Scholarlink

<p>Voltage stability is a fundamental requirement in modern smart grids, especially under conditions of fluctuating loads and the increasing use of renewable energy sources. In this context, maintaining adaptive and reliable reactive power control is becoming increasingly challenging. To address this issue, this paper proposes a machine learning-based adaptive controller to regulate the voltage of a static VAR compensator (SVC) in a 110&#xa0;kV substation environment. A total of nine supervised regression models were systematically trained and evaluated using a dataset of 2,000 samples generated from simulation scenarios representing realistic operating conditions, including load variations, voltage disturbances, and fault-recovery events. The dataset size was validated through five-fold cross-validation and consistent model performance across all folds. Input characteristics included node voltage deviation, real power (P<sub>t</sub>), and reactive power (Q<sub>t</sub>), and model performance was evaluated through a five-fold cross-validation method. Among the evaluated methods, the Random Forest regression model achieved the best performance, with RMSE = 3.0°, MAE = 2.2°, and R² = 0.92, significantly outperforming conventional control strategies. Furthermore, the feature importance analysis showed that reactive power (Q<sub>t</sub>) was the dominant predictor, contributing approximately 65% of the model’s predictive capability, consistent with established SVC control principles. Besides improved accuracy, computational efficiency and a compact model structure support the practical implementation of edge computing in embedded substation systems. Therefore, the proposed machine learning-based controller offers a scalable and ready-to-deploy solution for modernizing legacy SVC systems and enhancing adaptive voltage regulation capabilities in smart grid applications. The results highlight the potential of data-driven approaches for reliable and efficient voltage control in modern power systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data driven adaptive SVC control for 110 kV substations using machine learning and edge deployment validation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月22日

查看全文
  • Thanh Quang Nguyen,
  • Trung Kien Ngo

摘要

Voltage stability is a fundamental requirement in modern smart grids, especially under conditions of fluctuating loads and the increasing use of renewable energy sources. In this context, maintaining adaptive and reliable reactive power control is becoming increasingly challenging. To address this issue, this paper proposes a machine learning-based adaptive controller to regulate the voltage of a static VAR compensator (SVC) in a 110 kV substation environment. A total of nine supervised regression models were systematically trained and evaluated using a dataset of 2,000 samples generated from simulation scenarios representing realistic operating conditions, including load variations, voltage disturbances, and fault-recovery events. The dataset size was validated through five-fold cross-validation and consistent model performance across all folds. Input characteristics included node voltage deviation, real power (Pt), and reactive power (Qt), and model performance was evaluated through a five-fold cross-validation method. Among the evaluated methods, the Random Forest regression model achieved the best performance, with RMSE = 3.0°, MAE = 2.2°, and R² = 0.92, significantly outperforming conventional control strategies. Furthermore, the feature importance analysis showed that reactive power (Qt) was the dominant predictor, contributing approximately 65% of the model’s predictive capability, consistent with established SVC control principles. Besides improved accuracy, computational efficiency and a compact model structure support the practical implementation of edge computing in embedded substation systems. Therefore, the proposed machine learning-based controller offers a scalable and ready-to-deploy solution for modernizing legacy SVC systems and enhancing adaptive voltage regulation capabilities in smart grid applications. The results highlight the potential of data-driven approaches for reliable and efficient voltage control in modern power systems.