Abstract: For most of the magnetic field probes were calibrated in the frequency-domain and could not satisfy the measurement requirement of electromagnetic transient, this paper proposes a time-domain calibration method based on Elman neural network. The method uses Helmholtz coils, surge generators, oscilloscopes and other instruments to build a time-domain calibration platform, and collects the output current of surge generator and the induced voltage of magnetic field probe. The induced voltage and the magnetic induction intensity are taken as the input and output data, then the Elman neural network model is built. The results show that, the model can accurately predict the change tendency of the magnetic induction intensity and the calibration model is effective.