The present application provides a modeling method and apparatus for monitoring the operating state of a power secondary system, the method including extracting first training data from pre-obtained training data set, calculating Euclidean distances of first training vectors corresponding to the first training data and weight vectors of each initial neuron in preset initial neural network model respectively, acquiring a first initial neuron corresponding to a minimum Euclidean distance, and updating a weight vector of each initial neuron adjacent to the first initial neuron based on the firsttraining vector; if the first training data is not the last training data of the training data set, extracting a second training vector corresponding to the second training data until all the trainingdata of the training data set are extracted to obtain a training neuron model; calculating the adjacent region value of each neuron in the training neuron model; the neuron being tagged based on thecalculated adjacent region value and the preset adjacent threshold. The method and device can effectively monitor the running state of the secondary power system.