Heat demand and generation can be decoupled by the use of thermal energy storages, resulting in improved energy efficiencies and enhanced possibilities for integration of renewable energy resources. The design and sizing of TES systems need to satisfy the requirements of the building regarding the thermal power as well as the capacity of the storage. The ratio of those two parameters can be expressed as the specific storage thermal power q•_s. If this ratio of a selected storage technology does not match the requirements of the given application, then the storage has to be oversized either in terms of the thermal power or storage capacity, probably resulting in higher investment costs. In this research, the required values of q•_s for different kinds of buildings (residential, governmental, business) were determined by a frequency analysis of heat load profiles. Therefore, these profiles were resolved into single cosine oscillations, allowing the identification of dominant frequencies with high heat load amplitudes. The main idea was to use different, well adapted TES for elimination of these dominant cosine oscillations, resulting in a smoothed residual load profile for the heat generator. An analytical solution for q•_s in the case of a sinusoidal charge and discharge cycle is presented, showing a decrease of q•_s in proportion to the reciprocal of the charging/discharging period T with typical values for q•_s between 400 W/kW h to below 1 W/kW h for short-term (daily) and long-term (seasonal) applications, respectively. Latent heat and thermochemical storages often use packed bed reactors for heat and mass transfer. Optimizations regarding thermal power usually have an inverse effect on the storage density. This leads to the conclusion that unique TES designs should be used for different storage periods. To match the requirements of the building, multiple TES should be linked together.