Testing the accuracy of a measurement-based building energy model with synthetic data

Abstract A new class of building energy models has emerged which use short-term measured data to predict energy performance for a longer period or for average conditions (such as “weather normalization”). Few of these models have been fully validated because tests are complex, and it is difficult to control all variables. One method of testing a measurement-based model is to use “synthetic” consumption data generated by a conventional building loads simulation model. A procedure to test the accuracy of one aspect of a measurement-based model is described and applied to a recently developed model. The model was found to yield significant errors in predicted annual space heating use, but revised algorithms greatly improved the model's accuracy. The discovery, and subsequent correction of the errors, would have been much more difficult with field data.