Intelligent neural-network-based adaptive power-line conditioner for real-time harmonics filtering

Conventional approaches for harmonic filtering usually employ either passive or active filtering techniques or a combination of both. The paper proposes an alternative intelligent adaptive power line conditioner (I-APLC), which is a form of neural-network-based adaptive harmonic filtering. The I-APLC makes use of one supervised learning rule (backpropagation) which underlies the adaptive self-learning in realising the optimal filter weight vector. Experimental results obtained via a prototype model of the DC variable-speed motor verified that I-APLC is feasible in terms of real-time tracking, adaptive harmonic filtering, faster training and convergence speeds, and simplicity in the online hardware implementation.