Subharmonics Determination Method based on Binary Successive Approximation Feed Forward Artificial Neural Network: a first improvement

An improvement of a feedforward artificial neural network for individualizing electric network subharmonics frequencies has been realized increasing the neural network resolution. Whishing to have a quick system able to signal the presence of subharmonics in real time, It has been fixed a sampling window of 20 ms equal to a period of an ideal fundamental frequency of 50 Hz. This choice obliges to pick up only a fraction of the subharmonics period that is not ever completed inside the sampling window making the classical spectral frequency recognition method unsuited. This obliges to transfer the problem of the their recognition on the neural network trained phase obtained by an appropriate discretization of the training set. The binary successive approximation structure has the merit of the scalability exploited in this article to increase the resolution than previous work. The net has been widely tested on synthesized signals.