Abstract: The information network of the power system is an important part of the long-term continuous and effective operation in power industry. The complex network structure between power network and information network in the smart grid brings great challenges to the anomaly detection on network flow in information communication network security. Traditional machine learning algorithms and newly developing deep learning algorithms often have shortcomings such as low detection accuracy and poor real-time performance in solving the problem of network flow anomaly detection, while the network anomaly detection process that combines breadth learning and control chart has the advantages of faster training speed and more accurate detecting results. These advantages can meet the needs of anomaly detection requirement in smart grid to a certain extent, thereby achieving the purpose of improving the security of information network.

Key words: breadth learning, anomaly detection of network flow, artificial neural network, normal behavior model, control chart, smart grid