Study on the Adaptability of Magnetization Characteristic Model for Magnetically Saturated Controllable Reactor

The fundamental problem in accurately analyzing and calculating the performance and parameters of controllable reactors lies in establishing an accurate and practical magnetization characteristic model for the core material of controllable reactors. In this paper, the Magnetically-Saturated Controllable Reactor (MSCR) is taken as the analysis object, the magnetization characteristic model with different complexity of controllable reactor is established. According to the experimental measurement curve of the ferromagnetic material of the MSCR experimental prototype, the fitted magnetization curves under different magnetization characteristic models are obtained, and the fitting accuracy is compared and analyzed. Based on the nonlinear dynamic electromagnetic network model of MSCR, the working current of MSCR with different complexity magnetization characteristic models under different working conditions is calculated and compared with the experimental results. Finally, from the aspects of calculation accuracy and calculation speed, the relationship between the calculation accuracy and calculation speed of the electromagnetic parameters of the controllable reactor and the variation of the magnetization characteristic models with different complexities is revealed. And the selection criteria for controllable reactor magnetization characteristic model is proposed. The results show that both the single-valued magnetization characteristic model and the multi-valued magnetization characteristic model can be applied to the calculation of electromagnetic parameters of MSCR, and the multi-valued magnetization characteristic model is preferred in the no-load analysis. From the perspective of calculation accuracy, the neural network hysteresis model is the most superior, and the calculation accuracy of MSCR working current is positively correlated with the fitting accuracy of the magnetization characteristic model. From the perspective of calculation speed, the function fitting model has the fastest calculation speed. In general, the piecewise linear model and the neural network hysteresis model are more superior.