A pre-estimation filtering process of bad data for linear power systems state estimators using PMUs
The paper proposes a specific algorithm for the pre-estimation filtering of bad data (BD) in PMU-based power systems linear State Estimators (SEs). The approach is framed in the context of the so-called real-time SEs that take advantage of the high measurement frame rate made available by PMUs (i.e., 50–60 frames per second). In particular, the proposed algorithm examines PMU measurement innovations for each new received set of data in order to locate anomalies and apply countermeasures. The detection and identification scheme is based on: (i) the forecasted state of the network obtained by means of a linear Kaiman filter, (ii) the current network topology, (iii) the accuracy of the measurement devices and (iv) their location. The incoming measurement from each PMU is considered reliable, or not, according to a dynamic threshold defined as a function of the confidence of the predicted state estimated by using an AutoRegressive Integrated Moving Average (ARIMA) process. The performances of the proposed algorithm are validated with respect to single and multiple bad data of different nature and magnitudes. Furthermore, the algorithm is also tested against faults occurring in the power system to show its robustness during these unexpected operating conditions.