Baltazar Cervantes, Juan Carlos (2006-12). Development of an Automated Methodology for Calibration of Simplified Air-Side HVAC System Models and Estimation of Potential Savings from Retrofit/Commissioning Measures. Doctoral Dissertation. Thesis uri icon


  • This dissertation provides one methodology to determine potential energy savings
    of buildings with limited information. This methodology is based upon the simplified
    energy analysis procedure of HVAC systems and the control of the comfort conditions.
    Numerically, the algorithm is a tailored exhaustive search over all the independent
    variables that are commonly controlled for a specific type of HVAC system. The
    potential energy savings methodology has been applied in several buildings that have
    been retrofitted and/or commissioned previously. Results from the determined savings
    for the Zachry building at Texas A&M after being commissioned show a close
    agreement to the calculated potential energy savings (about 85%). Differences are
    mainly attributed to the use of simplified models.
    Due to the restriction of limited information about the building characteristics and
    operational control, the potential energy savings method requires the determination of
    parameters that characterize its thermal performance. Thus, a calibrated building is
    needed. A general procedure has been developed to carry out automated calibration of
    building energy use simulations. The methodology has been tested successfully on
    building simulations based on the simplified energy analysis procedure. The automated
    calibration is the minimization of the RMSE of the energy use over daily conditions.
    The minimization procedure is fulfilled with a non-canonical optimization algorithm, the Simulated Annealing, which mimics the Statistical Thermodynamic performance of
    the annealing process. That is to say, starting at a specified temperature the algorithm
    searches variable-space states that are steadier, while heuristically, by the Boltzmann
    distribution, the local minima is avoided. The process is repeated at a new lower
    temperature that is determined by a specific schedule until the global minimum is
    found. This methodology was applied to the most common air-handler units producing
    excellent results for ideal cases or for samples modified with a 1% white noise.

publication date

  • December 2006