Scholarlink

<p>Underground cable networks are essential in modern urban power distribution systems due to reliable electricity delivery in densely populated regions. The increasing penetration of renewable energy resources and fluctuating load demand is creating dynamic operational conditions that tend to influence the thermal behaviour of the underground cables. The conventional static thermal rating methods tend to rely on conservative environmental assumptions, thereby limiting the actual current-carrying capacity of cable systems and reducing network efficiency. The primary challenge lies in accurately predicting the thermal state of the underground cables under varying load currents, soil temperature, moisture conditions, and environmental uncertainties while simultaneously supporting intelligent operational decisions for power distribution networks. Conventional machine learning models provide accurate predictions, whereas fuzzy systems represent uncertainty, but they often lack adaptive learning capabilities. To address this problem, this study proposes a novel Adaptive Neuro-Fuzzy Thermal Decision Optimisation (ANFTDO) algorithm for intelligent underground cable management. The proposed method combines the adaptive neuro-fuzzy inference engine to model the nonlinear thermal interactions within underground cable systems. A deep learning–assisted prediction module estimates cable temperature dynamics, while a reinforcement-based optimisation layer continuously adjusts fuzzy membership functions based on historical operational data. The model further adopts a dynamic decision module that evaluates thermal risk levels and recommends an optimal load-allocation method for the cable network. Results are showing that the proposed ANFTDO model is achieving a MAE of 0.94&#xa0;°C, RMSE of the 1.39&#xa0;°C, prediction accuracy of the 96.9%, thermal risk detection rate of the 95.6%, and cable utilisation improvement of the 90.9%, which is performing better than the conventional AI- and fuzzy-based methods in both prediction reliability and operational efficiency.</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

AI-driven adaptive fuzzy decision intelligence for dynamic thermal rating and smart management of underground cable networks

  • 期刊论文

  • 发表日期：

    2026年7月13日

查看全文
  • S. Sahunthala,
  • K. D. V. Prasad,
  • A. Abinaya,
  • Adlin Sheeba,
  • S. Dinakar Raj,
  • P. Neelaveni

摘要

Underground cable networks are essential in modern urban power distribution systems due to reliable electricity delivery in densely populated regions. The increasing penetration of renewable energy resources and fluctuating load demand is creating dynamic operational conditions that tend to influence the thermal behaviour of the underground cables. The conventional static thermal rating methods tend to rely on conservative environmental assumptions, thereby limiting the actual current-carrying capacity of cable systems and reducing network efficiency. The primary challenge lies in accurately predicting the thermal state of the underground cables under varying load currents, soil temperature, moisture conditions, and environmental uncertainties while simultaneously supporting intelligent operational decisions for power distribution networks. Conventional machine learning models provide accurate predictions, whereas fuzzy systems represent uncertainty, but they often lack adaptive learning capabilities. To address this problem, this study proposes a novel Adaptive Neuro-Fuzzy Thermal Decision Optimisation (ANFTDO) algorithm for intelligent underground cable management. The proposed method combines the adaptive neuro-fuzzy inference engine to model the nonlinear thermal interactions within underground cable systems. A deep learning–assisted prediction module estimates cable temperature dynamics, while a reinforcement-based optimisation layer continuously adjusts fuzzy membership functions based on historical operational data. The model further adopts a dynamic decision module that evaluates thermal risk levels and recommends an optimal load-allocation method for the cable network. Results are showing that the proposed ANFTDO model is achieving a MAE of 0.94 °C, RMSE of the 1.39 °C, prediction accuracy of the 96.9%, thermal risk detection rate of the 95.6%, and cable utilisation improvement of the 90.9%, which is performing better than the conventional AI- and fuzzy-based methods in both prediction reliability and operational efficiency.