An anomaly detection method for daily line loss rate based on deep autoencoder

Collection and analysis of line loss data has a great significance for the smooth operation of power grid system. So, this paper apply the deep autoencoder to identify the outliers of line loss data. The deep autoencoder is constituted by a restricted Boltzmann machine, which can reconstruct original data. Firstly, the normal line loss rate sample is trained the deep autoencoder. Next, paper calculates the error between original value and reconstructed value output by the deep autoencoder. Finally, by setting the threshold, the data points with an error greater than the threshold are regarded as abnormal line loss rate data. Through experiments on real line loss rate data sets, the results reveal the anomaly detection method based on deep autoencoder is effective.