Online event-triggered switching load frequency control of islanded microgrids through machine learning

Area of Science:

• Electrical Engineering
• Control Systems
• Machine Learning

Background:

• Modern islanded microgrids face stability challenges due to high renewable energy penetration and dynamic characteristics.
• Conventional load frequency control (LFC) strategies exhibit performance limitations in these complex systems.

Purpose of the Study:

• To design and develop an adaptive online event-triggered LFC framework for islanded microgrids.
• To enhance frequency regulation and system stability using machine learning techniques.

Main Methods:

• Development of a Neural-PI controller pre-trained for complex system dynamics.
• Implementation of an intelligent event-triggered switching mechanism for real-time control gain adjustment.
• Evaluation on the IEEE 14-bus test system using Python and Jupyter Notebook.

Main Results:

• The Neural-PI controller demonstrated superior performance compared to conventional PID controllers.
• Achieved a 58% reduction in peak frequency deviations and a 33% improvement in settling time.
• Reduced control actions by 95% via event-triggered switching, increasing renewable hosting capacity by 19% and reducing operational costs by 37%.

Conclusions:

• The adaptive LFC framework effectively enhances stability and control efficiency in islanded microgrids.
• The Neural-PI controller offers significant improvements in frequency regulation, especially during renewable intermittency.
• The proposed method is highly suitable for islanded and low-inertia grids, optimizing operational costs and renewable integration.