engleski

Modelling influential factors of consumption in district heating systems

Assessing the influential factors on measured (or allocated) heat consumption in district heating systems is often limited by the available data. Within a project of modelling consumption in district heating systems using machine learning, an access to a complete billing database of the largest Croatian district heating company was granted. These data comprise of the relevant billing data per each consumer from the period from 2010 till 2017. The company supplies approximately 126, 400 final consumers (both households and business) over 375 km of distribution network. The billing database has 40 vectors in a few million single inputs. Additionally, to these data, a questionnaire is distributed to the final consumers in several buildings labelled as “model buildings”, gathering behavioural and demographic data of final consumers (such as occupancy, mode of space usage, heat comfort level, age of occupants, etc.). The two sets of data are then merged, and a correlation analysis is performed. Further, two step hierarchical regression analysis is performed based on variables from billing database in the first step, with added behavioural and demographic variables obtained from the questionnaires in the second step. The models from two steps are compared, tested and interpreted. . Results of the most influential factors on heat consumption in district heating systems are given and the influence of the behavioural/demographic variables on the prediction accuracy of heating consumption is quantified and interpreted.

district heating, heat cost allocator, energy efficiency, influential factors, consumption, correlation analysis, two step hierarchical regression analysis

nije evidentirano