A Data Analytics Method to Pinpoint Causes for Poor Performance of Real Time Distribution Power Flow

Precise real-time monitoring within Distribution Management Systems (DMS) is essential and power flow-based techniques are mostly utilized in practical applications for achieving this objective. The accuracy of near real time power flow results depends critically on the load estimation used. Therefore, this paper focuses on the causes for poor load allocation and proposes a three step data analytics-based approach to pinpoint possible causes affecting poor near real-time power flow solutions. The proposed method first identifies the main parameters that affect bus load allocation performance the most. In the second step, the approach employs logistic regression to pinpoint unique features associated with poor performing feeders. In the last step, these significant parameters are used to pinpoint the possible causes for the poor performance. Test results using a set of savecases from a real distribution system are given to illustrate the effectiveness of the proposed method.