Levenberg-Marquardt based MLP for Detection and Classification of Power Quality Disturbances
Serge Raoul Dzondé Naoussi1, Jean Paul Ngon2
1Serge Raoul Dzonde Naoussi, Laboratory of Technology and Applied Sciences, University of Douala, Douala, Cameroon.
2Jean Paul Ngon, Laboratoty of Electronics, Power Engineering, Automation and Telecommunication, University of Douala, Douala, Cameroon.
Manuscript received on 05 March 2019 | Revised Manuscript received on 11 March 2019 | Manuscript published on 30 July 2019 | PP: 155-161 | Volume-8 Issue-2, July 2019 | Retrieval Number: A2227058119/19©BEIESP | DOI: 10.35940/ijrte.A2227.078219
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: In recent years, power quality (PQ) has become an important issue for utilities and users. In order to improve PQ, a method for detecting and classifying power quality disturbances (PQDs) is proposed. Hence in addition to identifying the disturbance signals, the proposed method is able to determine its type when occurring. This approach is based on Multilayer perceptron and Levenberg-Marquardt training rule. It is inspired by the desire to take advantage of the parallelism inherent to neural networks in view of hardware implementation using reconfigurable chips. The inputs of these networks are the samples obtained on the power grid in various conditions. The proposed method is tested for sags and swells. To classify the disturbances, the neural architectures have been generalized and configured according to the number and type of disturbances to be treated. To validate and test the proposal, a grid model was built with a three-phase fault generator under Matlab / Simulink R2017a. After comparing the results with those obtained by certain methods in the literature, the proposal proves to be an efficient and reliable tool for monitoring PQ. In fact it has the smallest mean square error and a highperformance with precision of 96%.
Index terms: Electrical Disturbances; Multilayer Perceptron; Levenberg Marquardt Algorithm, Detection and Classification.
Scope of the Article: Classification.