Risk situation awareness of urban resilient distribution network cyber-physical system based on data density and improved N-HiTS

The urban resilient distribution network is a crucial component of a resilient power network, facing various risks such as operational failures, climate impacts, and network attacks. These risks pose significant threats to the security of the power networks. However, the existing methods still have some problems, such as incomplete risk awareness indicators and inaccurate identification of power grid operational state. To ensure the secure and stable operation of an urban power grid, a risk situational awareness model for the resilient distribution network cyber-physical system (CPS) based on data density is developed, along with an enhanced Neural Hierarchical Interpolation for Time Series Forecasting (N-HiTS). Firstly, a more comprehensive risk situation index system is proposed to reflect the characteristics of power grid security situation and provide assessment basis for situation understanding and prediction. Second, the data density is used to determine the typical state of each node in order to distinguish between risk and normal states. Thirdly, a neural hierarchical interpolation algorithm with enhanced predictive capabilities is presented for forecasting future operational characteristics. Finally, a neural random network is used to classify the future states of nodes, and the part that does not belong to the typical state is identified as the risk state, so as to realize the risk situational awareness. The model is validated using actual data from a regional distribution network in East China. The results indicate that the proposed method achieves an identification rate exceeding 95% for all types of risks, and both the index system and the model can improve the situational awareness accuracy of each node by about 2%. Therefore, the proposed method can perceive the future risk posture of CPS nodes in the distribution network and have higher accuracy compared to traditional methods.