Prediction and Correction of Missing Power Data Based on Combined CNN and LSTM

Data serves as the critical foundation for the stable operation of power grid dispatching control systems. Data loss during collection due to hardware failures compromises system data integrity, thereby affecting the intelligence and efficiency of grid dispatching. Accurate prediction of missing data is therefore essential for building smart grid dispatching systems. This paper improves existing CNN-LSTM joint prediction methods for addressing missing data in power grid energy collection systems by incorporating a correction model. Leveraging CNN convolutional neural networks and specialized opposite-side data scenario modeling in power data for different missing segments, experimental results demonstrate this method reduces mean absolute error to 0.142, enhancing prediction accuracy while ensuring data integrity and precision for intelligent grid dispatching systems.