To ensure core power grid data quality, this system proposes a four-dimensional technical system: 1) Data feature classification (volatility/trend/variability 11-dimensional features) and five verification scenario constructions; 2) PCA-LSTM-based time-series anomaly detection and prediction; 3) Unsupervised meta-learning technology for intelligent recommendation of detection methods; 4) Spark MLlib distributed anomaly detection framework. Through multidimensional feature analysis, intelligent algorithm fusion and distributed computing, it enables precise quality control of massive power data.

Technology provider:Nanjing University of Posts and Telecommunications