改进PSO-LSTM的水文时间序列预测

为更准确地预测中小河流水文时间序列变化,建立改进粒子群优化算法(PSO)与长短期记忆神经网络(LSTM)结合的预测模型.提出利用非线性惯性权重变化,加入自适应变异等操作的方法,改善PSO的寻优能力;实现LSTM与注意力机制(attention mechanism)的结合,建立PSO-LSTM组合模型,改变传统LSTM在...

Hydrological time series prediction based on improved PSO-LSTM

To predict the hydrological time series changes of small and medium rivers more accurately,a prediction model based on improved particle swarm optimization(PSO)and long short-term memory neural network(LSTM)was established.To improve the optimization ability of PSO,the nonlinear inertia weight change and adaptive mutation were used.The combination of LSTM and attention mechanism was realized,and the combination model of PSO-LSTM was established,which changed the difficulty of parameter selection and inaccurate prediction of traditional LSTM in hydrological prediction,and improved the fitting ability of hydrological time series.Through experiments,it is verified that the model can be better applied to hydrological time series prediction of small and medium rivers in complex hydrological data processing.