FFNet: An automated identification framework for complex power quality disturbances

Effective identification of complex power quality (PQ) disturbances is the premise and key to improve power quality issues in the current complex power grid environment. However, the influx of nonlinear loads and impact loads makes power system disturbance signals distorted and complex, which increases the difficulty of PQ disturbances identification. To address this issue, this work presents a novel feature fusion network (FFNet) for the automated detection and classification of complex PQ disturbances. Firstly, an adaptive double-resolution Stransform (ADRST) algorithm is proposed for PQ disturbance time-frequency analysis. The adaptive window parameters optimization method is developed in ADRST to improve time-frequency resolution based on energy concentration maximization. Next, an improved convolutional neural network (CNN) method is presented for feature automatical extraction and multiple disturbances classification. Integrating ADRST and improved CNN, a classification framework called FFNet is further proposed to identify various complex PQ disturbances. Finally, experiment cases based on simulation and experimental PQ signals are conducted, where the results demonstrate that the classification accuracy of proposed framework can reach 99.47% even in 20 dB noise level.

Automated summary

This study proposes a novel feature fusion network (FFNet) for the automated detection and classification of complex power quality (PQ) disturbances. By using an adaptive double-resolution Stransform (ADRST) algorithm for time-frequency analysis and an improved convolutional neural network (CNN) for feature extraction and disturbances classification, the accuracy of PQ disturbances identification is improved.