With the expansion and complexity of power system, ensuring the safe and stable operation of power backbone network becomes the key. Focusing on the multi-source data of power backbone network, this paper is committed to realizing accurate adaptive risk pre-alarm and anomaly detection. By deeply analyzing the characteristics of multi-source data in power backbone network, an algorithm model integrating machine learning and deep learning is constructed. Among them, the risk pre-alarm model combines random forest (RF) and long-term and short-term memory network (LSTM), and the anomaly detection model adopts convolutional neural network (CNN), and uses principal component analysis, adaptive learning rate adjustment and other strategies to optimize performance. The model is tested under the simulated actual operation environment. In the risk pre-alarm experiment, the false positive rate and the false negative rate of this algorithm are 10% and 5% in the gradual change scene from light load to heavy load, which are significantly lower than the traditional single data source algorithm. In the experiment of anomaly detection, the detection accuracy reaches 85%, far exceeding the traditional algorithm's 60%. The results show that the algorithm model has superior performance in processing multi-source data of power backbone network, which can effectively improve the accuracy of risk pre-alarm and anomaly detection, and provide strong support for the safe operation and maintenance of power backbone network.
Zhang et al. (Sun,) studied this question.