The residential sector is responsible for over 55% of the natural gas consumption in the Netherlands. In the climate accord of Paris, the Netherlands came to an agreement with the rest of the world leaders to limit the overall temperature rise by reducing the consumption of and by switching away from carbon-based fuels. The gas mining induced earthquakes in the northern part of the Netherlands increases the pressure on Dutch society to reduce natural gas consumption. The residential sector can play a role in reducing the consump- tion of natural gas in accordance with the Paris accord and to mitigate gas mining induced earthquakes. The amount of natural gas consumed per household is dependent on the behavioural aspects of residents, which are the biggest cause of uncertainty in estimating natural gas consumption. Current natural gas consumption based calculations are based upon dwelling characteristics and are not adjusted to individual behavioural as- pects. The inside temperature in a dwelling is seen as the primal indicator of residential heat consumption. The behavioural aspects of residents in the form of thermostat interaction are analysed in this thesis. The potential saving in the residential sector is addressed by including thermostat practices of residents in the estimation of potential savings.
The goal of this thesis is to identify thermostat practice in dwellings by the use of disaggregated energy con- sumption data and estimate the impact of adjusting individual thermostat practices on the natural gas con- sumption in the residential sector. Disaggregated energy consumption data is seen as detailed individual household consumption data. To reach the goal the following research question is answered:
What insights in thermostat practices that influence natural gas consumption of individual house- holds can be identified by a combined analysis of electricity and thermostat use?
Practice theory is used to understand the underlying mechanisms at play in household interaction with their thermostat. Disaggregated consumption data is used to gain insights in thermostat practices of individual dwellings. The thermostat practices of households are used to group specific practices and indemnify poten- tial savings in the residential sector.
The smart meter/ thermostat Toon is used to gather individual thermostat interactions and gas and electricity consumption data. Grouping of individual households with the use of clustering on the basis of thermostat settings is used to determine similar thermostat practices. Households with similar thermostat practices are grouped together, with the use of unsupervised classification in the form of hierarchal clustering. Similar thermostat practices groups are used to shape potential thermostat adjustments and assess the impact of these adjustments.
Thermostat practices of households are evaluated with the use of occupancy detection to identify potentials savings. A connection between thermostat practices and household occupancy is made with the use of elec- tricity consumption data. Individual electricity consumption data of households is used to determine the occupancy in a dwelling, by detecting moments of relative high consumption. Residential occupancy is de- tected with model ensemble of a Hidden Markov Model and a Rolling Mean Model to determine an overall occupancy schedule. The combined analysis of occupancy and thermostat practices is used to determine potential savings and determine the extent of these savings.
The generated insight in saving possibilities by a combination of thermostat practices and residential occu- pancy is used to develop and estimate the impact of 4 different thermostat practice adjustments. The esti- mations are based upon 3 different calculation methods: relative shift, heat demand and temperature and gas regression model, to gain a comprehensive understanding of the impact of each of the saving options. The potential savings are expressed in condition based lowering of the thermostat settings inside individ- ual dwellings. The relative shift method calculates the potential saving by relating gas consumption to the difference between in and outside temperature. Heat demand calculations are based upon the number of non-active heating hours for each of the saving options. In the regression method, a linear regression model for every individual household is build, to estimate the gas consumption on the basis of the difference be- tween the in and outside temperature.
More than half of the households in the sample group heats their home during daytime while the occupancy of these dwellings is detected at around 50%. The other half of the population has a clear heating pattern of
i
ii
morning and evening heating. For each of the detected thermostat practices, adjusting thermostat settings result in a potential gas saving. There is a factor 2.5 difference in potential saving between the different ther- mostat practices. Residents are able to save from 2% to 5% depending on their thermostat practice, resulting in an overall average saving of 34 euro per year. The largest saving potential of lowering the thermostat to 15 degrees overnight. Thermostat practices of households have a bigger impact on the gas consumption than currently used household characteristics. Natural gas consumption of households with similar thermostat practices have shown disparate consumption due to dwelling specific characteristics. Household specifics as thermostat practices and thermal inertia have shown to impact natural gas consumption. The detected ther- mal inertia of dwellings with the use of disaggregated consumption data is not in line with estimated values based upon physical characteristics of dwellings. The potential savings based upon physical characteristics of dwellings over estimates the potential saving in the residential sector. The impact of thermostat practices and thermal inertia of dwellings determined with the use of disaggregated consumption data is substantial. Cur- rent energy consumption studies and energy labelling are based upon physical characteristics of dwellings. The mismatch between estimated energy consumption and measured consumption indicates that estimat- ing residential gas consumption for individual households on the basis household characteristics alone is redundant. By including actual residential consumption data in energy labelling and shifting policy towards adapting household behaviour future policy measures can be improved.
The overall potential savings in the residential sector by adjusting thermostat settings are relatively small but can help to reach the climate goals. The results are based upon occupancy detection in a dwelling by analysing dis-aggregated consumption data. By improving the occupancy detection, the confidence in the results and number of included households can be improved. Expanding the size and representativeness of the sample group improves the applicability and confidence in the results. The personal preferences of residents in both the adoption adjustments and the actual impact on their thermal comfort is unknown. Fu- ture research is needed in order to close the gap between actual savings and potential savings by thermostat adjustments and to determine the impact on thermal comfort.