Power factor correction: performance comparison of an existing microcontroller-based system and a neuro-fuzzy system

Abstract

An existing microcontroller-based power factor correction system has been able to improve the overall conversion of electrical power into a useful work of a highly industrial load. However, more improvements are still desired to get the existing power factor value close to 1 as much as practically possible. With the current microcontroller-based power factor correction system, microcontroller has to be replaced often due to power fluctuation and a low-quality power available. The microcontroller requires ordering for new replacement as it is not reprogrammable to meet the new operational demands. Artificial intelligence tools, neural network and fuzzy logic are considered. Neuro-fuzzy system approach is settled for as an alternative to microcontroller-based power factor correction system. Neuro-fuzzy system is able to learn through training, testing, and validation processes and controls the automatic switching of the capacitor banks to adequately compensate for the lagging loads. Results obtained were compared to the existing microcontroller power factor correction system. Neuro-fuzzy system shows better performance over microcontroller-based system. The neuro-fuzzy system automatically adjusts itself to suit the present operational requirement to always have a power factor result closer to 1 as compared with that of a microcontroller-based system which does not give room for reprogramming making it static to a larger extent in its operational duties.