The following bachelor thesis deals with the possibilities of calibrating parameters regarding the energy consumption of a specific building model. For this purpose, different metaheuristics are analysed with regard to their suitability and a software module is developed on the basis of these, which is to carry out an optimization of the parameters entered at the beginning on numerical way.

The energy consumption of a building is subject to various factors. Aspects such as building typology, type of use, number of users and many other factors have to be considered. This paper focuses on the parameters that need to be calibrated when modelling a single-family home in order to produce a specific course of energy consumption. Since errors can occur at any time during data collection, well-founded assumptions are often made that reduce the potential for errors but do not guarantee the desired validity. If these assumptions are used to simulate energy consumption, it is not possible to say to what extent the simulation data can be used. The software module developed as part of this work uses certain smart meter data in the energy simulation carried out with the EnergyPlus software in order to carry out a step-by-step optimisation during an initial calibration, which adjusts the model and its parameters until the course of the smart meter data has been reached as far as possible. This fact thus describes a minimization process which, analogous to the method of the smallest squares, forms and minimizes the sum of the square deviations. Due to the fact that the function values are determined by input into a model, metaheuristics based on swarm intelligence are analysed in particular. At first the particle swarm optimization and the ant colony optimization are described in more detail.

Particle swarm optimization was chosen for the calibration module because of its advantages in implementation. In this context, the mode of operation of the particle swarm optimization is analysed by testing and documenting different settings with regard to different strategies up to the configuration of the metaheuristics. The particle swarm optimization offers many implementation possibilities and shows a high flexibility, with which several tests can give information about the exact mode of operation and the behaviour on different influences. It should be made clear which setting is most suitable for the specific situation. The aspects that were considered in the mode of operation, such as speed, time of convergence and diversity of the results, therefore played an important role in the analysis.

The analysis could finally show that particle swarm optimization is a suitable metaheuristic and thus a good possibility for optimizing the parameters to be entered. The results are particularly characterized by their informative value and diversity.