Online Pattern Recognition and Data Correction of PMU Data Under GPS Spoofing Attack

Abstract

Smart grids are increasingly dependent on data with the rapid development of communication and measurement. As one of the important data sources of smart grids, phasor measurement unit (PMU) is facing the high risk from attacks. Compared with cyber attacks, global position system (GPS) spoofing attacks (GSAs) are easier to implement because they can be exploited by portable devices, without the need to access the physical system. Therefore, this paper proposes a novel method for pattern recognition of GSA and an additional function of the proposed method is the data correction to the phase angle difference (PAD) deviation. Specifically, this paper analyzes the effect of GSA on PMU measurement and gives two common patterns of GSA, i.e., the step attack and the ramp attack. Then, the method of estimating the PAD deviation across a transmission line introduced by GSA is proposed, which does not require the line parameters. After obtaining the estimated PAD deviations, the pattern of GSA can be recognized by hypothesis tests and correlation coefficients according to the statistical characteristics of the estimated PAD deviations. Finally, with the case studies, the effectiveness of the proposed method is demonstrated, and the success rate of the pattern recognition and the online performance of the proposed method are analyzed.

A new four‐step method to identify the parameters of transmission line based on SCADA data

“…The general framework of the proposed four-step method is presented in this subsection, which can simplify and solve the formulation for the estimation of VPADs in (15), and the whole process is shown in Figure 5.…”

“…But only based on (15), the VPADs are difficult to solve. The reason is that, on the one hand, (15) contains n+ΔT variables to be solved, thus, the calculation burden is huge; on the other hand, (15) solves n+ΔT unknown variables based on the SCADA data at n+ΔT snapshots, which has low redundancy, thus, it is easy to be affected by noise, resulting in accurate estimation results.…”

“…2. In Equation (15), n+ΔT unknown variables are solved based on the n+ΔT snapshots of SCADA data, while in Equation (20), 1 unknown variable is solved based on the n+ΔT snapshots of SCADA data, which improves the redundancy and can resist the influence of noise better. 3.…”

“…For the methods of Category 1, as the voltage and current phasors can be provided by PMU [6], the linear equations based on least square estimation to identify ParTL can be established in [7–10], which work well even if there are measurement noise in PMU data. However, since bad data and synchronization errors may be introduced by the data loss [11, 12], gross error [13], manipulation attacks [14, 15], and time stamp shift [16, 17], the results with field data may contain errors [18].…”

“…Therefore, the methods based on live measurements have attracted a lot of attention, they could be generally classified into three categories according to the data source, that is, Category 1: methods based on phasor measurement unit (PMU) measurements; Category 2: For the methods of Category 1, as the voltage and current phasors can be provided by PMU [6], the linear equations based on least square estimation to identify ParTL can be established in [7][8][9][10], which work well even if there are measurement noise in PMU data. However, since bad data and synchronization errors may be introduced by the data loss [11,12], gross error [13], manipulation attacks [14,15], and time stamp shift [16,17], the results with field data may contain errors [18].…”