Application of Artificial Neural Network in TTU Frequency Analysis Module

In order to estimate the frequency of voltage and current signal in transformer terminal unit effectively, this paper applies artificial neural network (ANN) to estimate the signal frequency. The principle of ANN method is introduced, the structure, input and output of the neural network for frequency estimation is determined, and a computer program is written to implement ANN. The results show that the frequency estimation error is only about 0.1 mHz. On this basis, the influence of number of hidden layers, number of hidden layer neurons and goal on the accuracy of frequency estimation is investigated. The results show that the number of hidden layers have little effect on the frequency estimation error. With the increase of the number of hidden layer neurons, the frequency estimation error decreases gradually. The error remains pretty much the same when the number of hidden layer neurons larger than fifty. Therefore, it is suggested to be set to fifty. The frequency estimation error increases with goal. Taking into account all of these factors, the mean squared error is suggested to be set to 10
⁻⁸
HZ
²
.