Toward NZE: Optimising renewable energy distribution in Indonesia’s hybrid grids using Agentic AI

Supervisors

Dr Muhamad Risqi Utama Saputra, Monash University Indonesia

A/Prof Markus Wagner, Monash University

The project

This research proposes an Agentic AI framework powered by Multi-Agent Reinforcement Learning (MARL) to support intelligent energy management in hybrid grids. In this framework, each node, ranging from households, micro-grids, battery storage units, to solar farms, functions as an autonomous learning agent with its own behaviour and objectives, capable of interacting with the grid environment and adapting over time.

Through simulation and modelling that incorporate continuous interaction and reward-based learning, these agents learn to coordinate their actions, deciding when to consume, store, or supply electricity, to optimise energy distribution, enhance grid stability, and reduce carbon emissions. Potential research explorations include methods for balancing local agent objectives (e.g., cost savings, comfort preferences, battery health) with global system objectives (e.g., reducing total CO₂ emissions, minimising peak load) or developing approaches that remain robust in both stationary and non-stationary environments influenced by changing weather patterns and shifting social routines.

By enabling agents to coordinate in a decentralized yet system-aware manner, the resulting model is expected to achieve measurable reductions in CO₂ emissions per kilowatt-hour while improving operational efficiency and resilience. Ultimately, this research provides a scalable, human-centered, and data-driven pathway to support PLN’s long-term decarbonisation roadmap and contribute meaningfully to Indonesia’s journey toward Net Zero 2060.