Thermal Models for Intelligent Heating of Buildings

The Danish government has set the ambitious goal that the share of the total Danish electricity consumption, covered by wind energy, should be increased to 50% by year 2020. This asks for radical changes in how we utilize and transmit electricity in the future power grid. To fully utilize the high share of renewable power generation, which is in general intermittent and non-controllable, the consumption side has to be much more flexible than today. To achieve such flexibility, methods for moving power consumption in time, within the hourly timescale, have to be developed. One approach currently being pursued is to use the heat capacity of the thermal mass in buildings to temporarily store excess power production by increasing the electrical heating. Likewise can the electrical heating be postponed in periods with lack of production. To exploit the potential in thermal storage and to ensure the comfort of residents, proper prediction models for indoor temperature have to be developed. This paper presents a model for prediction of indoor temperature and power consumption from electrical space heating in an office building, using stochastic differential equations. The heat dynamic model is build using a grey box approach, i.e. by formulating the model using physical knowledge about heat flow, while the parameters in the model are estimated using collected data and statistics. The physical parameters in the model, e.g. heat capacities and resistances to transfer heat, have been estimated for an actual office building using a maximum likelihood technique.