CVT on-line error measurement hybrid-driven by domain knowledge and Stacking Model
Jingping Wang,Ying Shi,Rui Zhang,Zhonghua Wu,Hao Ye,Shenwei Li
DOI: https://doi.org/10.1016/j.engappai.2023.106710
IF: 8
2023-07-08
Engineering Applications of Artificial Intelligence
Abstract:The performance of Capacitive Voltage Transformer (CVT) degrades over time, making measurement error monitoring a research hotspot in the field of smart grid. At present, these are several challenges such as complex data features, a lack of criteria for selecting optimal measurement models, and low precision. CVT measurement errors can be classified into ideal error and additional one. The former is typically evaluated via mutual information and redundancy within the topology-level transformer group. Considering that a single model cannot process the time series, strong randomness and nonlinearity of the additional error, the Stacking model is selected. Based on the principle of heterogeneity and high-quality, correlation coefficient and feature contribution degree, Random Forest, eXtreme Gradient Boosting, Ridge Regression, K Nearest Neighbors, Support Vector Regression, and Long Short-Term Memory are chosen as base learners through correlation and feature contribution analysis; while extra-trees with strong generalization and robustness is chosen as the meta learner. To improve the measurement precision, the attention-like mechanism is used to scale time and accuracy weights. Finally, according to the power domain knowledge, a linear superposition model is developed to fuse the ideal and additional errors, and thus realize online error measurement for CVTs. The experimental results indicate that the improved Stacking model outperforms mainstream measurement models by an average reduction of 59.47%, and 52.58% in the root mean squared error and the mean absolute error with the best R 2 closest to 1. It not only effectively improves the accuracy but also meets speed requirement for online error measurement.
automation & control systems,computer science, artificial intelligence,engineering, electrical & electronic, multidisciplinary