Di Huang, Liangrong Xu, Bin Zhou, Chaoyue Zhu, Weiyan Zheng, and Xingping Yan
AI engine; Isolated forest; Spline interpolation; Data fusion; Infor-mation entropy; Spectral clustering
This study proposes an enhanced analysis framework based on an ar- tificial intelligence engine, aiming to address the challenges of highly complex data and dynamic changes in user behavior in emergency power protection scenarios of smart grids. This method incorporates a multi-stage data processing flow that includes anomaly detection using isolated forests, data repair via cubic spline interpolation, and behavioral decoupling analysis, which combines information entropy features with spectral clustering. The experimental results showed that the proposed method in the study achieved a classification accu- racy rate of 91.4% through kernel space transformation and density peak adaptation. The case analysis revealed significant behavioral insights. Cluster I exhibited characteristic dual peaks (0:00-8:00 and 18:00-24:00 hours) as well as daytime gas slots (8:00-18:00 hours), which precisely matched the three-shift manufacturing operation. On the contrary, Cluster V showed the double peaks of the office (9:00- 11:00 and 15:00-17:00 hours), reflecting the working rhythm of the administrative agency. The research found that among the five power consumption patterns, the classification accuracy rate reached 91.4%, and the overall processing efficiency was 37.6% higher than that of the traditional method. This research provides feasible data analysis support for dynamic demand response and stable operation in smart grids.
Important Links:
Go Back
