Research on Anomaly Detection Method for Meteorological Data Based on LSTM-TCN Neural Network
Y.-C. Zhang, Fei Cui, Zhuxin Xue, Guochao Fan · 2024
Meteorological data is of great significance in climate research and weather forecasting, but due to factors such as equipment failures and sensor deviations, outliers often exist, which affect the accuracy and reliability of the data. The traditional methods for quality control of meteorological data are mostly based on principles of atmospheric science and statistical laws to detect anomalies in meteorological data, which have problems such as large workload, low efficiency, and lack of flexibility. With the development of deep learning technology, in order to solve the above problems, considering the typical time series characteristics and parameter correlation features of meteorological data, a time series anomaly detection algorithm based on LSTM-TCN is proposed and applied in meteorological data anomaly detection to improve the efficiency and accuracy of meteorological data quality control. The experimental results show that the proposed model outperforms single time series models (LSTM and TCN), with stronger ability to extract time series features and better prediction accuracy. And through the verification of meteorological station observation data, it has been proven that the model can accurately identify anomalies in the time series of meteorological data, thereby achieving quality control of meteorological data.