| 引用本文: | 王冠智,粟晓玲,张特, 等.基于DWT-WFGM(1,1)-ARMA组合模型的农业用水量预测[J].灌溉排水学报,2021,(11):106-114. |
| WANG Guanzhi,,SU Xiaoling,ZHANG Te, et al.基于DWT-WFGM(1,1)-ARMA组合模型的农业用水量预测[J].灌溉排水学报,2021,(11):106-114. |
|
| 摘要: |
| 农业用水量预测对于区域水资源规划与管理具有重要意义。【目的】针对农业用水量序列的振荡性以及传统模型预测结果输出单一的问题，提出一种新的组合预测模型DWT-WFGM(1,1)-ARMA对区域农业用水量进行预测。【方法】通过离散小波变换将原始用水量序列分解为近似序列和细节序列，并分别采用自回归滑动平均模型和分数阶灰色模型预测细节序列和近似序列，并结合加权马尔可夫链对近似序列进行误差修正，将不同成分序列的预测结果进行线性叠加得到农业用水量的预测值和预测区间。利用该模型分别对陕西省和内蒙古自治区的农业用水量进行预测，并与灰色模型GM(1,1)、DWT-GM(1,1)-ARMA模型和DWT-FGM(1,1)-ARMA模型对比分析。【结果】DWT-WFGM(1,1)-ARMA模型在陕西省和内蒙古自治区的评价指标平均绝对百分比误差分别为1.25%和1.01%，预测精度高于其他模型，且预测区间为研究区未来时期的农业用水量提供了合理的波动范围，具有一定的实际参考价值。【结论】本文构建的组合模型能够有效提高农业用水量预测的精度，同时预测区间的提出可以为区域农业用水量预测提供更加可靠的依据。 |
| 关键词: 农业用水；分数阶灰色模型；加权马尔可夫链；离散小波变换；预测区间 |
| DOI：10.13522/j.cnki.ggps.2021177 |
| 分类号: |
| 基金项目: |
|
| Predicting Agricultural Water Demand Using the DWT-WFGM (1,1)-ARMA Model |
|
WANG Guanzhi, SU Xiaoling, ZHANG Te, et al
|
|
1. College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling 712100, China;
2. Key Laboratory of Agricultural Soil and Water Engineering in Arid Area of Ministry of Education,
Northwest A&F University, Yangling 712100, China
|
| Abstract: |
| 【Objective】The increased demand for water due to economic development coupled with dwindling water supply is the double whammy facing most provinces in north China. Given agriculture is the biggest water use sector, understanding the change in agricultural water demand is critical to improving water resources management. The purpose of this paper is to present a new method to estimate agricultural water use changes at provincial scale in the north of China.【Method】The proposed method was based on discrete wavelet transform (DWT), fractional-order grey model (FGM(1,1)), weighted Markov Chain (WMC), and autoregressive moving average (ARMA) model. The time series of agricultural water demand was firstly decomposed into approximate series and detailed series, respectively, using DWT. FGM(1,1) was then used to describe the approximate series, with the errors corrected by WMC. The Fisher optimal segmentation method was used to divide the state intervals of the predicted errors, and the state intervals were predicted using a probability transfer matrix. The predicted intervals and values of the approximate series were obtained from the boundary values and median of the predicted error state intervals. In comparison, the detailed series were predicted using the ARMA model based on the Akaike Information Criteria. These were used to predict the agricultural water demand and its intervals. We applied the models to agricultural water demand in Shaanxi and Inner Mongolia provinces, with data measured from 2002 to 2015 used to train the model and those measured from 2016 to 2019 to validate the model. We compared the results calculated from the proposed model with those estimated from the traditional GM (1,1), DWT- GM(1,1)-ARMA, and DWT-FGM(1,1)-ARMA models.【Result】The average absolute error of the proposed model for the two provinces was 1.25% and 1.01%, respectively, much less than those given rise to by other models. The predicted agricultural water demand intervals showed that after correction by WMC, the model provided reliable short-term fluctuation intervals in agricultural water demand in both provinces.【Conclusion】The proposed model for predicting agricultural water demand at provincial scale was accurate and robust. It can also predict the intervals which describe the short-term fluctuation in agricultural water demand. The model has an implication in helping improve water management and developing sustainable agriculture. |
| Key words: agricultural water consumption; fractional-order grey model; weighted Markov chains; discrete wavelet transform; predicted intervals |
