A novel iron loss reduction technique for distribution transformersbased on a combined genetic algorithm - neural network approach

摘要

The paper presents an effective method to reduce the iron lossesnof wound core distribution transformers based on a combined neuralnnetwork/genetic algorithm approach. The originality of the worknpresented is that it tackles the iron loss reduction problem during thentransformer production phase, while previous works concentrated on thendesign phase. More specifically, neural networks effectively usenmeasurements taken at the first stages of core construction in order tonpredict the iron losses of the assembled transformers, while geneticnalgorithms are used to improve the grouping process of the individualncores by reducing iron losses of assembled transformers. The proposednmethod has been tested on a transformer manufacturing industry. Thenresults demonstrate the feasibility and practicality of this approach.nSignificant reduction of transformer iron losses is observed inncomparison to the current practice leading to important economic savingsnfor the transformer manufacturer