Building thermal simulation based parametric methods are computationally intensive for optimizing the building design. This work uses experimental design techniques, i.e. fractional factorial design and response surface methodology, for sensitivity analysis and surrogate modeling respectively. These techniques find the solution in a reasonable time. Their application for building design optimization has not been found in the literature before. Fractional factorial design has been used to identify the significant design variables. These variables are used to form a correlation for annual cooling load prediction, using response surface methodology. These methods are illustrated using two cases to minimize the life cycle cost of a single-storeyed, air-conditioned, solar powered, detached home, with 64 sq. m. floor area, for the warm and humid Mumbai climate. For this climate, window solar heat gain coefficient, window to wall ratio, overhang depth and roof reflective coatings turn out to be the most important among the design variables used for this case study. The created response surface models show an error of less than 5% for more than 99% of the test data, which is comparable to other such models. Strategies are suggested to bring the error for the entire search space to less than 10%. Life cycle cost minimization using the model for case 2 does 12 million iterations as opposed to 250 iterations using a parametric EnergyPlus simulation run at the same time. The solution is better and the design achieved is also different. The optimum design has a cooling load of 55 kWh m-2 yr-1, while it varies from 46 to 118 kWh m-2 yr-1. This work adds an intuitive method for building design and opens up possibilities for optimization.
分享
引用
