Research on Drought Assessment in the Yellow River Basin Based on Deep Learning
Drought is a crucial factor influencing agricultural production in the Yellow River Basin.The assessment of drought holds significant importance for promptly identi-fying drought occurrences and formulating disaster prevention and mitigation measures.In this study,centered on the Yellow River Basin,the concept of data fusion is employed.The Copula function is utilized to merge the feature layers of meteorological drought and hy-drological drought,resulting in the creation of the Drought Convergence Index(DCI).An automatic model is proposed that dynamically adjusts the inertia weight and learning factor using the Adaptive Particle Swarm Optimization algorithm(APSO).This model is combined with the Long Short-Term Memory neural network(LSTM)to form the APSO-LSTM model for predicting DCI.The findings indicate that:1)DCI combines the advantages of both the standardized precipitation index and the standardized runoff index,accurately describing the onset and duration of drought;2)The APSO algorithm enhances both global search capability and local convergence ability,improving the accuracy of the early search range of particles.3)By utilizing the APSO algorithm to optimize the number of iterations,batch processing,the number of hidden layers in the LSTM model,and the learning rate of the Adam algorithm,re-sults are comparable to manually adjusted LSTM models.In comparison to the APSO-LSTM model,the prediction error is smaller,with an average reduction of 65.6％in Mean Squared Error(MSE)and an average increase of 12.7％in R2.The model exhibits a higher fit and demonstrates a superior predictive effect.
NETL
NSTL
万方数据
