Data-driven fault diagnosis of power converters

Abstract

Development of machine learning algorithms for multi-classification makes many unsolved classification problems approachable. The only way to find the best classifier is not just to get the best accuracy level by using it. So, we considered some more statistical measures to prove the efficiency of our model. Considering the wide use of neural network architectures in different fields we have selected one of the emerging architectures, Stochastic Configuration Network (SCN). It not just influentially impacts the accuracy and outperforms compared with some traditional algorithms in term of time complexity. In our work of classifying converter open circuit faults where we deal with 22 types of fault classes, SCN shows an efficient result to be used with. In this final year project, MATLAB-based programming will be used to conduct the system models and simulate the test results.