Joint causal and anti-causal segmentation and location of transitions in power disturbances

An important step in power disturbance data analysis is to first partition the disturbance sequence into transition and event segments. One main contribution of the paper is the propose of a new segmentation scheme that jointly uses the causal and anti-causal segmentation. We show that the actual location of a trigger point can be accurately estimated by combining the causal and anti-causal transition segments. Another main contribution is a new method for setting the threshold of detection parameter. From the residuals of Kalman filter, a detection parameter is defined and the threshold for this parameter is computed based on pdf estimates and hypothesis tests on measurement sequences with/without disturbances. Using the proposed segmentation and the statistically-based threshold, we show that transition segments, especially the trigger points, are allocated more accurately. Tests are performed on semi-synthetic disturbance sequences containing step, ramp and multi-stage transition. The proposed method is shown to be able to locate fast transitions with very high accuracy. The performance of segmentation against the point on wave, depth of transition, speed of transition, and polluted noise level are also evaluated.