基于多因子XLSTM模型的地下水位动态模拟与预测——以中国邢台为例
Dynamic Simulation and Prediction of Groundwater Level Based on Multi-Factor XLSTM Model—A Case Study of Xingtai, China
DOI: 10.12677/ag.2026.164057, PDF,   
作者: 刘 标*, 张海波：新疆邢美矿业有限公司，新疆 库尔勒；王 锴：中国煤炭地质总局水文地质局，河北 保定；张心月, 王浩冉, 范 博：中国煤炭地质总局第三水文地质队，河北 邯郸
关键词: XLSTM；地下水位；深度学习；时序建模；XLSTM； Groundwater Level； Deep Learning； Time Series Modeling
摘要: 地下水是重要的战略性淡水资源，在开采过程中出现的大规模地下水位的下降是一个普遍存在的问题，地下水有效预报对地下水资源管理有着重要意义。本文采用最新提出的xLSTM模型来解决上述难点。本文以河北省邢台地区选取的5口观测孔2018~2023年实际资料为依据，并根据降水量、温度、蒸发量等因素对模型进行率定分析。上述试验表明，在模型精度及适应性方面xLSTM模型优于传统的LSTM及BP，对于水位波动较大的情况具有更好的识别效果，可以更好地指导可持续开采困难地区的实践工作。
Abstract: Groundwater is a key freshwater resource, yet long-term extraction has led to widespread declines in water levels, posing challenges for sustainable management. Accurate forecasting is therefore essential. This study applies a recently developed xLSTM model to simulate groundwater dynamics using data from five observation wells in Xingtai, Hebei Province (2018~2023), incorporating precipitation, temperature, and evaporation as driving factors. Results show that, for most wells and metrics, xLSTM generally achieves higher accuracy and more stable performance than LSTM and BP models. It is particularly effective in capturing periods of pronounced water level fluctuations, indicating its potential for supporting groundwater management in complex hydrological settings.
文章引用：刘标, 张海波, 王锴, 张心月, 王浩冉, 范博. 基于多因子XLSTM模型的地下水位动态模拟与预测——以中国邢台为例[J]. 地球科学前沿, 2026, 16(4): 623-635. https://doi.org/10.12677/ag.2026.164057

参考文献

[1] 曹文庚, 郭华明, 郑焰, 等. 华北平原地下水砷异常识别-成因-预警关键技术[J/OL]. 地球学报, 1-11.
https://link.cnki.net/urlid/11.3474.P.20260326.1426.006, 2026-04-28.
[2] 王铂淮, 周帅, 李紫妍, 等. 地下水超采区地下水时空动态演变特征预测及其驱动力研究[J/OL]. 水电能源科学, 1-10. 2026-04-28.[CrossRef] 
[3] 李传科, 徐扬, 鲁帆. 基于逐步回归的地下水超采区水位预测及控制指标研究[J]. 中国水利水电科学研究院学报(中英文), 2023, 21(4): 360-368.
[4] 王军进, 张洪伟, 张国珍, 等. 地下水数值模拟方法的研究与应用进展[J]. 环境与发展, 2018, 30(6): 103-104+106.
[5] 陈霞, 陈少华, 金震, 等. 地下水污染数值模拟与应用进展[J]. 安徽化工, 2025, 51(4): 120-122.
[6] 刘玲, 魏亚强, 陈坚, 等. 动网格在非饱和饱和界面数值模拟中的应用研究进展[J]. 地质科技通报, 2023, 42(1): 360-368+377.
[7] 王辰辰. 南水北调中线滏阳河段生态补给地下水数值模拟研究[D]: [硕士学位论文]. 邯郸: 河北工程大学, 2025.
[8] 惠磊, 乔鑫, 高国琰, 等. 基于MODFLOW与CNN-LSTM-Attention模型的月牙泉水位预测模拟[J/OL]. 水电能源科学, 1-12.
https://link.cnki.net/urlid/42.1231.TK.20260306.0933.006, 2026-04-28.
[9] 周宇. 基于深度学习与数值模拟的河谷雨洪资源优化利用研究[D]: [硕士学位论文]. 三河: 防灾科技学院, 2025.
[10] 饶庆阳, 杨琼波, 崔东文. 基于WPT二次分解与CPO优化的KAN地下水位预测模型[J/OL]. 人民珠江, 1-12.
https://link.cnki.net/urlid/44.1037.TV.20260105.1427.002, 2026-03-17.
[11] 潘思成, 潘秀昌, 崔东文. 基于数据处理与超参数优化的高斯过程回归月地下水位预测[J/OL]. 三峡大学学报(自然科学版), 1-8.
https://link.cnki.net/urlid/42.1735.TV.20251111.1704.002, 2026-03-17.
[12] Seidu, J., Ewusi, A., Kuma, J.S.Y. and Ziggah, Y.Y. (2026) Optimised Stacking Generalisation Methodology for Groundwater Level Prediction. Environmental Research Communications, 8, Article 035006. [Google Scholar] [CrossRef] 
[13] Zhang, Y., Li, H., Zhong, Y., Liu, W., Chen, S., Zhang, X., et al. (2026) Comparative Assessment of Machine-Learning Models for Daily Groundwater Level Prediction in a Metropolis, Southwestern China. Journal of Hydrology: Regional Studies, 64, Article 103233. [Google Scholar] [CrossRef] 
[14] Yashooa, N.K., Boo, K.B.W., Cherubini, C., Huang, Y.F., Sham, F.F., Sherif, M., et al. (2026) Integration of an Adaptive Neuro-Fuzzy Inference System (ANFIS) Model for Groundwater Level Prediction Utilizing Feature Engineering Techniques in the Training Process. Neural Computing and Applications, 38, Article No. 79. [Google Scholar] [CrossRef] 
[15] Wei, H., Wei, G., Yu, B., Peng, Y., Wang, M., Xu, B., et al. (2026) A Coupled Spatial Reduction-Reconstruction and LSTM Framework (SRR-LSTM) for Groundwater Level Prediction in Large Irrigation Districts. Scientific Reports, 16, Article No. 7450. [Google Scholar] [CrossRef] 
[16] Banadkooki, F.B., Ghanbari-Adivi, E., Sayyahi, F. and Ehteram, M. (2026) An Intelligent Hybrid Deep Learning-Machine Learning Model for Monthly Groundwater Level Prediction. Scientific Reports, 16, Article No. 4132. [Google Scholar] [CrossRef] 
[17] Hu, S., Du, M., Yang, J., Liu, Y., Tuo, Z. and Ma, X. (2025) Application of a Hybrid CNN-LSTM Model for Groundwater Level Forecasting in Arid Regions: A Case Study from the Tailan River Basin. ISPRS International Journal of Geo-Information, 15, Article 6. [Google Scholar] [CrossRef] 
[18] 刘婧. 河北省地下水监测站网建设与成果应用的有关思考[J]. 河北水利, 2025(5): 14+30.
[19] 郭天辰, 钱睿智, 刘海婧, 等. 地下水位对河水位的响应暨时序预测分析[J]. 陕西水利, 2025(5): 22-24.
[20] Kanito, D., Benaafi, M. and Baalousha, H.M. (2025) Machine Learning Models for Groundwater Level Prediction and Uncertainty Analysis in Ruataniwha Basin, New Zealand. Hydrology, 12, Article 282. [Google Scholar] [CrossRef]