Optimized Fault Prediction in Power Distribution Transformers Using Grey Wolf Optimizer-Based SVM and MLP Models
DOI:
https://doi.org/10.30865/jurikom.v13i1.9435Keywords:
Distribution Transformer, Fault Prediction, Grey Wolf Optimizer, Multilayer Perceptron, Support Vector MachineAbstract
Distribution transformers are critical components of power distribution systems, and their reliability directly affects the continuity and quality of electrical energy supply. However, early-stage transformer faults are difficult to detect because their operational characteristics often closely resemble normal operating conditions, which can lead to undetected degradation and unexpected failures. This study aims to improve the accuracy and robustness of fault prediction in distribution transformers by proposing a hybrid approach that integrates the Grey Wolf Optimizer (GWO) with Support Vector Machine (SVM) and Multilayer Perceptron (MLP) models. The main contribution of this research is a direct and systematic performance comparison between baseline machine learning models and their GWO-optimized counterparts, highlighting the effectiveness of metaheuristic optimization in enhancing classification performance. GWO is employed to optimize key model parameters, enabling improved convergence behavior, higher classification accuracy, and better generalization capability. The proposed models are evaluated under four transformer operating conditions, namely Light Load Imbalance, Light Overload, Normal, and Normal High Temperature, which represent practical scenarios in power distribution networks. Model performance is assessed using standard classification metrics, including Accuracy, Precision, Recall, and F1-Score. Experimental results show that the baseline SVM achieved an accuracy of 68%, while the baseline MLP reached 87% accuracy. After GWO-based optimization, the SVM–GWO model demonstrated a significant improvement, achieving 92% accuracy, whereas the MLP–GWO model produced the best overall performance, achieving 93% accuracy, precision, recall, and F1-score. These findings confirm that GWO-based optimization substantially enhances transformer fault prediction performance and demonstrates the strong potential of the proposed hybrid models for real-time monitoring and preventive maintenance of power distribution transformers.
References
[1] R. Shintabella, C. E. Widodo, and A. Wibowo, “A Hybrid Machine Learning-Genetic Algorithm (ML-GA) to Predict Power Tranformer Life,” International Journal of Health, Education &Social (IJHES), vol. 6, no. 12, p. 12, 2023, [Online]. Available: www.ijhes.com
[2] V. Rokani, S. D. Kaminaris, P. Karaisas, and D. Kaminaris, “Power Transformer Fault Diagnosis Using Neural Network Optimization Techniques,” Mathematics, vol. 11, no. 22, Nov. 2023, doi: 10.3390/math11224693.
[3] G. Santamaria-Bonfil, G. Arroyo-Figueroa, M. A. Zuniga-Garcia, C. G. Azcarraga Ramos, and A. Bassam, “Power Transformer Fault Detection: A Comparison of Standard Machine Learning and autoML Approaches,” Energies (Basel)., vol. 17, no. 1, Jan. 2024, doi: 10.3390/en17010077.
[4] A. Esmaeili Nezhad and M. H. Samimi, “A review of the applications of machine learning in the condition monitoring of transformers,” Feb. 01, 2024, Springer Science and Business Media Deutschland GmbH. doi: 10.1007/s12667-022-00532-5.
[5] A. Esmaeili Nezhad and M. H. Samimi, “A review of the applications of machine learning in the condition monitoring of transformers,” Feb. 01, 2024, Springer Science and Business Media Deutschland GmbH. doi: 10.1007/s12667-022-00532-5.
