Data-Driven Short-Term Voltage Stability Assessment Using Convolutional Neural Networks Considering Data Anomalies and Localization

Abstract

Short-term voltage stability of power systems is governed by load dynamics, especially the proportion of small induction motors prevalent in residential air-conditioners. It is essential to efficiently monitor short-term voltage stability in real-time by detailed data analytics on voltage measurements acquired from phasor measurement units (PMUs). It is likewise critical to identify the location of faults resulting in short-term voltage stability issues for effective remedial actions. This paper proposes a timeseries deep learning framework using 1D-convolutional neural networks (1D-CNN) for real-time shortterm voltage stability assessment (STVSA), which relies on a limited number of phasor measurement units (PMU) voltage samples. A two-stage STVSA application is proposed. The first stage comprises a 1D-CNN-based fast voltage collapse detector. The second stage comprises of 1D-CNN-based regressor to quantify the severity of the short-term voltage stability event. Two novel indices are presented, and their predicted future values are used to quantify the severity of short-term voltage stability events. This work also considers DB-SCAN clustering-based fault detection and physics-based fault localization for effective short-term voltage stability assessment and remedial actions by identifying the most critical PMUs. A bad data pre-processing technique is also included to mitigate the impact of missing data and outliers on shortterm voltage stability assessment accuracy. The proposed framework is validated using the standard IEEE test systems and compared against other machine learning models to demonstrate the superiority of 1D-CNN-based time-series deep learning for short-term voltage stability assessment.

Data-Driven Short-Term Voltage Stability Assessment Considering Sample Imbalance and Overlapping

“…To further demonstrate the superiority of the proposed method, four other methods, including LightgGBM, LSTM (Zhang et al, 2021), 1D-CNN (Rizvi et al, 2021), DT, and ANN, are selected for comparison. Table 3 compares the classification performance of the proposed CasLightGBM with other machine learning methods.…”

“…The ensemble models improve the prediction accuracy by integrating weak learners, but they are prone to overfitting. In Rizvi et al's (2021) study, a new approach based on a convolutional neural network (CNN) is introduced for STVS assessment considering data anomalies and fault localization. Moreover, the long-short-term-memory network (LSTM) is used to extract voltage stability information from the time-varying features in Zhang et al (2021).…”

“…Combined with the spatial-temporal characteristics, the deep graph neural network proposed by Luo et al (2021) and Zhong et al (2022) has shown higher accuracy and better adaptability. The deep learning algorithms in Luo et al (2021), Rizvi et al (2021), Zhang et al (2021), andZhong et al (2022) can establish an excellent input-output mapping relationship, but problems such as large demand for data samples and long training time remain to be solved.…”

Power system short‐term voltage stability assessment based on improved CatBoost with consideration of model confidence

“…To fully learn the time dependencies from the system dynamic trajectories after the fault, a long short-term memory (LSTM)-based model is employed to determine STVS status in Zhang et al 14 Moreover, an LSTM-based deep recurrent neural network model is employed to learn both spatial and temporal dependencies from smart grids in Zhu et al 15 In Luo et al 16 and Cai and Hill, 17 a graph neural network-based model and a gated recurrent graph attention network-based model are proposed, respectively, to estimate the STVS status, and they can adapt to topological changes of the power grid. The deep learning algorithms [13][14][15][16][17] can establish an excellent input-output mapping relationship, but problems such as large demand for data instances and long training time remain to be solved.…”

“…Moreover, above typical machine learning-based STVS assessment approaches [11][12][13][14][15][16][17] have some limitations, for example, they do not consider the credibility of assessment results. In Zhang et al 18 and Ren et al, 19 a decision-making rule is adopted for the classification credibility check of the hybrid randomized ensemble model, and credible decision parameters are optimized to achieve the optimal assessment speed and accuracy.…”