Learning from success: A machine learning approach to guiding solar building envelope applications in non-domestic market

Abstract

Solar building envelopes, also known as Building Integrated PV (BIPV) show significant growth in Asia and Europe, although other regions such as Australia are still lagging. The decision to uptake BIPV is complex due to the heterogeneous interest of adopters and multi-dimensional features. Instead of redesigning BIPV in hypothetical buildings, we built a machine learning model using a database of real BIPV and building-attached PV (BAPV) applications, for the purpose of learning and predicting a BIPV adoption decision-making in non-domestic buildings in western countries. We used Australia as a case study to execute the support vector machine (SVM) prediction model. It was revealed that the combination of project determinants such as geographical conditions, equivalent building materials, interest rates and capital cost influenced the decision of BIPV. The prediction model provides pieces of information for stakeholders across the BIPV ecosystem to take their decision on investment, policymaking, and research directions. The current global industry transformation and innovations in technology are favourable to politically promoting and investing in BIPV. Such promotion and investment would help both expand the current market and reach the greenhouse targets.