Study On Building-Photovoltaic Interface Optimization Approach Oriented By Building Performance

The construction industry currently leads in civil energy consumption and therefore has a significant potential to achieve energy conservation goals and reduce GHG emissions.Recently,there is a rising interest in the integrated design of buildings and photovoltaics in the early design stage,aiming to determine convincing schemes under building performance priorities,sometimes called ‘computational design’.Unfortunately,the application of such conventional approaches in practices is limited due to their high computational costs.Although the concerned building attributes and building performance objectives are finite,it is almost impossible for even a skilled practitioner to optimize building design among alternatives in a many-dimensional search space driven by reasonable data rather than empirically,much less with many-criteria decision-making.With oriented by the indoor environment comfort,the economy,and the ecology of a building,this study attempted to run many-objective optimization and determine the optimal building design scheme from over a hundred Pareto-optimal alternatives fast and efficiently,which benefited from the implementation of a surrogate model,Artificial Neural Network model,rather than conventional Building Energy Models and complicated statistical equations.To improve the interpretability of multi-criteria decision-making,two global sensitivity analysis methods were also conducted to quantify the uncertainty of concerned building attributes for building performance during propagation in the black-box model.This study proposed a feasible and reliable framework and tool driven by a machine-learning model,focusing on improving building performance by optimizing attributes related to the building-photovoltaic interface.According to literature review,the research gaps and limitations of current building optimization studies can be summarized as follows:(1).Building interface optimization studies usually take reference building models as archetypes.However,the actual effect in practice remains to be seen.(2).Many studies for building optimization problems driven by surrogate models focus on improving model performance and taking Pareto solutions but lack exploration of the rationality and interpretability of decision-making.(3).Building optimization studies are supposed to include parametric simulation,gradient-free optimization,and decision-making at least,which requires association among multiple modules.Therefore,the efficiency and data resolution of such approaches are challenged due to the inevitable data exchange among multi-thread processes with heterogeneous protocols,unless an integrated framework is applied.Section 2 clarified the theories of building optimization design,defining key concepts,and analyzing the connotation of building optimization problems from objective-oriented and mechanism of interactions.Then two main approaches,multi-(or many-)objective optimization and decision-making algorithms were introduced,as well as sensitivity analysis.Section 2 gave solid theoretical support for the next contexts.Section 3 studied from epistemology and methodology of the model of the building-photovoltaic interface optimization design,analyzing concerned model types,determining inputs and outputs of the presented model,as well as how to build the model.Section 4 was the methodology of this study,presenting the framework of the building-photovoltaic interface optimization design and decision-making on the basis of the mentioned theories,concepts,and models in the above sections.In Section 5,the self-developed graphical user interface was used to validate the feasibility and reliability of the presented framework in a case study of an elderly apartment in Jian City,Shandong Province.The results showed:(1).Practitioners could take many-objective building performances across multidisciplinary into comprehensive consideration in early design phase through the integration of building-photovoltaic interface optimization design.(2).A high-performance artificial neural network model with significant prediction accuracy and robustness that reached a reasonable agreement between simulated and predicted values was developed and implemented.(3).The presented framework and tool could be applied for multi-criteria decision-making and sensitivity analysis in a limited time.In this study,the main achievements included:(1).A novel framework for building optimization design under building performance priorities was presented.(2).On the basis of utilization of Grasshopper and Python programming,a graphical user interface for building-photovoltaic interface optimization design was developed for interactive operation and result visualization.(3).The workability of the proposed approach and self-developed plugin was assessed in a case study of an elderly apartment in Jinan City.