陶志勇, 胡启振, 任晓奎. 基于二层分解技术和改进神经网络的河流溶解氧预测研究[J]. 云南大学学报(自然科学版), 2022, 44(2): 262-270. doi: 10.7540/j.ynu.20210194
引用本文: 陶志勇, 胡启振, 任晓奎. 基于二层分解技术和改进神经网络的河流溶解氧预测研究[J]. 云南大学学报(自然科学版), 2022, 44(2): 262-270. doi: 10.7540/j.ynu.20210194
TAO Zhi-yong, HU Qi-zhen, REN Xiao-kui. Prediction of dissolved oxygen in river based on two level decomposition and improved neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2022, 44(2): 262-270. DOI: 10.7540/j.ynu.20210194
Citation: TAO Zhi-yong, HU Qi-zhen, REN Xiao-kui. Prediction of dissolved oxygen in river based on two level decomposition and improved neural network[J]. Journal of Yunnan University: Natural Sciences Edition, 2022, 44(2): 262-270. DOI: 10.7540/j.ynu.20210194

基于二层分解技术和改进神经网络的河流溶解氧预测研究

Prediction of dissolved oxygen in river based on two level decomposition and improved neural network

  • 摘要: 针对河流溶解氧质量浓度序列的非线性和不稳定性导致的预测精度低的问题,提出二层分解技术和改进神经网络相融合的预测模型. 首先,引入自适应噪声的完整集成经验模态分解对溶解氧时序数据进行分解,通过计算分解后各本征模函数(Intrinsic Mode Functions,IMF)的排列熵值以量化序列的复杂性,用变分模态分解对熵值较高的IMF进行二次分解,进一步削弱序列的非线性和不稳定性从而保证预测精度;其次,使用麻雀搜索算法优化神经网络的权值和阈值并对各分量进行预测;最后,将各分量预测结果重构后得到最终预测结果. 实验结果表明,所提预测模型平均绝对误差为0.091,均方根误差为0.14,平均绝对百分比误差为0.96%,决定系数为0.948,优于其它预测模型.

     

    Abstract: Aiming at the problem of low prediction accuracy caused by the nonlinearity and instability of dissolved oxygen concentration series in rivers, a prediction model based on the fusion of two-level decomposition technology and improved neural network is proposed in this paper. Firstly, the complete ensemble empirical mode decomposition of adaptive noise is introduced to decompose the dissolved oxygen time series data. The permutation entropy of each Intrinsic Mode Function (IMF) after decomposition is calculated to quantify the complexity of the sequence, and the variational mode decomposition is used to decompose the IMF with higher entropy. The nonlinearity and instability of the sequence are further weakened to ensure the prediction accuracy. Then, sparrow search algorithm is used to optimize the weights and thresholds of the neural network, and each component is predicted. Finally, the final prediction result is obtained by reconstructing the prediction results of each component. The experimental results show that the average absolute error of the proposed model is 0.091, The root mean square error is 0.14, The average absolute percentage error was 0.96%, The coefficient of determination is 0.948, which is better than other prediction models.

     

/

返回文章
返回