Balance Risk Assessment Method of Power System Based on CNN-BiLSTM-Attention

In view of the dual balance risks of power supply protection and consumption protection, this paper constructs the balance risk index system of power system from the two aspects of power supply protection and consumption protection. Then, on this basis, the balance risk classification standard of power system is constructed, and the balance risk levels of power system are divided. A multi-time scale evaluation method for power system balance risk based on CNN-BiLSTM-Attention is proposed. In the offline training session, the deep neural network structure was constructed based on convolutional neural network and bidirectional long short-term memory network to extract multi-spatial-temporal scale features, and the attention attention mechanism was introduced to improve the model’s ability to represent key information. In online application, transfer learning method is used to update the offline CNN-BiLSTM-Attention model under the condition of small samples in the target domain, so as to make up for the defect that the risk data characteristics cannot be well captured due to the lack of samples. Thus, the evaluation accuracy and efficiency of CNN-BiLSTM-Attention model under the new operation point can be effectively improved to realize the rapid and efficient evaluation of power system balance risk. Experiments are carried out on IEEE 30-nodes systems to verify the effectiveness of the proposed method.