Generative Design of 110 kV Condenser Bushing Using Artificial Neural Networks

Abstract

Bushings are an essential component of high-voltage equipment. Particularly in systems rated above 66 kV, condenser bushings are preferred due to their reduced weight and high electric field utilization factor. Traditionally, optimal design parameters for a condenser bushing at a given voltage rating are the number of foils, foil spacing, and length of foils and are arrived at using analytical methods to ensure uniform field distribution in the bushing. Alternately, numerical tools like the finite element analysis (FEA) and other techniques might be used to generate and validate new designs subject to various design and manufacturing criteria. The next step in the process involves validating the design through FEA and tuning the design to meet certain specified design and manufacturing criteria. Recent advancements in machine learning and artificial intelligence have led to the development of models employing neural network architectures for the generative design of complex mechanical, thermal, and electrical systems. In this work, we propose an artificial neural network (ANN) model capable of generating design parameters for a condenser bushing rated at 110 kV. The ANN model uses a dataset of previously simulated bushing designs to learn the dependencies between design parameters and their impact on the performance of the bushing. The field utilization factor is taken as the performance marker. The model is then trained using this data and can generate optimized design parameters for achieving the best field utilization factor for the bushing of the same rating. Finally, the design generated by the ANN model is validated using FEA.