Real-time voltage security assessment using adaptive fuzzified decision tree algorithm

This paper presents the adaptive machine learning approach for voltage security classification. The online probabilistic assessment of voltage security is done using decision tree, which are updated periodically. The advantage of fuzzified decision tree support is robust classification of voltage security in the upcoming samples. Offline learning datasets are generated for each N-1 contingency conditions using continuation power flow method. Security classes are defined by threshold value of maximum loadability margins, calculated using the continuation power flow method. The proposed method is tested on two IEEE bus systems. Classification accuracy from a value of 88% to finally 100% is achieved for line outage no. 5 in IEEE-30 bus system and 100% for line outages no. 51 and 172, in IEEE-118 bus system. The result shows the fast and accurate classification for online decisions. This confirms the proposed method validity and suitability for the energy management system in online control decisions.