The concentration of air pollutants is closely related to people’s production and life. Air quality prediction is the premise for environmental management departments to make decisions and put forward pollution control measures. A novel air pollutant prediction model was proposed in this paper to predict air pollutant concentration more accurately. Firstly, the data were decomposed into several subsequences by a complete ensemble empirical mode decomposition with adaptive noise and calculated the sample entropy of the subsequence. Secondly, variational mode decomposition is used to decompose the sequence with the highest sample entropy, and a fast correlation-based filter is used to select the features of the second decomposed sequence and the remaining sequences. Then, a multi-layer perceptron is used to predict the processed quadratic decomposition sequence, and a gated recurrent unit is used to predict the remaining sequences. According to the experimental results, three main conclusions can be drawn. First, through two groups of comparative experiments, it is found that the model has a good prediction effect. Second, after adding the decomposition algorithm, the average improvement levels of mean absolute error and root mean squared error were 44.50% and 34.77%, respectively. Third, after the re-decomposition of intrinsic mode functions 1, the mean absolute percentage error can be reduced by 22.98% on average on the original basis. The results of this study can provide a valuable reference for the prediction of atmospheric pollutants.

空气污染物浓度与人们的生产生活密切相关。空气质量预测是环境管理部门做出决策、提出污染控制措施的前提。本文提出了一种新的空气污染物预测模型，以更准确地预测空气污染物浓度。首先，通过带自适应噪声的完全系综经验模态分解将数据分解为多个子序列，并计算子序列的样本熵。其次，使用变分模式分解来分解样本熵最高的序列，并使用基于快速相关的滤波器来选择第二个分解序列和剩余序列的特征。然后，多层感知器用于预测处理后的二次分解序列，门控循环单元用于预测剩余序列。根据实验结果，可以得出三个主要结论。首先，通过两组对比实验，发现该模型具有良好的预测效果。其次，加入分解算法后，平均绝对误差和均方根误差的平均改善水平分别为44.50%和34.77%。第三，对本征模态函数1重新分解后，平均绝对百分比误差在原来的基础上平均可以降低22.98%。本研究结果可为大气污染物预测提供有价值的参考。门控循环单元用于预测剩余序列。根据实验结果，可以得出三个主要结论。首先，通过两组对比实验，发现该模型具有良好的预测效果。其次，加入分解算法后，平均绝对误差和均方根误差的平均改善水平分别为44.50%和34.77%。第三，对本征模态函数1重新分解后，平均绝对百分比误差在原来的基础上平均可以降低22.98%。本研究结果可为大气污染物预测提供有价值的参考。门控循环单元用于预测剩余序列。根据实验结果，可以得出三个主要结论。首先，通过两组对比实验，发现该模型具有良好的预测效果。其次，加入分解算法后，平均绝对误差和均方根误差的平均改善水平分别为44.50%和34.77%。第三，对本征模态函数1重新分解后，平均绝对百分比误差在原来的基础上平均可以降低22.98%。本研究结果可为大气污染物预测提供有价值的参考。发现该模型具有良好的预测效果。其次，加入分解算法后，平均绝对误差和均方根误差的平均改善水平分别为44.50%和34.77%。第三，对本征模态函数1重新分解后，平均绝对百分比误差在原来的基础上平均可以降低22.98%。本研究结果可为大气污染物预测提供有价值的参考。发现该模型具有良好的预测效果。其次，加入分解算法后，平均绝对误差和均方根误差的平均改善水平分别为44.50%和34.77%。第三，对本征模态函数1重新分解后，平均绝对百分比误差在原来的基础上平均可以降低22.98%。本研究结果可为大气污染物预测提供有价值的参考。