This paper presents a new joint optimization method for the design of sharp linear-phase finite-impulse response (FIR) digital filters which are synthesized by using basic and multistage frequency-response-masking (FRM) techniques. The method is based on a batch back-propagation neural network algorithm with a variable learning rate mode. We propose the following two-step optimization technique in order to reduce the complexity. At the first step, an initial FRM filter is designed by alternately optimizing the subfilters. At the second step, this solution is then used as a start-up solution to further optimization. The further optimization problem is highly nonlinear with respect to the coefficients of all the subfilters. Therefore, it is decomposed into several linear neural network optimization problems. Some examples from the literature are given, and the results show that the proposed algorithm can design better FRM filters than several existing methods.

A neural network method combined with simulated annealing algorithm is proposed for power system harmonic analysis. This method is aimed at the system in which the sampling frequency cannot be locked on the actual fundamental frequency. By updating the relevant parameters including the learning rate of fundamental frequency, fundamental frequency, harmonic phases and amplitudes, the accurate harmonic estimating results can be obtained. The simulating results show that the harmonic estimation accuracy by the proposed approach is relatively better than that by the conventional harmonic analysis methods in the asynchronous case.

No abstract