An Integrated Method for Electric Power Data Cleaning and Anomaly Detection Based on Multidimensional Semantic Similarity and Deep Auto-Encoding

This paper proposes an integrated method for electric power data cleaning and anomaly detection based on multidimensional semantic similarity and deep auto-encoding. The background of the study is the increasing complexity and size of electric power data, which pose challenges for data cleaning and anomaly detection. The purpose of this study is to develop an effective method to address these challenges. The method consists of two main steps: multidimensional semantic similarity calculation and deep auto-encoding-based anomaly detection. Firstly, multidimensional semantic similarity is calculated to measure the similarity between different data points and detect potential anomalies. Then, deep auto-encoding is applied to learn the high-level representations of the power data and detect anomalies based on the reconstruction error. The experimental results demonstrate that the proposed method achieves high accuracy in data cleaning and anomaly detection for electric power data. In conclusion, the integrated method based on multidimensional semantic similarity and deep auto-encoding is effective in electric power data cleaning and anomaly detection, providing a useful tool for im-proving the reliability and quality of power grid operations.