Abstract:Power systems transport an increasing amount of electricity, and in the future, involve more distributed renewables and dynamic interactions of the equipment. The system response to disturbances must be secure and predictable to avoid power blackouts. The system response can be simulated in the time domain. However, this dynamic security assessment (DSA) is not computationally tractable in real-time. Particularly promising is to train decision trees (DTs) from machine learning as interpretable classifiers to predict whether the system-wide responses to disturbances are secure. In most research, selecting the best DT model focuses on predictive accuracy. However, it is insufficient to focus solely on predictive accuracy. Missed alarms and false alarms have drastically different costs, and as security assessment is a critical task, interpretability is crucial for operators. In this work, the multiple objectives of interpretability, varying costs, and accuracies are considered for DT model selection. We propose a rigorous workflow to select the best classifier. In addition, we present two graphical approaches for visual inspection to illustrate the selection sensitivity to probability and impacts of disturbances. We propose cost curves to inspect selection combining all three objectives for the first time. Case studies on the IEEE 68 bus system and the French system show that the proposed approach allows for better DT-selections, with an 80% increase in interpretability, 5% reduction in expected operating cost, while making almost zero accuracy compromises. The proposed approach scales well with larger systems and can be used for models beyond DTs. Hence, this work provides insights into criteria for model selection in a promising application for methods from artificial intelligence (AI).

Multi-support Vector Machine Power System Transient Stability Assessment Based on Relief Algorithm

Real-time Transient Stability Assessment in Power System Based on Improved SVM

Clustering Based on Support Vector Regression Model and Its Application in Power System Transient Stability Prediction

Selecting Decision Trees for Power System Security Assessment

Transient Stability Assessment Model With Sample Selection Method Based on Spatial Distribution

Transient Stability Assessment of Power Systems Based on the Transformer and Neighborhood Rough Set

A multifeature fusion approach for power system transient stability assessment using PMU data

An Improved Power System Transient Stability Prediction Model Based on mRMR Feature Selection and WTA Ensemble Learning

A Novel Stability Classifier Based on Reformed Support Vector Machines for Online Stability Assessment

Support Vector Machine For Transient Stability Assessment: A Review

Conservative Online Transient Stability Assessment in Power System Based on Concept of Stability Region

A Novel Feature Selection Method for Power System Transient Stability Assessment Based on Interaction Gain

Power system transient security assessment based on multi-channel time series data mining

Interaction Gain and Rotation Forest Based Key Feature Selection Method for Power System Transient Stability Assessment

Data-driven Transient Stability Assessment Using Sparse PMU Sampling and Online Self-check Function

Transient stability assessment based on neighborhood rough set theory and support vector machine

Information Entropy Based Prioritization Strategy for Data-Driven Transient Stability Batch Assessment

Real-Time Transient Stability Assessment Based On Deep Recurrent Neural Network

Power System Transient Stability Analysis Based on PMU and Hybrid Support Vector Machine

Deep learning-based transient stability assessment framework for large-scale modern power system

Real-time Emergency Control Decision in Power System Based on Support Vector Machines