model results, and strategies for fixing temperature bias.
OpenEEmeter working group recaps the elastic net model selection, discusses interpolation approaches, supplemental data inputs, and hyperparameter optimization for the hourly model.
OpenEEmeter working group recaps the completed 4.0 daily model release, discusses porting hourly model R&D work to the API, hyperparameter optimization, and plans for commercial sample analysis.
OpenEEmeter working group celebrates the public release of 4.0, compares LSTM neural network vs. elastic net models for the hourly model, and discusses incorporating supplemental data inputs.
LF Energy OpenEEmeter 4.0 is released, featuring an 84% reduction in seasonal bias, 95% reduction in weekend/weekday bias, and up to 100x faster performance with monthly data.
OpenEEmeter working group discusses integrating EEweather functions into OpenEEmeter, simplifications in 4.0 code, data sufficiency updates, and an elastic net model outperforming alternatives for the hourly model.
OpenEEmeter working group discusses switching from stratified K-fold to rolling test/train cross-validation, replacing CVRMSE with PNRMSE, and incorporating solar irradiance (GHI) into the hourly model.