A Comparative Analysis of Decision Trees, Support Vector Machines and Artificial Neural Networks for On-line Transient Stability Assessment

Transient instability is considered the most severe form of instability in power systems with grave socioeconomic repercussions if not prevented. Conventional methods, such as time domain simulations and direct methods impose limitations to fast on-line transient stability assessment in modern power systems. The development of phasor measurement units paved the way for transient stability assessment by means of artificial intelligence for pattern recognition and classification. Many classification algorithms have been reported in the literature for assessing transient stability. This paper aims to provide insights regarding which algorithm is more suitable for a given dataset for power system stability assessment. For this purpose, decision trees, support vector machines and artificial neural networks are investigated for their ability to address the binary stability classification problem in a comparative analysis for two datasets. The two datasets differ in terms of class distribution so that the impact of imbalanced datasets on classification accuracy could also be studied. The above datasets are created using MATLAB with two extension packages of MATPOWER and MATDYN to simulate different contingency scenarios in IEEE-9 bus test system.

I. Introduction

Since the end of the 18th century, power systems have been evolving as the need for electric power has increased rapidly. Although modern power systems have increased in complexity and size since then, the basic structure remains the same. Power generation, power transmission and power distribution are the basic subsystems of a power system [1] . Each subsystem, however, consists of many, hundreds or even thousands, different components that have distinct characteristics, such as generator inertia and circuit breaker response time. These characteristics form the dynamics of a power system and its reaction after a contingency. Power system stability can be defined as the ability of a power system to remain at an equilibrium state of operation or reach a new equilibrium state of operation after a disturbance [1] . Transient stability assessment (TSA) is associated with the determination of whether or not a system will remain in synchronism after a severe disturbance.