Securing long-term dispatch of isolated microgrids with high-penetration renewable generation: A controlled evolution-based framework

Isolated microgrids with high-penetration renewable generation are facing significant challenges due to the increasing uncertainties in renewable generation and load demand. Energy storage within microgrids plays a key role in overcoming the challenges, yet optimizing the dispatch for long-term energy supply remains a critical issue due to inevitable nonanticipativity constraints. This paper addresses the long-term dispatch problem in isolated microgrids with a high share of renewable generation. Firstly, a mathematical model of the isolated microgrid is developed, incorporating both long-duration energy storage (LDES) and virtual energy storage (VES). The uncertainties in renewable generation and load demand are characterized and dynamically updated using kernel regression techniques and sequentially observed data. The stochastic response capability of VES is then convexified through chance constraints. Secondly, we propose a long-term decision-making framework for isolated microgrids, where the controlled evolution layer is first introduced to coordinate the SOC variation of LDES and the real-time operation of the microgrid. Finally, a nonanticipativity cut generation algorithm is devised to efficiently solve the long-term dispatch problem while ensuring the nonanticipativity constraints. Case studies, based on actual data from southwestern Germany, demonstrate that the proposed method significantly reduces load loss compared to mainstream methods, as well as offering substantial advantages in terms of economic efficiency and renewable energy utilization.