**4. Discussion**

Air-to-air or ground-source heat pumps were utilized in all optimal retrofit solutions. Thus, some buildings will increase their electric power demands on the grid. Other studies have therefore assumed that new heat pumps cannot use the average low-emission electricity of the grid and would instead need to utilize the marginal production that is typically high emission coal generation [25]. However, the issue can be bypassed if the existing loads are lowered at the same time as new ones are added. This was the case in this study on the building stock level. New heat pumps increased electricity demand in the retrofitted wood heated or district heated buildings, but this was o ffset by the same solutions (heat pumps and envelope improvements) reducing electricity demand in the most power intensive buildings that were heated directly with electric radiators or electric boilers.

The optimization of retrofit solutions favored very large PV systems (4400 MWp and 5600 MWp) for the scenarios aiming for the largest emission reduction. However, the majority of the solar electricity generated by the oversized systems could not be used at the buildings and had to be exported to the grid. The maximum power levels of the exports were several times larger than the peak power demand in building without electric main heating systems. In those cases, the high-power requirement of the solar arrays could be a problem for the distribution grid, if it is not designed to handle such power. However, in electrically heated buildings the peak winter demand was on a similar level as the exports and the grid would presumably be able to handle the loads. However, a study on integration of variable renewables in the Finnish grid estimated that more than 1100 MWp of solar electricity would decrease wind energy integration potential and significantly increase costs [39]. This shows the need for an additional study that looks at the building stock in more detail, while also including the e ffects of the national power grid and international transfers through the Nordic electricity market.

With the large amounts of excess solar energy going to the grid, the electricity spot price would likely drop. Solar electricity is produced in all buildings at the same time, so with enough excess power the price could go to zero or even to negative values. This would influence the LCC of the building retrofits, by lowering the lifetime value of solar electricity generation. Thus, if large scale retrofitting was done, the cost-optimal PV array size would go down. To avoid this, more ways to use the electricity are needed. Communities could use seasonal thermal energy storage to shift the use of electricity in summer to meet heating needs in winter. For example, solar electricity combined with borehole thermal energy storage for Finnish conditions was examined in [40]. Typically, demand response and short-term thermal energy storage in water tanks is also useful for increasing the value of solar electricity [41,42], but in the retrofit cases of the current study, solar thermal collectors were also included and handled most of the heating demand in summer. District heating could be produced with heat pumps [43]. Totally new uses for electricity are likely to appear. For example, the number of new electric car registrations in Finland has almost tripled in a year, though the absolute numbers are still low [44]. Other uses for excess electricity are in the energy intensive Finnish industry [45] or synthetic fuel production (also known as power-to-X) [46].

When GSHP was utilized, the annual peak electricity demand was significantly lower for the retrofit B cases than the original or retrofit D cases. This was due to higher heat pump thermal power capacity relative to the heating demand. When the heat pumps were sized to 60% or so of peak demand, electric backup heaters saw more use. Sizing the heat pump to above 90% made the peak electricity demand drop, also making the daily variance in demand smaller. This helps in sizing the electricity distribution infrastructure and designing energy storage systems. It is easier to optimize an energy storage system for power or energy capacity compared to having to maximize both.

The energy demand data for all the buildings were obtained through simulation. While dynamic simulation with IDA-ICE has been shown to be accurate, the results are sensitive to the background assumptions. Di fferent age classes of single-family houses were modelled, but the shape and size of the basic building was the same for every case. The results could be di fferent for smaller houses. In addition, only the southern climate zone of Finland was used for weather input data, creating a southern bias in the data. Further north, heating demand would be higher while solar energy generation would su ffer. However, the majority of houses are located in the southern zone. Doing detailed calculations for two more climate zones would have tripled the number of cases and the need for time-consuming optimization. The results were obtained using the Test Reference Year 2012. Since building retrofitting is a long-term investment, climate change can influence the energy demand of the buildings during the lifetime of the buildings, as we move towards the year 2050. Cooling demand was ignored in this study, but it could be that as air-to-air heat pumps become more common, people will start using them for cooling as well, even though the heat pumps were purchased mainly for reducing heating expenses. This would increase the electric loads during summer, though this increase could mostly be mitigated by the increased amount of solar power.

The changes in the building stock were accounted for in a simplistic way, assuming all buildings are immediately retrofitted. In practice, many buildings in regions with declining populations and house values would likely not be retrofitted, due to the resident's unwillingness to do long-term investments. A separate study is needed to calculate more feasible retrofit pathways, taking into account that change happens gradually and that new buildings are added while some old buildings are completely torn down. No flexibility or demand response methods were utilized, which removes the balancing element that appears when a large amount of buildings with di fferent use profiles and energy storage systems are combined. In practice, on the building stock level, the peak power demand could thus be expected to be lower than in the cases presented in this study. The houses were assumed to be oriented south for solar energy purposes. In practice, some buildings are oriented badly, receive a lot of shading or are otherwise not suitable for solar energy installations. Thus, only a fraction of houses would be feasible for solar energy production.

There is uncertainty in the heating systems in use in the current building stock. Building owners do not always report changes to their heating system, such as when replacing an oil boiler with a heat pump. Some wood-heated buildings might actually use wood only as a backup energy source, while others use it as the main heat source. Thus, the real distribution of heating systems is not known. Possible changes to the electricity use of equipment and appliances in the buildings were not considered in the study. On the one hand, old appliances are gradually upgraded into more energy-e fficient devices, which should reduce electricity consumption, but on the other hand, people are adding new electricity consuming equipment, which increases power demand. These trends could have an influence on the heating demand of buildings in the future through the excess heat they release.
