Water Conservancy College, North China University of Water Resources and Electric Power, Zhengzhou, 450046, China.
Collaborative Innovation Center of Water Resources Efficient Utilization and Protection Engineering, Zhengzhou, 450046, China.
Environ Monit Assess. 2022 Jun 2;194(7):471. doi: 10.1007/s10661-022-10149-x.
Industrial water consumption is a major component of the total regional water consumption. Accurate and scientific prediction of industrial water consumption is an essential guide to the rational use of natural resources. In this paper, we proposed a combined model of CEEMD (collective empirical modal decomposition) and ARIMA (autoregressive integrated moving average) for forecasting industrial water consumption to establish an accurate and efficient forecasting model, because of the poor generalization ability of most current industrial water consumption forecasting models. The influencing factors of industrial water consumption are complex, and the data are non-stationary. "Decomposition-prediction-reconstruction" is one of the significant methods for forecasting time series data, and the data decomposition has a suppressive influence on the modal mixing problem in the EMD decomposition procedure. Based on the smoothing ability of CEEMD for non-smooth signals and the better adaptation of the autoregressive moving average prediction model (ARIMA), a combined CEEMD-ARIMA model was established for industrial water consumption forecasting. This study was conducted for industrial water consumption in Henan Province in central China. The results suggest the combined CEEMD-ARIMA model has a favorable forecasting effect, with an average relative percentage error of 1.96%, and mean square error (MSE) of 0.35, a Nash efficiency coefficient (NSE) of 0.95, a prediction pass rate of 100%, and a better prediction accuracy than the ARIMA model and the combined EEMD-ARIMA model. It provides an effective prediction method for the prediction of industrial water consumption and has good application prospects.
工业用水量是总区域用水量的主要组成部分。准确、科学地预测工业用水量是合理利用自然资源的重要指导。在本文中，我们提出了一种基于 CEEMD（集合经验模态分解）和 ARIMA（自回归综合移动平均）的组合模型来预测工业用水量，以建立一个准确高效的预测模型，因为目前大多数工业用水量预测模型的泛化能力较差。工业用水量的影响因素复杂，数据非平稳。“分解-预测-重构”是时间序列数据预测的重要方法之一，数据分解对 EMD 分解过程中的模态混合问题有抑制作用。基于 CEEMD 对非平稳信号的平滑能力和自回归移动平均预测模型（ARIMA）的更好适应性，建立了用于工业用水量预测的组合 CEEMD-ARIMA 模型。本研究以中国中部河南省的工业用水量为例。结果表明，组合 CEEMD-ARIMA 模型具有较好的预测效果，平均相对百分比误差为 1.96%，均方误差（MSE）为 0.35，纳什效率系数（NSE）为 0.95，预测通过率为 100%，预测精度优于 ARIMA 模型和组合 EEMD-ARIMA 模型。为工业用水量预测提供了一种有效的预测方法，具有良好的应用前景。
