Distribution network voltage transformer abnormal state evaluation method based on CNN-LSTM-RBF network

The invention relates to a CNN-LSTM-RBF network-based distribution network voltage transformer abnormal state evaluation method, and the method comprises the following specific steps: collecting parameters of a transformer in a normal state, and constructing an input matrix X; processing the input matrix X based on the correlation difference between the actually measured transformation ratio and other parameters to obtain a difference matrix psi, and adopting a CNN-LSTM training model to obtain a transformation ratio evaluation value K '; and learning by adopting an RBF network model to obtain a transformation ratio evaluation value # imgabs0 #, and estimating the transformation ratio coefficient of each transformer based on the operation data of the current moment. Based on the primary voltage deduced by the transformer and the primary voltage deduced by the mutual inductor, evaluating the state of the mutual inductor; and calculating the error of the mutual inductor. According to the invention, the CNN-LSTM