A machine learning method for locating subsynchronous oscillation source of VSCs in wind farm induced by open-loop modal resonance based on measurement

In recent years, sub-synchronous oscillation incidents have been reported to happen globally, which seriously threatens the safe and stable operation of the power system. It is difficult to locate the oscillation source in practice using the parameterized model of open-loop modal resonance. Therefore, this paper aims at the problem of oscillation instability caused by the interaction between the multiple voltage source converters in the wind farm grid-connected system, proposes a method for locating the oscillation source of a wind farm using measurement data based on the transfer learning algorithm of transfer component analysis. At the same time, in order to solve the problem of the lack of oscillation data and the inability to label in the real system, a simplified simulation system was proposed to generate large batches of labeled training samples. Then, the common features of the samples from simulation system and the real system were learned through the transfer component analysis algorithm. Afterward, a classifier was trained to classify samples with common features. Finally, two grid-connected wind farms with VSC access are used to verify that the proposed method has good locating performance. This has important reference value for the practical application of power grid dispatching and operation using measurement to identify oscillation sources.