Energy simulation models are often used to help improve the sustainability of homes, but there is a slight problem with that: in reality, the predictions they make are often wrong. PhD researcher Paula van den Brom analysed actual and theoretical energy consumption in more than a million homes and developed a calibration method to improve the models at housing stock level.

Two dwellings may both have been given energy label A, but in practice the amount of energy they consume may differ considerably. Why is that? The behaviour of residents obviously plays a role. If they want their home to have a constant temperature of 24°C and they take two 10-minute showers a day, the model’s prediction will be way off the mark. This is because the model is based on ‘standard’ behaviour. Building characteristics also play a role. “This can affect the payback period for measures taken to improve the sustainability of dwellings, but it can also result in flawed policies. That is because they are based on the models,” explains Paula van den Brom. “At an individual level, the difference between actual and theoretical consumption is easy to explain, but we are also seeing major differences at housing stock level, despite the fact that you would expect residents’ behaviour here to ‘average out’. So you really have to make the models much more accurate. To do that, you have to carry out a range of analyses – right down to individual dwelling level.”

And that is possible in the Netherlands. Statistics Netherlands (CBS) keeps accurate records of the energy consumption of Dutch households and links these to microdata such as family composition and the age of the residents. The AEDES SHAERE database is also used, which contains a wealth of information on building characteristics. For example, it is possible to use this database to look up the insulation level of many houses in the Netherlands.

As part of her PhD research, entitled ‘Energy in dwellings. A comparison between theory and practice’, Van den Brom conducted static analyses using information from almost 1.4 million dwellings. Surprisingly enough, these analyses revealed that building and environmental characteristics have almost as much of an impact on the difference between theoretical and actual energy consumption as occupant behaviour. 

Four times higher

In her analysis, Van den Brom pays particular attention to the outliers: the 10 percent with the highest and lowest levels of energy consumption. She also zooms in on the amount of energy used in dwellings before and after they were renovated. In particular, she focuses on 90.000 dwellings to which the same residents return following renovation. This minimises the impact of occupant behaviour.

Older dwellings often have the highest energy consumption levels within an energy label, even after renovation; using renovation measures to upgrade houses to new-build level is proving to be quite the challenge.

The general condition of the house before renovation also appears to influence the result. And single-person households use much less energy than families.

The differences with the models can be really quite striking. Van den Brom: “The level of consumption is sometimes three to four times higher or lower.” Relatively speaking, the user has a greater impact in energy-efficient homes than in energy-inefficient homes. In homes with energy label A, balanced ventilation appears to be linked to relatively low energy consumption, while consumption increases with natural ventilation.

Algorithm

That is all very interesting, but how can you use all these different parameters to give energy advice to tens or hundreds of thousands of households? “We can do this by using an algorithm to establish connections between actual energy consumption data in an intelligent way,” explains Van den Brom. “That facilitates the calibration of existing theoretical models at housing stock level.”

When she put it to the test, the average difference between actual and theoretical energy consumption in the models almost completely disappeared. That is an excellent result. Follow-up research should focus on making sure that policymakers and decision-makers will soon be able to use the improved models in practice for energy renovations and subsidy policy.