Dry-type air-core reactor with high voltage and large capacity is one of the important electrical devices in large scale power system. Inter-turn short circuit fault is the main damage form of the reactor. Both online monitoring of reactor and early warning of inter-turn short circuit fault are of great significance for the power system reliability. The inductive voltages of equipotential self-difference coils installed at the two ends of the reactor are used to monitor the variation of the magnetic field of the reactor. Both high-speed sampling of digital signal processor（DSP） with self-power supply and ZigBee wireless communication are adopted to insulate high voltage. The magnetic field of the reactor will suddenly change when inter-turn short circuit occurs

therefore

a new algorithm for the mean value of moving center of sampled data is proposed to fit the working waveform of the running reactor

in which the deviation between the sampled data and the working waveform is regarded as a stochastic process

and the t-test model for confidence interval is used to recognize the inter-turn short circuit fault signals. The test results show that the proposed methods are accurate and reliable to detect the early inter-turn short circuit fault in the range of 5%～100% load variation

and they can be suitable for online monitoring of dry-type air core reactor with high voltage and large capacity.