Active power line conditioner with neural network control.

摘要

Harmonics in a power system can lead to communication interference, transformer heating, or malfunction of solid-state devices. Active power line conditioners (APLC) are one important way to achieve harmonic reduction. The purpose of this dissertation is to propose a novel voltage-type APLC which cancels harmonic currents by injecting a compensation current. The proposed APLC consists of a variable dc voltage source, an inverter, and a neural network controller that is trained with the genetic algorithm and back-propagation. Computer simulations for two load current test cases show that the neural network can provide switch control signals for the proposed APLC to generate compensation currents that reduce line current THDs significantly. Measurements on a simple, not yet optimized hardware prototype of the proposed APLC are also very encouraging. These results and the simplicity of the system suggest that the proposed APLC and its unique neural network-genetic algorithm control system constitute a viable solution alternative for line current harmonics.