A Deep Learning through DBN Enabled Transmission Line Fault Transient Classification Framework for Multimachine Microgrid Systems

The reliable operation of power systems becomes a formidable job these days due to the high amount of complexities in the expanded power system networks. The power system networks often comprise microgrids that encounter over 80% of faults due to their exposure to unpredictable weather conditions, which reduce the insulation strength of the conductors which damaged the distribution system. Therefore, detection, classification, and location of such faults in the distributing systems are a must to ensure the flawless operation of the power systems. Machine learning methodologies are getting more attention to detect these types of faults due to their capability to handle complex fault information. Nevertheless, the obstacles are even now on the board as the traditional machine learning techniques rely on oversimplified frameworks that are incapable of analyzing a wide range of latent and explicit parameters and are also time-consuming. In this work, a unique defect diagnosis technique based on a multiblock deep belief network (DBN) and the fundamental discrete wavelet transform (DWT) is proposed, allowing the architecture to identify the deterministic reconstructing throughout its inputs. This method enables a strong multilevel generative network to utilize fault-related properties, decodes high variability functionalities, and requires minimal previous knowledge. The suggested approach is validated using a wide set of input data at various sampling frequencies. To evaluate the efficiency of DBN, a benchmark methodology based on the International Electrotechnical Commission (IEC) standard was used. White Gaussian Noise (WGN) was also implemented to test the envisioned network’s resilience. The findings show that the approach is capable of executing exact diagnosis procedures.