A Convolutional Neural Network based Pattern Classification Approach for Dynamic Security Assessment with Inadequate Energy Sources

Abstract

The security-state categorization of complex power system networks based on "transient stability" is proposed in this research study as a "Convolutional Neural Network (CNN)" based pattern classification approach. The “pre- contingency operating states” of a “power system network” served as CNN’s input. The focus was on predicting the system's post-contingency stability condition, so the Critical Clearance Time (CCT) was used as the primary metric for categorizing the “pre-contingency operational states” into "secure" and "insecure" classes. The recommended method was successfully applied to the “IEEE 39-bus system”, and it was discovered that the CNN classifier can classify the power system's pre-contingency operational states based on an accurate forecast of the conditions that will lead to future post- contingency transient stability.