Given the expected high penetration of renewable energy production in future electricity systems, it is common to consider buildings as a valuable source for the provisioning of flexibility to support the power grids.
Motivated by this concept, a wide variety of control strategies for building energy management has been proposed throughout the last decades and especially for the previously mentioned components.
However, these algorithms are usually implemented and evaluated for very specific settings and considerations.
Thus, a neutral comparison, especially of performance measures, is nearly impossible.
Inspired by recent developments in reinforcement learning research, we suggest the use of common environments (i.e. benchmarks) for filling this gap and finally propose a general concept for standardized benchmarks for the evaluation of control strategies for building energy management.