Redes neurais artificiais para a estimativa do sinal de saída de amplificadores ópticos considerando sinais de entrada inclinados

Abstract

Optical Communication is the main way to supply the amount of data traffic that only grows in the last years. Optical communication systems need to be robust for dynamic operating conditions to obtain the best transmission signal. For the optical network to be adaptable it is necessary to have autonomous devices. Optical amplifiers are important devices for establishing optical communication, to make this device autonomous it is necessary to estimate how the signal is modified when passed through it. The studies were realized to estimate the signal at the output of the optical amplifier, although only flat signal characterization data were considered, which differs from what happens in a real scenario. In this work, models based on artificial neural networks are proposed to estimate the output power signal of the optical amplifier, using characterization data with non-fiat signals. The results showed that the previous model, proposed with flat signals, does not perform well with data from non-flat signals. With the results it was possible to observe that the models that use the entire signal spectrum, obtained a better result than those that consider the isolated channels. The model with the best result reached 0.15dB in the median of the medium erro distribution, a reduction of 2.14 compared to the model previously proposed. In addition, by optimizing the model with the best result, it was possible to obtain a median below 2.5dB, with amplifier cascade scenarios with more than 20 devices.