Scholarlink

<p>This research introduces a new methodology incorporating the Gramian Angular Difference Field (GADF) approach with an adaptive particle swarm optimised convolutional neural network (APSO-CNN) model in order to classify various types of faults occurring in power system. In this work, firstly, voltage time series signal is converted into a two-dimensional image file using GADF. Subsequently, CNN is applied to extract features and classify the images. Recently, CNNs have demonstrated outstanding performance on a variety of image classification tasks. However, the architectures of CNNs have a significant impact on their performance. The architectures of the most advanced CNNs are frequently hand-crafted by experts in both CNNs and the problems under investigation. Therefore, in order to successfully handle the fault classification problems, this research introduces an optimized CNN framework that determine the architecture and hyperparameters simultaneously using the adaptive particle swarm optimization (APSO) algorithm. The obtained results are compared with optimally tuned architectures acquired through Genetic algorithm optimized CNN and the Bayesian optimization-based built-in KerasTuner tool. The acquired observations clearly show a considerable reduction in the computational workload of the improved CNN architectures using the APSO algorithm. The outcomes demonstrated competitive results compared to other cutting-edge approaches, with the proposed approach achieving favourable results with an accuracy higher than 99%.</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

Imaging Time Series Technique with an Adaptive Particle Swarm Optimized CNN Model for Classification of Faults in Transmission Line

  • 期刊论文

  • 发表日期：

    2025年1月18日

查看全文
  • Jyoti Shukla,
  • Basanta K. Panigrahi,
  • Ramesh C. Bansal

摘要

This research introduces a new methodology incorporating the Gramian Angular Difference Field (GADF) approach with an adaptive particle swarm optimised convolutional neural network (APSO-CNN) model in order to classify various types of faults occurring in power system. In this work, firstly, voltage time series signal is converted into a two-dimensional image file using GADF. Subsequently, CNN is applied to extract features and classify the images. Recently, CNNs have demonstrated outstanding performance on a variety of image classification tasks. However, the architectures of CNNs have a significant impact on their performance. The architectures of the most advanced CNNs are frequently hand-crafted by experts in both CNNs and the problems under investigation. Therefore, in order to successfully handle the fault classification problems, this research introduces an optimized CNN framework that determine the architecture and hyperparameters simultaneously using the adaptive particle swarm optimization (APSO) algorithm. The obtained results are compared with optimally tuned architectures acquired through Genetic algorithm optimized CNN and the Bayesian optimization-based built-in KerasTuner tool. The acquired observations clearly show a considerable reduction in the computational workload of the improved CNN architectures using the APSO algorithm. The outcomes demonstrated competitive results compared to other cutting-edge approaches, with the proposed approach achieving favourable results with an accuracy higher than 99%.