An Autonomous Energy Management Concept for Sustainable Smart Cities

In the course of the extreme expansion of distributed renewable energy generation systems and the increasing bottlenecks of energy resources due to political tensions, it is necessary to adapt the energy demand in cities and districts to availability. In addition, it is required to enable balancing across different energy resources, e.g., compensating gas shortages by an increased use of electricity.In this paper, we introduce a concept that makes such a load management control technically possible by using the flexibility of consumers of a city or district. To control the loads of a building autonomously multi agent deep reinforcement learning is applied, which automatically adapts to the consumption system. Dynamic energy prices are used as communication signals, they reflect the availability and demand of energy resources. We show that this concept meets all criteria for a load management system suitable for practice. Such are scalability, adaptability, data security, convergence speed, changeability and interoperability.