Optimization of transformer winding voltage distribution algorithm under lightning impulse

Author： GOZ Electric Time：2024-08-04 09:24:12 Read：13

    At present, 35kV dry-type transformers are widely used in wind farms. The frequent operation of vacuum circuit breakers under wind farm conditions will produce ultra-fast transient overvoltages, which will cause insulation breakdown failures in a few extreme cases. Although the price of 35kV dry-type transformers is low and the failure rate is low, the cost of shutdown and replacement transportation is high after the failure, which seriously affects the economic benefits. Therefore, it is necessary to optimize the design of the winding structure. A manufacturer still had insulation failure when conducting a chopped wave impulse test on the improved transformer. To this end, for impulse signals with short time and large steepness, simulation operations need to be performed to obtain more accurate transformer winding voltage distribution results. Two problems arise in this process: First, when inputting lightning impulse chopped wave signals, large calculation errors are generated; second, when optimizing the winding structure, multiple calculations are required. When using the MTL model for calculation, there are problems of long time consumption and low model efficiency. At present, there are few studies on lightning chopped waves with complex spectrum in voltage distribution simulation research, and the methods to improve model efficiency lack pertinence. Therefore, achieving error control and improving model efficiency are of great practical significance.

    The MTL model is an effective tool for simulating and analyzing the transformer winding wave process. The solution of the voltage distribution at the beginning and end of the transformer coil based on this model can usually be divided into the following three steps:

1) Calculate the distribution parameter matrix of MTL;

2) Use FFT to decompose the input signal into a series of simple harmonics of a single frequency, and combine the MTL boundary conditions to solve the telegraph equation in the frequency domain;

3) Convert the frequency domain calculation results to the time domain, and superimpose them to obtain the voltage time domain distribution waveform of each coil turn. Among them, the setting of FFT will affect the model operation speed and accuracy, and will lead to an increase in the calculation cycle of the frequency domain solution, especially when facing large transformers, the number of turns increases, and a certain calculation accuracy is required, which will lead to problems such as slow model convergence speed and long program running time. According to actual requirements, in the process of selecting effective calculation accuracy, it is necessary to continuously adjust the calculation parameter settings, and the model itself has a long calculation time and low work efficiency.

    For this reason, taking an actual transformer as an example, the lightning impulse signal is input, and the cause of the error in the calculation process of the lightning chopping wave is analyzed. A calculation idea is proposed to effectively control the error generated in the calculation process. On this basis, the parameter calculation process in the model operation is analyzed. The fitting curve method is used to avoid the complex operation involving the parameter matrix, and the amplitude and phase of the high-frequency part are directly obtained. The model maintains a certain calculation accuracy while improving the calculation efficiency and shortening the running time, which provides convenience for the optimization design of the winding structure of 35kV dry-type transformers in wind farms.

Conclusion

   For transformers with many turns, when using the multi-conductor transmission line (MTL) model for voltage distribution calculation, there are problems such as insufficient calculation accuracy, large amount of calculation and poor calculation efficiency. By analyzing the calculation principle of the model itself, the following conclusions are obtained:

1) When the lightning impulse chopped wave is input to calculate the voltage distribution, the error generated in the calculation process can be effectively controlled by processing the input signal in steps.

2) By fitting the curve, the complicated parameter matrix cyclic integral solution can be avoided, and the calculation of 5000 frequency points can be reduced to 1200 points. While improving the calculation efficiency, it does not affect its calculation accuracy, shortening the program running speed by 74.85%, and improving the work efficiency of the model.