Harmonic current model for medium-frequency furnace based on general regression neural network

A general regression neural network (GRNN) is proposed for harmonic current modeling of medium-frequency furnace (MFF) in steady-state frequency-domain. In the model, a new concept of fundamental active power load degree (FAPLD) is introduced, which is the ratio of load fundamental active power to its rated power; the nonlinear mapping between FAPLD and each order of harmonic current amplitude is established by GRNN. The interrelationships between FAPLD and current amplitude of each harmonic are discussed. GRNN has the advantages of good nonlinear mapping, small samples for modeling, few man-determined parameters, etc. Numerical results show that the proposed model has the characteristics of short training time, high precision and dynamic modeling; it is an effective method for building up harmonic current model of MFF.