MIC-DNN: Data-Driven Correlation Analysis Framework for Power Data

Power systems generate large volumes of diverse data, but the most valuable information is often sparsely distributed. Effectively leveraging this valuable data has become an important challenge. Research on power data analysis methods that balance accuracy and efficiency to analyze and process massive and complex power data efficiently has become an important challenge in power data-related research. In this paper, we propose a correlation analysis framework based on the maximal information coefficient (MIC) and a deep neural network (DNN) to efficiently and accurately assess the relationships among large-scale power system data. First, we introduce the MIC method along with its underlying concepts. Then, we develop the MIC-DNN correlation analysis framework for power data and describe the design of the DNN-based feature extraction models in detail. The experimental results show that the MIC-DNN correlation analysis framework can discover the implied correlations between massive multi-source power data, thus providing valuable references for power data classification and grading, power system planning, and power market development.