DRIFT

Real-time optimization of low-voltage distribution network operation with integrated flexibility using deep reinforcement learning

The DRIFT (Deep Reinforcement learning for optimisation of LV distribution network operation with Integrated Flexibility in real-Time) project seeks to develop an advanced, real-time distribution network control algorithm using deep reinforcement learning (DRL) to address the challenges posed by increasing renewable energy sources, electric vehicles, and heat pumps on modern power systems. By optimally exploiting the flexibility of active network users—such as heat pumps, solar power plants, and battery storage – the project aims to improve voltage profiles, prevent transformer overloading, and enhance network hosting capacity without costly infrastructure reinforcement. The algorithm will be designed, tested offline with historical and simulated data, and then implemented in an actual low-voltage distribution network. Collaboration among experts in network modeling, artificial intelligence, and network operation from the University of Ljubljana and DSO Elektro Gorenjska, along with support from key industry stakeholders like SODO and ELES, ensures the solution’s scalability, replicability, and acceptance within the industry.

Project funding: