10 Conclusion

During the course of this project, several contributions were made. Firstly, the QBuildings software was redesigned into a new fully functional version, giving the possibility to generate consistent energy data on buildings in the region of interest. For this purpose, a structure containing three schemas - Aggregated, Processed, Smoothed - has been created. These patterns are available on a server GIS database, allowing quick and easy access to this data. QBuildings was used to provide data for the canton of Geneva. The generated data were compared to statistical standards on building heating and to SIG data. The results were very satisfactory for estimating heating demand and solar potential per m2. Outside of some outliers, the ERA considered in both data sets were also very close. However, the hot water demand did not match between the two data sets and the QBuildings calculation method requires validation with field measurements.

A district typification method has also been proposed. The method allows to group together similar districts in terms of energy, i.e. by looking at the district’s demand for heat, hot water and electricity and comparing them to the resources available to the district to meet them. Two clustering algorithms can be used with a predefined number of clusters, in case the knowledge available allows to determine the number of clusters to be expected, or without in the opposite case. The method was used in the canton of Geneva and gave different results depending on the algorithm used:

• Kmedoids obtains a very stable clustering, considering two different archetypes These groups can be summarised as:
  1. Residential-dominated districts,
  2. Non-residential-dominated districts. Although it is a significant distinction, one can wonder if it is a the only one that should be made.
• GaussianMixture obtains also good stability results, although not to the same extent as the stability of Kmedoids clusters, considering twenty-two groups. The main factors that differentiated the archetypes are the type of buildings mix, the biomass potential and the geothermal capacity. This clustering offers more possibility to differentiate the districts. This number of clusters compares to the numbers of land use categories in Geneva. Before testing the clusters with REHO, this clustering appears as the most preferable.

A typification validation with the REHO optimisation model still has to be done. This prevents conclusions from being drawn on the exact method to be used to group districts for energy optimisation purpose. Nevertheless, the current methodology is fully functional and leaves out for now different options that are interesting to investigate in the future.