Wide-Area Monitoring of Power System Using Dynamic Mode Decomposition on Nonlinear Observables

An improved dynamic mode decomposition for real‐time electromechanical oscillation monitoring in power systems: The impact of ultra‐low frequency modes and its removal strategy

“…DMD is one of the data-driven methods that has the potential to be used in the power system application due to its robustness and estimation speed. Specifically, DMD has been used for electromechanical oscillation monitoring [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16] estimation of inertia [17] and frequency [18,19], harmonic distortion monitoring [20,21], fault detection [22], load forecasting [23][24][25] and data-driven control for enhancing rotor angle stability [26,27]. Regarding the oscillation monitoring using DMD, there are still some challenges related to the amount of input data that need to be addressed for achieving real-time oscillation monitoring.…”

“…Feature addressed Target Data type New technique proposed [3][4][5] Modal identification ability Analysis Ringdown None [6] Robustness [7,8] Computational speed [9] Accuracy [2], [10][11][12], [14][15][16] Robustness Improve optimization, data stacking, randomization, data processing, energy evaluation [13] Accuracy Non-linear mapping FIGURE 1 The behaviour of speed signal during real-time operation (ambient and ringdown data).…”

“…ULFMs can also be observed in Figure 9 of [11]. However, the characteristics of ULFMs and trends contained in the ringdown data are not discussed and analysed in [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16].…”

“…In addition, it is still unclear if (and how) we can further reduce the window length as it will increase the computational speed. Therefore, in comparison to our previous work in [9] and the literature presented in [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16], this paper proposes a detrending strategy as a preprocessing step where the measurements from the PMUs are passed through a high-pass IIR filter. This trend elimination improves the DMD's accuracy for the considered three time-window lengths and as a result the time-window length can be reduced up to 4 s. In addition, this paper addresses the following open yet fundamental questions regarding the ULFM: What are the components in the power system which excite the ULFMs?…”

Practical Implementation and Operational Experience of Dynamic Mode Decomposition in Wide-area Monitoring Systems of Italian Power System

“…This approach (also called data stacking) was proposed in [17] and combined with an optimal hard threshold to select the best model order to deal with noise. Nonlinear observables [18] can extend the DMD to better capture the system dynamics. Randomized DMD combined with data stacking [19] can increase the computing efficiency without losing accuracy.…”