Random Forest-assisted Dynamic TRM Estimation for Improved ATC in Renewable-rich Power Systems

Abstract

Accurate Available Transfer Capability (ATC) assessment is critical for reliable operation of renewable-rich power systems. Conventional dynamic Transmission Reliability Margin (TRM) methods based on rolling-window statistics mainly react to recent volatility, limiting responsiveness during rapid system changes. This paper proposes a Random Forest (RF)-assisted dynamic TRM framework where a lightweight RF regression model predicts short-term volatility using historical forecast errors and operating-condition features. The predicted volatility is incorporated into an adaptive confidence factor, K(t), enabling anticipatory TRM adjustment while preserving the original probabilistic framework. Validation on the IEEE 24-bus Reliability Test System with integrated wind generation demonstrates improved performance, achieving 97.2% coverage during ramp events compared with 91.7% for the rolling-window approach and 69.4% for the static method. The RF model achieved RMSE = 3.60 MW and R2 = 0.825 for 3-step-ahead volatility prediction, showing strong potential for renewable-integrated Australasian power systems.