A condition evaluation ensemble for power metering HPLC units within complex data scenarios

Abstract

Abstract The High-Speed Power Line Carrier Communication (HPLC) enables the connections among power metering devices in integrated energy systems, and thus their satisfying operations are indispensable for system reliabilities. In order to more precisely diagnose their conditions especially in real complex data scenes, a multi-model evaluation ensemble is proposed in this paper. Firstly, typical IoT application contexts of customer-side metering equipment are analyzed, thus the corresponding main impact factors along with their performance evaluation indices can be probed. Next, to handle the multi-source, heterogeneous, high-dimensional datasets during applications, the Kernel Independent Component Analysis (KICA) is established to diminish data dimensionalities, thus the individual weights of each index can be rated. On the other hand, the Component Importance Measure (CIM) model is built to differentiate the impact degree of each indicator on the overall IoT connection performance, where the influence of dissimilar index on the entire performance, rather than the proportion or frequency, will be directly assessed to determine their impact weights. Ergo, a comprehensive diagnosis can be achieved via these two-fold total weights accordingly. Finally, the feasibility and effectiveness of the proposed method can be verified by an empirical case study, which is conducive to further improving the accuracy and rationality of HPLC condition evaluations.