Data-driven recommendations for building energy retrofitting at urban scale

Shevchenko, Oleksandr

Домівка
→
Students Research & Project Works | Роботи студентів
→
Факультет прикладних наук
→
Освітня програма наук про дані
→
2024
→
Перегляд матеріалів

dc.contributor.author	Shevchenko, Oleksandr
dc.date.accessioned	2024-08-23T11:26:49Z
dc.date.available	2024-08-23T11:26:49Z
dc.date.issued	2024
dc.identifier.citation	Shevchenko Oleksandr. Data-driven recommendations for building energy retrofitting at urban scale. Ukrainian Catholic University, Faculty of Applied Sciences, Department of Computer Sciences. Lviv 2024, 65 p.	uk
dc.identifier.uri	https://er.ucu.edu.ua/handle/1/4677
dc.language.iso	en	uk
dc.subject	Data-driven recommendations	uk
dc.subject	building energy retrofitting	uk
dc.subject	urban scale	uk
dc.title	Data-driven recommendations for building energy retrofitting at urban scale	uk
dc.type	Preprint	uk
dc.status	Публікується вперше	uk
dc.description.abstracten	Decreasing the number of retrofitting recommendations based on building stock data without using expensive and computational-consuming UBEM models or en- ergy advisors’ work could allow upscale of retrofitting decision-making for cities or districts and also for smaller units, such as associations of property owners or development companies. Despite the extensive amount of research in building rennovaiton area, the ques- tion of decreasing the number of retrofitting measures, that should be validated for every single building is commonly out of the research interests area. This work aims to investigate an approach to identify feasible retrofitting recom- mendations for existing building stock using data-driven approaches and machine- learning techniques based on urban-level datasets. We approached the problem as a classification task using the Swedish EPCs dataset as data input and created a dataset for using retrofitting measures recom- mendation from EPCs declarations as classification labels. In this study, we tested multi- and single-label classification approaches and var- ious classification algorithms. Results confirmed that this research area is promising, but obtained classification performance is insufficient for the industry usage. The bi- nary classification on single retrofitting measures achieving high precision, but low recall. This makes this method possible to be used in the task of enhancing building stock datasets with missing retrofitting measures.	uk