LF Energy’s OpenDSM project (formerly OpenEEmeter) has recently approved and released a new model for measuring the energy impacts of demand-side interventions using hourly utility meter data. Developed via a collaborative effort of the OpenDSM working group, the new model replaces the previous hourly model (released with OpenEEmeter 3.0 and reflecting the CalTRACK 2.0 methodology) with an entirely new, modern, flexible, and extensible modeling framework. These improvements enable reliable application across a wide range of modeling contexts while avoiding many of the pitfalls that are commonly encountered with general-purpose models. 

The new OpenDSM hourly model has a number of advantages over the previous version, including reduced overfitting, improved handling of outliers, increased flexibility, a more developer-friendly API, and computing times that are faster by a factor of 4-5. Most notably, the model opens the door to improved reliability for meters with solar PV generation. While the prior hourly model performed well for solar meters when cloudiness was consistent from year to year, annual variability in cloud cover created the potential for bias in measured savings. As the subset of meters with solar PV continued to grow, this bias represented a growing source of risk for demand-side portfolios. By introducing the option to include solar irradiance as a predictive variable, the new OpenDSM hourly model drastically reduces the risk of measurement bias for solar meters.

In this webinar, we will introduce the new OpenDSM hourly model, describe how it functions, and summarize key improvements. We will also detail the model’s performance on the 33,000 meter sample on which it was developed and tested.