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).

Online Distributed Security Feature Selection Based on Big Data in Power System Operation

Distributed Feature Selection for Power System Dynamic Security Region Based on Grid-Partition and Fuzzy-Rough Sets

A Causality Based Feature Selection Approach for Data-Driven Dynamic Security Assessment

Sequential Feature Selection for Power System Event Classification Utilizing Wide-Area PMU Data

A Distributed Computing Platform Supporting Power System Security Knowledge Discovery Based on Online Simulation

Selecting Decision Trees for Power System Security Assessment

Hybrid Model and Data Driven Concepts for Power System Security Feature Selection and Knowledge Discovery:Key Technologies and Engineering Application

An Integrated Online Dynamic Security Assessment System for Improved Situational Awareness and Economic Operation

Online multi-fault power system dynamic security assessment driven by hybrid information of anticipated faults and pre-fault power flow

Power system online security operational trend analysis and simulation results

Power System Security Weak-Point and Kernel-Feature Identification Based on Hierarchical Pattern Recognition

A Collaborative Characteristic Event Sequence Based Identification Method for Cyberattacks in Cyber-Physical System

Enhancing Security And Resilience Of Bulk Power Systems Via Multisource Big Data Learning

Study on Practical Dynamic Security Region of Power System Based on Big Data

Knowledge Management and Security Early Warning Based on Big Simulation Data in Power Grid Operation

Power Grid On-line Security and Stability Analysis and Forewarning System

A data-driven approach for online dynamic security assessment with spatial-temporal dynamic visualization using random bits forest

Development and application of on-line dynamic security early warning and preventive control system in China

Preventive Dynamic Security Control of Power Systems Based on Pattern Discovery Technique

Automatic Learning of Fine Operating Rules for Online Power System Security Control

A Vulnerability Identification Method of Power System Based on Infomap Community Discovery Algorithm