An Investigation on the Selection of Filter Topologies for Passive Filter Applications by Neural Network

Abstract

This project presents an alternative way in designing 5th, 7th and 11th harmonic filters by creating a neural network program that can calculate harmonics which are generated in the system by ac or dc drives and it can verify the result using the simulation in verification program. Our program can choose the right harmonic filter and help designing harmonic filter faster, easier and more accurate by concerning the cause of other factors that exclude the formulas such as limitation of the devices in a suitable cost. It will be more and more accurate every time the users use this program. Our program will train itself after inputting new data. The more we use the program, the more accurate it will be. We created the program with easy understandable user interface to facilitate any harmonic filter designers with low experiences to have an ability as an expert harmonic filter designer because our program has a memory of an experienced designer and ability to train itself. Furthermore, our program will help supporting the development in harmonic filter designing for any kinds of future research.