Topology Identification for Distribution Systems Based on Bayesian Nonparametric Model and Gaussian Mixture Ensemble Kalman Filter

To ensure the secure and reliable operation of active distribution systems (ADSs), it is essential to efficiently and accurately identify their time-varying topology. This paper proposes an estimation-based topology identification (TI) method for ADSs considering topology changes. The TI problem is formulated by a forecasting-aided estimation model, which is depicted by a discrete-time switching state-space model. The topology change process is described by a Bayesian nonparametric model that allows TI with limited knowledge of topology libraries. Then the Gaussian mixture ensemble Kalman filter (GM-EnKF) is employed to carry out the TI, where the underlying probability density is approximated by using Gaussian mixture model. In this case, the proposed method become capable of handling the unavoidable non-Gaussian noises in realistic distribution systems. Test results on various ADSs validates the effectiveness of the proposed GM-EnKF-based TI method on accurate and efficient topology identification.