Small-signal stability analysis in smart grids: An approach based on distributed decision trees

• Published in: Electric power systems research Vol. 203; p. 107651
• Main Authors: da Cunha, Guilherme L., Fernandes, Ricardo A.S., Fernandes, Tatiane Cristina C.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.02.2022; Elsevier Science Ltd
• Subjects: Contingency; Control systems; Damping; Decision analysis; Decision tree; Decision trees; Electricity distribution; Electromechanical oscillations; Evaluation; Machine learning; Measurement; Measuring instruments; Oscillations; Phasor measurement units; Phasors; Small-signal stability; Smart grid; Stability analysis; Systems stability; Electromechanical oscillations; Small-signal stability; Phasor measurement units; Decision tree; Machine learning
• ISSN: 0378-7796; 1873-2046
• DOI: 10.1016/j.epsr.2021.107651

•Each distributed DT uses only local data to assess the small-signal stability.•The TSO can assess the stability from any of the available individual DT predictions.•The distributed DTs require a low computational burden and a reduced number of features.•The higher classification ratios provided by distributed DTs show their efficiency. Accurate information on the dynamic behavior of a smart grid is essential for its effective control and reliable operation. Specifically, ensuring that the electromechanical oscillations are properly damped is needed for the dynamic security operation of an interconnected system. Since the model-based method to evaluate these oscillations is subject to uncertainties, the data from a wide-area measurement system can assist in this challenging task. In this context, this paper proposes a decentralized approach based on decision trees to classify the interarea oscillations through data measured in the generation buses. For each generation bus, there is an individual decision tree capable of evaluating the interarea oscillations damping. The decision trees presented a low computational burden and, consequently, they can be embedded in phasor measurement units with low-cost hardware. The results obtained with the 68-bus test system show that the classifications provided by individual decision trees are accurate, even when contingency scenarios were evaluated, reaching accuracies greater than 0.93. Thus, the possibility of analyzing the power system stability to small disturbances from local information is one of the main contributions of the proposed approach to advance the state-of-the-art, since the use of distributed decision trees is robust even when there is loss of information from a specific generator.