Design Optimization of Insulation Structure for HVDC converter transformer using Genetic Algorithm with Reliability Analysis

Abstract

This thesis presents a design optimization strategy for insulation structure of High Voltage Direct Current (HVDC) converter transformer. Recently, HVDC power transfer system is drawing attention because of its efficient bulk power capability. HVDC power transfer system is typically composed of a transformer, an AC-DC converter, and transmission lines. Because of the AC-DC converter connected to the transformer, electric stresses such as AC, DC, and DC polarity reversal (DCPR) are excited onto the transformer, whereas only AC stress is excited on the conventional AC power transmission systems. Since DC and DCPR stresses cause significantly different electric field distributions in the insulations, transformer insulations designed for AC systems do not guarantee safety under those stresses against electrical breakdown. Therefore, when designing insulation for HVDC converter transformer, each insulation material, transformer-oil and barriers, should be differently placed compared to AC transformers. This problem of placing insulation material can be solved by identifying design variables of transformer insulation and optimizing them with an objective function that is constructed to improve the robustness of transformer insulation. This thesis focuses on the design optimization of two types of insulations, parallel oil-barrier insulation between conductors and end-winding insulation, because electric fields are stronger in those areas than other areas. The proposed algorithm can optimize the two insulations either separately or at the same time in a sequential way. Because no computer-aided design optimization algorithm has been suggested until now, the algorithm is expected to save a considerable amount of effort and time comparing to the approaches depending on experimental data and human experience. Genetic Algorithm is used for the optimization, and the analyses of each candidate solution structure are performed with Finite Element Analysis. Reliability analysis is included in the optimization process to consider the unexpected insulation failure due to positional deviations of barriers, which can be caused by alternating electric field and manufacturing inaccuracy. In the end of this thesis, the resultant insulation structures are compared with reference structures which are provided from a manufacturer. The suggested algorithm produces parallel oil-barrier insulations having 5% less insulation lengths than reference insulations on average under both AC and HVDC stresses, while maintaining their robustness, which means the achievement of the design objective. If the reliability analysis is incorporated into the algorithm, 4.6% increase in the insulation length is observed.