SEM Energy Modeling Method Selection Guide
The purpose of this paper is to outline the selection and application of a variety of SEM M&V methods. First a method selection decision tree provides guidance to determine the appropriate M&V method under common circumstances. Detail on the application, merits, demerits, and special considerations for each of seven methods is then provided. The appendices include references, standard metrics for model fitness (statistical verification), and discussion of some common analytical choices.
Source: Northwest Strategic Energy Management Collaborative
Published: June 2019
The Following is a Brief Summary of the Seven Methods Included in this Paper:
• Method 1 - The Forecasting Method: uses baseline data to create a predictive energy model. This is the preferred approach when baseline independent variable (energy driver) data is available. Assuming data is available, the challenge with this method is that if conditions significantly change during the reporting period, the model may not be able to predict accurately, in which case a revision or alternate method is needed.
• Method 2 - The Backcasting Method: follows the same methodology as forecasting but uses the reporting period to create the model. Energy use is then predicted for the baseline period. This method is useful when finer interval data is available during the reporting period (i.e. electrical meters are upgraded from monthly manual readouts to automated daily reads).
• Method 3 - A Mean Model: is applicable when there is insufficient variation in the independent variables or energy usage over time to create a predictive energy model. Use per period is then used as a model. The baseline is simply the average energy consumption for the facility during a defined baseline interval. This method should only be used when other modeling approaches are not effective and independent variables do not fluctuate much.
• Method 4 - The Pre-Post Method: uses both the baseline and reporting periods to generate one model. One of the model variables is an indicator variable for the reporting period, the coefficient of which is used to calculate energy savings. This method is useful when the baseline data set is too small to adequately capture the impact of key independent variables.
• Method 5 - The Standard Conditions Method: models are specified for the baseline and each reporting period. A set of “standard” conditions are specified and used as model inputs. Thus the baseline model energy use is adjusted for the standard conditions as is the energy use for the reporting period. An example of when this method is useful is when measures are highly weather dependent.
• Method 6 - The Chaining Method: is useful for situations where key independent variables, such as production, expand beyond the range observed in the baseline and reporting periods. In this case a single forecasting or backcasting model cannot adequately quantify savings. Instead a model is specified using an intermediate period. The model is then applied twice, once to backcast to the baseline period and once to forecast the reporting period performance.
• Method 7 - A Bottom-up Analysis: is essentially the calculation of energy savings through engineering principles and calculations for each of the identified and implemented measures. This method is useful when whole facility energy modeling provides inadequate analysis and when sufficient data is available for the implemented measures. SEM stakeholders including program administrators, evaluators and implementers can benefit from a greater understanding of the methods available to calculate energy savings beyond the common forecasting method.
In addition to the summaries presented in this paper, key references are included in Appendix D of this document to provide readers further background. Download the full resource to continue reading.