用水总量预测的研究进展及展望

Research progress and prospects of total water consumption prediction

摘要

【目的】用水总量管控是最严格水资源管理制度的关键指标，面对持续波动的用水需求，传统单一预测模型难以有效捕捉这种动态调整。【方法】为了更好地预测用水总量，结合当前用水总量预测领域的研究进展，从统计规律、用水机理、用水定额、模型构建等多维度，系统梳理了现有用水总量预测方法的发展动态及其适用范围。【结果】结果显示：我国早期对中长期用水总量的预测结果普遍高于实际数据；基于统计规律的用水预测，高度依赖于历史数据，短期预测参考性较好；基于用水机理的用水预测，在试验点或者范围比较小的地方适用，对于宏观规划无法直接应用；基于用水定额的用水预测，适用中长期、区域、不同用水户的预测；基于模型构建的用水预测，对数据要求高、依赖性强，预测准确性高，适应性强。【结论】早期的长期用水总量预测存在准确性不足问题，且我国用水趋势呈现非线性演变特征，现有预测方法在适用性和准确性方面也存在一定的局限性，因此，未来研究需转向多元方法综合集成或多层模型构建以提升科学性和实用性。

Abstract

[Objective] Total water consumption control serves as a key indicator of the most stringent water resource management system. Faced with continuously fluctuating water demand, traditional single prediction models fail to effectively capture such dynamic adjustments. [Methods] To enhance the accuracy of total water consumption prediction, based on current research progress in this field, the development trends of existing prediction method for total water consumption and their applicability are systematically reviewed from the dimensions of statistical patterns, water consumption mechanisms, water consumption quotas, and model construction. [Results] The results show that China's early prediction result of medium-and long-term total water consumption are generally higher than the actual data. Water consumption predictions based on statistical patterns rely heavily on historical data, demonstrating better reference value for short-term predictions. The prediction method based on water consumption mechanism is suitable for experimental sites or at relatively small-scale areas, but cannot be directly applied to macro-planning. The prediction method based on water consumption quota is applicable to medium-and long-term, regional predictions for different water users. The model construction-based prediction method, while requiring high-quality data and strong dependence, demonstrates high prediction accuracy and strong adaptability. [Conclusion] Early predictions of long-term total water consumption suffer from limited accuracy. Furthermore, the water consumption trend in China demonstrates nonlinear evolution characteristics, and the existing prediction method have limitations in applicability and accuracy. Therefore, future research should focus on comprehensive integration of multiple method or multilayer model construction to enhance scientific validity and practical applicability.

关键词

用水总量 / 用水预测 / 适用性 / 研究进展 / 水资源 / 非线性 / 人工智能 / 机器学习

Key words

total water consumption / water consumption prediction / applicability / research progress / water resources / nonlinearity / artificial intelligence / machine learning

基金资助

国家重点研发计划项目“南水北调西线工程调水对长江黄河生态环境影响及应对策略”(2022YFC3202401)

江苏省水利科技项目“江苏省区域用水总量预测模型技术研究”(2024024)

中央级公益性科研院所基金项目(Y525005)