Empirical testing has shown correlation between energy consumption and season in residential buildings, that can have an effect on the savings error. The CalTRACK 2.0 working group has proposed accounting for this seasonal effect by shortening the 12-month baseline period to 3-month weighted baseline periods. The figure below shows the effect of shortening baseline periods to 3-month on Normalized Mean Bias Error (NMBE) for residential buildings.
The working group has established the following NMBE thresholds to define buildings that require 3-month weighted baselines and those where an annual baseline period is acceptable.
- If there are 2 or more months that have an NMBE greater than 0.01, then 3-month weighted baselines are required.
- If there are less than 2 months that have an NMBE greater than 0.01, then 3-month weighted baselines are optional.
Since February, the CalTRACK 2.0 process has tackled several major issues. Below is a quick synopsis of the major tasks addressed and outcomes for these topics.
- Improved weather station mapping.
- Weighted regression for months in billing period methods.
- Expanded grid search range for variable balance points.
- Maximum lengths for baseline and reporting periods in billing period and daily methods.
- For portfolio-based cases, buildings with a high uncertainty metric can still be included in a portfolio as long as the defined portfolio-level uncertainty threshold is not exceeded.
- For site-based applications, ASHRAE Guideline 14 thresholds were recommended.
- Lawrence Berkeley National Lab’s Time-of-Week and Temperature model was used as a template for hourly methods.
- The aggregated hourly energy savings estimates were stable when considered at the portfolio-level for residential houses, which was an encouraging finding.
- When price signals are applied to hourly savings estimates, the temporal and locational value of energy efficiency projects can be calculated.
- Various price signals were analyzed and it was shown that they can provide different values to different types of load shapes based on priorities of the procurer.
The direction of CalTRACK 2.0 methods development was guided by feedback from use cases that required “payable savings”. For example, PG&E’s pay-for-performance energy efficiency program decided to increase compensation for energy savings during peak hours during the second iteration of their program. This required CalTRACK 2.0 to develop methods that generate savings estimates at the hourly level. Similarly, we expect CalTRACK 3.0’s tasks will be guided by the demands of stakeholders that implement programs using CalTRACK 2.0 methods.
In addition, we have designed a CalTRACK 3.0 sandbox on GitHub to document issues that require further investigation. We encourage working group members to continue adding ideas to the CalTRACK 3.0 sandbox as they arise.
- Use the CalTRACK 2.0 methods
- Contribute to the CalTRACK 3.0 sandbox
- Keep an eye out for the open source CalTRACK-based engine release (eemeter 2.0)