A NOVEL ANALYTICAL EXPRESSIONS MODEL FOR CORONA CURRENTS BASED ON CURVE FITTING METHOD USING ARTIFICIAL NEURAL NETWORK

Abstract

The analytical expressions for corona discharge currents are usually represented by the mathematic models based on curve fitting method. For the complex mechanisms, none of these currently models can describe a measured corona current with arbitrary waveforms. A novel curve fitting method using BP neural network (BPNN) technique is applied to describe the mathematic model of the corona currents in time domain. The analytical expressions for the currents can be established via extracting the weights and thresholds parameters of the trained BPNN. The expressions all have the same structure which has only four types of parameters, and the structure is independent of the corona current waveforms. The curve fitting for the measured corona currents with arbitrary waveforms by different models was carried out, and the results were analyzed, which indicate that the BPNN method performs best. Compared with the current expressions fitted by the double exponential function and Gaussian function, the expressions by BPNN can fit the current waveforms with the lowest mean square error (MSE) in time domain and the highest accuracy to spectra of the currents in frequency domain. The proposed method is suitable for establishing a unified analytical expressions model for corona currents with arbitrary shapes.