This article presents an algorithm for identification of anomalous magnetic field and loss in silicon steel sheet (SSS). Since the magnetic field in an SSS consists of static hysteresis, eddy current, and anomalous magnetic fields, the anomalous magnetic field is extracted from the measured hysteresis field by excluding the static hysteresis and eddy current fields. In order to predict the anomalous magnetic field for an arbitrary $B$ -waveform, an artificial neural network model with deep learning algorithm is suggested. Through comparisons with experimental measurements over a non-oriented electrical steel sheet, 35PN440, the proposed algorithm is validated.