Efficiency and sustainability play an increasingly important role in most industrialized countries. Within the context of building energy systems the integration of renewables represents a key challenge. As these forms of energy show very volatile characteristics, local energy systems need more flexibility regarding the time of energy purchase. There are numerous approaches in research for the optimal use of storage units, such as rule-based control, model predictive control or adaptive control. However, most of these methods depend on detailed models of the system dynamics. A major hurdle to the manual development of physical white-box models for building energy systems is the low investments possible in most countries due to low energy costs. Furthermore, the creation of such models is very time-consuming and error prone, even for domain experts. Another weakness is, that changes in the systems are not automatically adapted within the models. However, with the steadily increasing availability of computing power and collected data in recent years, the practical application of methods from the field of machine-learning has yielded increasingly good results. Machine-learning methods can help to obtain data-driven, self-calibrating models, which can be learned from operation data directly. In this paper, we apply methods for automated data-driven model generation. We demonstrate how these techniques can be used to model individual subsystems as well as a complete energy supply infrastructure. The considered system is integrated into a district cooling network and consists of two compression chillers and an ice storage unit. This work is part of an ongoing research project with the aim to optimize the operation of the entire system. Our results suggest that, if detailed monitoring data is available, data-driven modelling represents a viable alternative to the labor-intensive white-box modelling approach.