[Paper Review] A Review of Neural Network Based Machine Learning Approaches for Rotor Angle Stability Control

Reza Yousefian, Sukumar Kamalasadan|arXiv (Cornell University)|Jan 5, 2017

Power System Optimization and Stability80 references21 citations

TL;DR

This paper reviews neural network-based machine learning approaches—particularly Reinforcement Learning (RL) and Supervised Learning (SL)—for rotor angle stability control in modern power systems. It evaluates their effectiveness in wide-area monitoring and control, transient stability assessment, and renewable energy integration, demonstrating that RL and SL methods outperform classical controllers in handling nonlinearities and uncertainties.

ABSTRACT

This paper reviews the current status and challenges of Neural Networks (NNs) based machine learning approaches for modern power grid stability control including their design and implementation methodologies. NNs are widely accepted as Artificial Intelligence (AI) approaches offering an alternative way to control complex and ill-defined problems. In this paper various application of NNs for power system rotor angle stabilization and control problem is discussed. The main focus of this paper is on the use of Reinforcement Learning (RL) and Supervised Learning (SL) algorithms in power system wide-area control (WAC). Generally, these algorithms due to their capability in modeling nonlinearities and uncertainties are used for transient classification, neuro-control, wide-area monitoring and control, renewable energy management and control, and so on. The works of researchers in the field of conventional and renewable energy systems are reported and categorized. Paper concludes by presenting, comparing and evaluating various learning techniques and infrastructure configurations based on efficiency.

Motivation & Objective

• To evaluate the current state of neural network-based machine learning techniques for rotor angle stability control in modern power systems.
• To analyze the application of Supervised Learning (SL) and Reinforcement Learning (RL) in wide-area control (WAC) and transient stability assessment.
• To compare the performance of AI-based controllers with classical control methods in handling nonlinearities, uncertainties, and time-varying dynamics.
• To examine the implementation challenges and reliability concerns of learning-based controllers in real-time power system applications.
• To assess the role of neural networks in renewable-integrated systems, especially for voltage and power oscillation control.

Proposed method

• The paper reviews various neural network architectures, including deep neural networks and radial basis function networks (RBFNN), applied to power system dynamics.
• It analyzes Supervised Learning (SL) methods using labeled input-output data for system modeling, control, and prediction in microgrids and renewable integration.
• Reinforcement Learning (RL) techniques such as Adaptive Dynamic Programming (ADP) and Heuristic Dynamic Programming (HDP) are evaluated for optimal control policy learning through interaction with the system.
• The study examines Adaptive Critic Designs (ACD) and Neural Network-based controllers for wide-area monitoring and control (WAM), including damping inter-area oscillations.
• The paper evaluates the use of neural networks in transient stability assessment, including clustering coherent areas and classifying system behavior post-fault.
• It compares different infrastructure configurations and learning techniques based on computational efficiency, accuracy, and real-time feasibility.

Experimental results

Research questions

• RQ1 How effective are Reinforcement Learning and Supervised Learning algorithms in improving rotor angle stability in large-scale power systems?
• RQ2 What are the key advantages and limitations of using neural networks for wide-area control (WAC) in power systems with high renewable penetration?
• RQ3 How do AI-based controllers compare to classical control methods in handling nonlinearities and uncertainties in power system dynamics?
• RQ4 What are the main challenges in deploying learning-based controllers in real-time power system applications?
• RQ5 To what extent can neural networks enhance transient stability assessment and control in power systems with multiple contingencies?

Key findings

• Neural network-based controllers, especially those using Reinforcement Learning (RL), demonstrate superior performance in damping inter-area oscillations and improving transient stability under fault conditions.
• Supervised Learning (SL) methods effectively model nonlinear system dynamics and have been successfully applied in microgrid control, DFIG-based wind turbines, and rectifier/inverter control with better dynamic response than PID controllers.
• RL-based controllers such as those using Adaptive Dynamic Programming (ADP) and Heuristic Dynamic Programming (HDP) show strong adaptability and predictive capability in wide-area control schemes.
• Neural networks have been shown to outperform classical controllers in handling uncertainties and nonlinearities, particularly in renewable-integrated systems.
• The use of RBFNN in Wide-Area Monitoring (WAM) improves system response during transient events, although communication time delays and multiple wind farm integration remain open challenges.
• Despite promising results, the heuristic nature and high dimensionality of learning-based controllers raise reliability concerns, limiting full deployment in real-time power systems.