Calibration of Building Simulation Models: Assessment of Current Acceptance Criteria

In this paper, the most used acceptance criteria in calibration of building simulation models are introduced and tried out by means of a practical exercise. In order to simplify the process of obtaining “simulated data” and to avoid carrying out a formal calibration procedure; two yearly testing profiles (hourly time scale) are created from real building electrical metered data (¼ hour profile). Both testing profiles represent two model responses that could possibly be obtained in a common calibration procedure.

The objective of this work is to test the capabilities of the method to determine (1) the model adequacy to represent an existing situation; (2) the reliability of the model when predicting a future or different scenario and also (3) the ability of the method to orient the practitioner to upgrade the model when it provides a non-satisfactory response.

To do this, the real accuracy of both testing profiles is verified by means of a complementary statistical bin analysis. This crosschecking analysis allows highlighting the strengths and weakness of the current criteria and determining whether they need to be revised, modified or complemented.

At the end of the analysis, it is concluded that the capabilities of the current acceptance criteria are limited because don’t provide any satisfactory answer, indication or clue for none of the three points aforementioned and some other complementary tests (such as bin analysis) must be implemented and performed in order to properly declare a model as calibrated.