Dissolved oxygen prediction based on GRU-CNN optimized with improved whale optimization algorithm
In order to enhance the predictive accuracy of dissolved oxygen in the Yangtze River's water quality profile,a hybrid prediction method incorporating the Gated Recurrent Unit(GRU),Convolutional Neural Networks(CNN),and the Whale Optimization Algorithm(WOA)has been proposed.The model utilizes the GRU to capture long-term historical features of the water quality data,while employing the CNN to extract short-term features from the GRU's input and output through convolutional layers.The final prediction results are obtained by concatenating the outputs of both models using arithmetic operations.To address the issues of low population quality,premature convergence,and sensitivity to parameter settings in the initialization stage of the WOA algorithm,the Tent chaotic map,adaptive inertia weight,and Genetic Algorithm(GA)are introduced into the WOA algorithm.Subsequently,the Improved Whale Optimization Algorithm(IWOA)is applied to optimize the model's parameters.The experimental results demonstrate that the proposed Improved Whale Optimization Algorithm optimized GRU-CNN(IWOA-GRUC+)model delivers outstanding performance.It achieves a notably low MAPE of 2.27％and exhibits excellent results with RMSE,MAE,and R2 values of 0.339,0.216,and 0.913％,respectively.The IWOA-GRUC+model significantly enhances the performance of traditional models in predicting dissolved oxygen(DO)levels.
国家科技期刊平台
NSTL
万方数据
