A novel inverse data driven modelling approach to performance-based building design during early stages

•Energy-based design consists of iterative process of divergent & convergent phases.•Designers make decision under high level of uncertainties at earlier stages.•Current building industry lacks an effective approach to energy-based design.•An inverse data-driven method is proposed combining divergent & convergent phases.•Input to inverse method is energy target & output is probabilities of design parameters. Energy analysis at the early stage of building design is a critical, yet difficult task in performance-based design. The difficulty arises from the complex, iterative, and uncertain nature of building design and the challenges of integration with well-posed energy assessment tools. The purpose of this article is to first review characteristics of performance-based design and establish requirements for a methodology that includes generating promising design alternatives, assessing the energy performance in tandem with the generation of alternatives, and choosing an alternative design solution with confidence. The study then proposes a novel systematic data-driven method, based on linear inverse modeling that generates plausible ranges for design parameters given a preferred energy target. The energy performance in this method is described as a linear function of the design parameters for a particular scenario of design. The application of the proposed method in a case study shows that it is capable of helping designers make informed decisions regarding energy performance iteratively and confidently at the early stages of building design.