Hoeffding-Tree-Based Learning from Data Streams and Its Application in Online Voltage Security Assessment

• Main Author: Nie, Zhijie
• Other Authors: Electrical and Computer Engineering
• Format: Others
• Published: Virginia Tech 2017
• Subjects: Power Systems Stability; Voltage Security Assessment; Machine Learning; Decision Trees; Hoeffding Trees
• Online Access: http://hdl.handle.net/10919/78805

According to the proposed definition and classification of power system stability addressed by IEEE and CIGRE Task Force, voltage stability refers to the stability of maintaining the steady voltage magnitudes at all buses in a power system when the system is subjected to a disturbance from a given operating condition (OC). Cascading outage due to voltage collapse is a probable consequence during insecure voltage situations. In this regard, fast responding and reliable voltage security assessment (VSA) is effective and indispensable for system to survive in conceivable contingencies. This paper aims at establishing an online systematic framework for voltage security assessment with high-speed data streams from synchrophasors and phasor data concentrators (PDCs). Periodically updated decision trees (DTs) have been applied in different subjects of security assessments in power systems. However, with a training data set of operating conditions that grows rapidly, re-training and restructuring a decision tree becomes a time-consuming process. Hoeffding-tree-based method constructs a learner that is capable of memory management to process streaming data without retaining the complete data set for training purposes in real-time and guarantees the accuracy of learner. The proposed approach of voltage security assessment based on Very Fast Decision Tree (VFDT) system is tested and evaluated by the IEEE 118-bus standard system. === Master of Science