Abnormal behaviour recognition technology in smart grid monitoring based on artificial intelligence

Abstract This paper aims to solve the problem of insufficient accuracy of abnormal behaviour recognition in smart grid monitoring. By applying the CNN (Convolutional Neural Network)- BiLSTM (Bidirectional Long Short-Term Memory) model, combined with spatial feature extraction and time series analysis, the detection ability of complex abnormal patterns is improved. The safety and stability of power grid operation are guaranteed, and efficient management and accurate fault diagnosis of smart grid are realized. Multidimensional time series data, including voltage, current, power factor, and other information, are obtained from sensors and equipment of smart grids, and data cleaning, abnormal annotation, and standardization are performed. The sliding window method is used to divide the time series data to adapt to the input of the deep learning model. The model structure includes a CNN layer for extracting local features, a BiLSTM layer for capturing time series dependencies, and finally a fully connected layer for abnormal behaviour classification. In the test set, the average accuracy of epochs within 30–50 times reaches 97.6 %. Experimental findings demonstrate that the accuracy, precision, and recall of the CNN-BiLSTM model on the test set are better than those of the traditional CNN-LSTM, BiLSTM, and Transformer models. The CNN-BiLSTM model can effectively enhance the accuracy of abnormal behaviour detection in smart grids and provide a reliable solution for the safety monitoring of power grids.