Distinguishing method for zoning frequency stability of electric power system

The invention discloses a power system partition frequency stability judgment method, which comprises the following steps of: performing simulation operation on an interconnected power system under different load conditions by using power system simulation software to obtain frequencies of each node in different operation time, and taking the frequencies as a training data set of a convolutional neural network; constructing indexes based on Grubrum angle field imaging; feature extraction is carried out on the constructed Gramer angle field, and a convolutional neural network is trained; determining a weak point of frequency instability based on the dynamic partition of the dynamic time warping distance; after time sequence data accuracy and safety evaluation is carried out on the data collected at the frequency instability weak point, a corresponding Gramer angle field is established, and the data is converted to a polar coordinate system; inputting the converted data into the trained convolutional neural net