Dissolved Gas Analysis Interpretation and Intelligent Machine Learning Techniques

This chapter provides a comprehensive study on intelligent Dissolved Gas Analysis (DGA) interpretation techniques and their applications in power transformer condition assessment. It summarizes some major work and contributions from CIGRE in dissolved gas measurement and its interpretation. The chapter discusses the experience of using propene and propane as a new tool for DGA interpretation and identifies the drawbacks and uncertainties of the conventional DGA interpretations. It proposes a general framework for applying intelligent pattern recognition techniques to transformer insulation diagnosis and provides a literature review on various pattern recognition algorithms for transformer insulation diagnosis. To overcome the drawbacks of the conventional pattern recognition techniques for health assessment of power transformers, a hybrid SMOTEBoost algorithm is introduced. This hybrid algorithm can facilitate pattern recognition algorithms, achieving desirable diagnosis accuracy and generalization capability. The chapter presents a case study to illustrate the procedure of applying pattern recognition techniques and the SMOTEBoost algorithm to practical transformer insulation diagnosis.