Research On Prony For Zero-crossing Prediction Of Controlled Fault Interruption

Nowadays, the power system is developing towards higher voltage, larger capacity and more networks, along with the increase of short fault currents. However, the interrupt capacity of produced high voltage circuit breakers cannot satisfy the cut-off demands of the growing system. Even though measures, such as changing the prototype of the grid, inserting the current limiting reactors and so on have been proposed, the delivery capacity and the reliability of the power system will be affected. However, Controlled Fault Interruption(CFI) not only mitigates the erosion of the contacts, spans the circuit breakers' electric life, but also enhances their existing breaking level, making the CFI being one of the focuses for the intelligent electrical apparatus.Fast and exact zero-crossing prediction is the foundation of the CFI technique. This paper compared the existing algorithms and proposed to use the Prony method. Based on a deep research, the inherent defects of Prony method were found and a new algorithm named Prony-like method was offered. Besides, By means of process optimization, the Prony-like method becomes more stable. In the end, the new algorithm was tested with set signals in Matlab and relationship between sampling frequency, data length and algorithm precision was analyzed, finally forming a comprehensive comparison result among the five mentioned prediction algorithms in this paper.Effects of Prony and Prony-like were validated by recorded data from a low-voltage simulation. Meanwhile, the error distribution of the new algorithm was formed in statistics with multiple recorded data processed. After that, this paper also inspected the Prony-like algorithm with short fault data recorded in the practical power system.Both results show that the Prony-like is accurate with very samll error margin for zero-crossing prediction.Finally, this paper also designed a controller used for the synchronous switch and strategies for the conventiaonal load types and the short fault were integrated to make it more flexible. Besides, the processing time of the Prony-like algorithm was checked in the kernel of the controller. It shows that the algorithm finished prediction in 1ms with 1kHz sampling frequency. Considering the property of the selected chip, the proposed Prony-like algorithm was also verified with fault signals in the DSP.