Methodology for estimating thermal performance of buildings for better planning and energy efficient investments

Abstract: This master thesis aims to find a method to separate the household-electricity usage fromheating-electricity and to estimate the thermal performance of buildings using the datasetavailable. The dataset includes electricity consumption of buildings, number of occupants, typeof heating system, climate data (temperature, wind speed etc.) which would be used toestimate the thermal performance of the single-family dwellings. The data set comprises ofhourly data of total electric consumption for those buildings that do not have solar photovoltaicsystem. For households that do have solar PV systems, it is gross PV production, sold and boughtelectricity (hourly based).The challenge in this study was developing a dynamic data driven model that could be applied toa large building dataset. Each building is located in a different climate zone, have a different size,number of inhabitants, and a different type of heating system. The buildings were thus analyzedaccording to the climate zone they were located in. The model gets building id from the userand locates other buildings that are situated in the same climate zone. It then develops energysignature for all these buildings and disintegrates the total consumption into base load, spaceheating load and intermittent load.The model evaluated 162 buildings, out of which 100 buildings had error in the data (all of theenergy consumption values were 0). The 62 buildings left were analyzed, and some buildingsshowed good results whereas some showed strange and absurd results. Overall, the model findsgood results for 45% of the buildings (28 out of 62) whereas the rest of the buildings (34 out of62) show bad results due to error in data such as missing hourly electricity consumption values,cooling load and the behavior of the power meter.The overall heat loss coefficient of the building is found by using regression analysis on thepower signature curve. The slope of the regression line gives the overall heat loss coefficient ofthe building which is a representation of the sum of transmission and ventilation losses and is animportant energy performance characteristic of a building. By comparing the value of heat losscoefficient obtained from the model to the theoretical value of heat loss coefficient calculated,we can see that when data quality is good the model calculates a good estimate of the heat losscoefficient.The results show that the data analysis and machine learning techniques can provide valuableinsights for the building stock. The thermal performance can be evaluated by developing modelssuch as energy signature to estimate overall heat loss factor which has big importance for theenergy performance. These insights can help to understand how the energy efficiency can beimproved in buildings. The overall heat loss factor can be used to find the U-value of the buildingwhich is also an important indicator of the thermal performance of the building.