cybersecurity challenges. Existing Energy Management Systems (EMS), Supervisory Control and Data Acquisition (SCADA) systems, and Artificial Intelligence (AI)-based decision support tools primarily function as passive advisory systems, requiring human operators to interpret recommendations and execute control actions. Such centralized approaches are increasingly inadequate for managing the dynamic, data-intensive, and highly distributed nature of future power systems.

Recent advances in Agentic Artificial Intelligence (Agentic AI) provide an opportunity to redefine smart grid operation. Unlike conventional AI models that perform isolated prediction or optimization tasks, Agentic AI comprises autonomous intelligent agents capable of perceiving their environment, reasoning over multiple information sources, planning complex actions, collaborating with other agents, learning continuously from experience, and safely executing decisions with minimal human intervention. This paradigm enables the development of self-managing, resilient, and adaptive power systems capable of autonomous operation while maintaining compliance with engineering constraints and regulatory requirements.

This project aims to develop an Agentic AI-enabled Autonomous Smart Grid Ecosystem that integrates distributed AI agents with digital twins, physics-informed machine learning, explainable AI, Internet of Things (IoT) sensing, and advanced power system analytics to create the next generation of intelligent electrical networks. The proposed ecosystem will consist of multiple specialized AI agents operating collaboratively across transmission systems, distribution networks, renewable energy plants, microgrids, substations, industrial facilities, and consumer premises. Each agent will possess dedicated responsibilities, including monitoring, forecasting, diagnosis, optimization, decision-making, communication, and autonomous control, while coordinating with neighboring agents through secure multi-agent communication protocols.

The proposed framework will incorporate a real-time Digital Twin of the electrical network that continuously synchronizes operational data obtained from SCADA systems, Phasor Measurement Units (PMUs), Advanced Metering Infrastructure (AMI), weather services, renewable generation forecasts, market information, and IoT-enabled sensors. The digital twin will enable Agentic AI to evaluate multiple operational scenarios, simulate contingencies, predict system responses, and verify corrective actions before implementation in the physical grid, thereby significantly improving operational safety and reducing the risks associated with autonomous control.

To ensure engineering reliability, the project will introduce Physics-Informed Agentic AI, whereby autonomous decision-making is constrained by fundamental electrical engineering principles including power flow equations, voltage stability limits, frequency regulation requirements, thermal equipment ratings, protection coordination, and network operational constraints. By embedding domain knowledge directly into AI reasoning processes, the proposed framework seeks to overcome one of the major limitations of current black-box AI systems while enhancing trustworthiness, robustness, and regulatory acceptance.

Another key innovation of this project is the integration of Explainable Agentic AI (XAgenticAI), allowing every autonomous decision to be accompanied by interpretable engineering justifications. Grid operators will receive clear explanations regarding the causes of disturbances, selected control actions, expected operational impacts, and confidence levels, thereby improving situational awareness and facilitating human-AI collaboration in mission-critical environments.

The research programme is organised into ten interconnected research thrusts that collectively establish a comprehensive autonomous smart grid ecosystem:

1. Agentic AI for Autonomous Smart Grid Operation and Self-Healing Power Networks.
2. Agentic AI-Based Distributed Energy Resource Coordination.
3. Physics-Informed Agentic AI for Secure and Reliable Power System Control.
4. Agentic AI-Driven Digital Twin Framework for Smart Grid Management.
5. Explainable Agentic AI for Intelligent Grid Decision Support.
6. Agentic AI for Autonomous Power Quality Monitoring and Corrective Action Recommendation.
7. Multi-Agent Agentic AI for Grid Resilience Against Cyber-Physical Threats and Extreme Events.
8. Agentic AI for Renewable-Based Microgrid Energy Management.
9. Agentic AI for Predictive Maintenance and Intelligent Asset Health Management.
10. Agentic AI-Enabled Hierarchical Energy Management System for Next-Generation Smart Grids.

The project adopts a hierarchical multi-layer architecture spanning the grid edge, feeder, substation, distribution, transmission, and system operator levels. Intelligent agents deployed at each layer will cooperate through secure communication networks to optimize renewable integration, Volt-VAR control, congestion management, demand response, energy storage coordination, predictive maintenance scheduling, market participation, and power quality mitigation. Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), engineering knowledge graphs, reinforcement learning, graph neural networks, and multi-agent reinforcement learning will be integrated with conventional optimization methods to enable advanced reasoning and collaborative autonomous decision-making.

Experimental validation will be conducted using IEEE benchmark systems (e.g., IEEE 33-bus, 69-bus, 118-bus, and transmission networks) through real-time co-simulation using platforms such as OpenDSS, MATPOWER, GridLAB-D, DIgSILENT PowerFactory, MATLAB/Simulink, OPAL-RT, or Typhoon HIL. Hardware validation will incorporate IoT devices, Raspberry Pi, ESP32/ESP8266 controllers, PMU emulators, renewable energy simulators, and industrial communication protocols including IEC 61850 and MQTT, enabling practical demonstration of autonomous grid operation.

The anticipated outcomes include novel Agentic AI algorithms for autonomous power system operation, a modular multi-agent software platform, digital twin-enabled smart grid demonstrators, explainable autonomous control frameworks, open benchmark datasets, high-impact journal publications, patents, industry-ready technologies, and policy recommendations supporting the transition toward intelligent, resilient, and carbon-neutral power systems. The project is expected to establish a new paradigm in electrical power engineering by shifting from operator-assisted grid management to fully autonomous, collaborative, and self-optimising smart grids capable of supporting future net-zero energy ecosystems. Such advances will significantly enhance grid reliability, renewable energy integration, operational efficiency, cybersecurity resilience, and sustainability, positioning the research at the forefront of next-generation intelligent energy systems.

For enquiries, please contact Ir. Ts. Dr. Charles Raymond A/L Sarimuthu, Monash University.

cybersecurity challenges. Existing Energy Management Systems (EMS), Supervisory Control and Data Acquisition (SCADA) systems, and Artificial Intelligence (AI)-based decision support tools primarily function as passive advisory systems, requiring human operators to interpret recommendations and execute control actions. Such centralized approaches are increasingly inadequate for managing the dynamic, data-intensive, and highly distributed nature of future power systems.

Recent advances in Agentic Artificial Intelligence (Agentic AI) provide an opportunity to redefine smart grid operation. Unlike conventional AI models that perform isolated prediction or optimization tasks, Agentic AI comprises autonomous intelligent agents capable of perceiving their environment, reasoning over multiple information sources, planning complex actions, collaborating with other agents, learning continuously from experience, and safely executing decisions with minimal human intervention. This paradigm enables the development of self-managing, resilient, and adaptive power systems capable of autonomous operation while maintaining compliance with engineering constraints and regulatory requirements.

This project aims to develop an Agentic AI-enabled Autonomous Smart Grid Ecosystem that integrates distributed AI agents with digital twins, physics-informed machine learning, explainable AI, Internet of Things (IoT) sensing, and advanced power system analytics to create the next generation of intelligent electrical networks. The proposed ecosystem will consist of multiple specialized AI agents operating collaboratively across transmission systems, distribution networks, renewable energy plants, microgrids, substations, industrial facilities, and consumer premises. Each agent will possess dedicated responsibilities, including monitoring, forecasting, diagnosis, optimization, decision-making, communication, and autonomous control, while coordinating with neighboring agents through secure multi-agent communication protocols.

The proposed framework will incorporate a real-time Digital Twin of the electrical network that continuously synchronizes operational data obtained from SCADA systems, Phasor Measurement Units (PMUs), Advanced Metering Infrastructure (AMI), weather services, renewable generation forecasts, market information, and IoT-enabled sensors. The digital twin will enable Agentic AI to evaluate multiple operational scenarios, simulate contingencies, predict system responses, and verify corrective actions before implementation in the physical grid, thereby significantly improving operational safety and reducing the risks associated with autonomous control.

To ensure engineering reliability, the project will introduce Physics-Informed Agentic AI, whereby autonomous decision-making is constrained by fundamental electrical engineering principles including power flow equations, voltage stability limits, frequency regulation requirements, thermal equipment ratings, protection coordination, and network operational constraints. By embedding domain knowledge directly into AI reasoning processes, the proposed framework seeks to overcome one of the major limitations of current black-box AI systems while enhancing trustworthiness, robustness, and regulatory acceptance.

Another key innovation of this project is the integration of Explainable Agentic AI (XAgenticAI), allowing every autonomous decision to be accompanied by interpretable engineering justifications. Grid operators will receive clear explanations regarding the causes of disturbances, selected control actions, expected operational impacts, and confidence levels, thereby improving situational awareness and facilitating human-AI collaboration in mission-critical environments.

The research programme is organised into ten interconnected research thrusts that collectively establish a comprehensive autonomous smart grid ecosystem:

1. Agentic AI for Autonomous Smart Grid Operation and Self-Healing Power Networks.
2. Agentic AI-Based Distributed Energy Resource Coordination.
3. Physics-Informed Agentic AI for Secure and Reliable Power System Control.
4. Agentic AI-Driven Digital Twin Framework for Smart Grid Management.
5. Explainable Agentic AI for Intelligent Grid Decision Support.
6. Agentic AI for Autonomous Power Quality Monitoring and Corrective Action Recommendation.
7. Multi-Agent Agentic AI for Grid Resilience Against Cyber-Physical Threats and Extreme Events.
8. Agentic AI for Renewable-Based Microgrid Energy Management.
9. Agentic AI for Predictive Maintenance and Intelligent Asset Health Management.
10. Agentic AI-Enabled Hierarchical Energy Management System for Next-Generation Smart Grids.

The project adopts a hierarchical multi-layer architecture spanning the grid edge, feeder, substation, distribution, transmission, and system operator levels. Intelligent agents deployed at each layer will cooperate through secure communication networks to optimize renewable integration, Volt-VAR control, congestion management, demand response, energy storage coordination, predictive maintenance scheduling, market participation, and power quality mitigation. Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), engineering knowledge graphs, reinforcement learning, graph neural networks, and multi-agent reinforcement learning will be integrated with conventional optimization methods to enable advanced reasoning and collaborative autonomous decision-making.

Experimental validation will be conducted using IEEE benchmark systems (e.g., IEEE 33-bus, 69-bus, 118-bus, and transmission networks) through real-time co-simulation using platforms such as OpenDSS, MATPOWER, GridLAB-D, DIgSILENT PowerFactory, MATLAB/Simulink, OPAL-RT, or Typhoon HIL. Hardware validation will incorporate IoT devices, Raspberry Pi, ESP32/ESP8266 controllers, PMU emulators, renewable energy simulators, and industrial communication protocols including IEC 61850 and MQTT, enabling practical demonstration of autonomous grid operation.

The anticipated outcomes include novel Agentic AI algorithms for autonomous power system operation, a modular multi-agent software platform, digital twin-enabled smart grid demonstrators, explainable autonomous control frameworks, open benchmark datasets, high-impact journal publications, patents, industry-ready technologies, and policy recommendations supporting the transition toward intelligent, resilient, and carbon-neutral power systems. The project is expected to establish a new paradigm in electrical power engineering by shifting from operator-assisted grid management to fully autonomous, collaborative, and self-optimising smart grids capable of supporting future net-zero energy ecosystems. Such advances will significantly enhance grid reliability, renewable energy integration, operational efficiency, cybersecurity resilience, and sustainability, positioning the research at the forefront of next-generation intelligent energy systems.

For enquiries, please contact Ir. Ts. Dr. Charles Raymond A/L Sarimuthu, Monash University.