Home

Welcome to the mave wiki!

Mave is a tool for automated measurement and verification. At it's most simple, the aim is to read energy consumption data from before and after a retrofit (pre-retrofit and post-retrofit data) and to predict how much energy the retrofit saved. Mave does this by training a model(s) to the data and using that model to predict energy consumption during a different period. Mave uses three different approaches to do this.

Single Model Measurement and Verification

In this case mave trains one model to the pre-retrofit data and uses the model to predict what the energy consumption would have been during the post-retrofit period if no retrofit had occurred. It then compares the prediction to the actual measured post-retrofit data to estimate how much energy the retrofit actually saved during the post-retrofit period.

Dual Model Measurement and Verification

In this case mave trains two models: one to the pre-retrofit data and one to the post-retrofit data. Mave uses each model to predict energy consumption over the same period. It then compares the two predictions to estimate how much energy would have been saved if the retrofit was done at the beginning of the period (a theoretical scenario). The default period that mave uses is the entire period contained within the input file (i.e. both pre- and post-retrofit). This is the longest period voer which it is certain that we will have all the required data (e.g. outside air temperature, if used). However, mave can also use other periods that may be more useful, such as one year of Typical Meteorological Year Data. This allows mave to estimate the difference between the pre- and post-retrofit models, normalized to a typical year. However, this approach only works if the model was trained on datetime and weather input features alone. For example, if 'total building occupancy' or 'units produced' were an input feature in the input file, this information will not be present in the TMY file, and thus a predication will not be possible. The previous method is more useful in this case.