[1]关思源,张巧珍.基于滤波重构时间序列回归算法的需水量预测[J].中国给水排水,2025,41(7):63-68.
GUANSi-yuan,ZHANG Qiao-zhen.Water Demand Prediction Utilizing Filtering Reconstruction Time Series Regression Algorithm[J].China Water & Wastewater,2025,41(7):63-68.
点击复制
GUANSi-yuan,ZHANG Qiao-zhen.Water Demand Prediction Utilizing Filtering Reconstruction Time Series Regression Algorithm[J].China Water & Wastewater,2025,41(7):63-68.
基于滤波重构时间序列回归算法的需水量预测
中国给水排水[ISSN:1000-4062/CN:12-1073/TU]
卷:
第41卷
期数:
2025年第7期
页码:
63-68
栏目:
出版日期:
2025-04-01
- Title:
- Water Demand Prediction Utilizing Filtering Reconstruction Time Series Regression Algorithm
- 关键词:
- 需水量预测; Savitzky-Golay滤波器; 相关性分析; 滤波重构; 时间序列
- Keywords:
- water demand prediction; Savitzky-Golay filter; correlation analysis; filtering reconstruction; time series
- 摘要:
- 为合理分配供水量、提高供水效率、保障供水安全，自来水厂对用户需水量的准确预测是十分必要的。采用基于样本特征的滤波重构自回归模型，能够在保留趋势性数据的同时，对异常数据进行剔除。通过相关性分析发现，城市需水量与时间具有高度的线性相关性；采用滑动窗口对原数据进行时间序列分析，结合滤波重构，平均绝对百分比误差为2.26%，且近92%的数据落在误差5%以内，明显优于单一时序分析法和其他机器学习。采用该算法进行需水量预测，建模后应用于某自来水公司生产调度，其降低了出厂水压力和供水压力电耗，降低幅度分别为5.64%和4.55%。
- Abstract:
- Accurate prediction of user water demand is essential for water treatment plants to rationally allocate water supply, improve water supply efficiency, and ensure water supply safety. The auto-regressive model, reconstructed through feature-based filtering, effectively eliminates anomalous data while preserving trend information. Through correlation analysis, it was determined that there existed a significant linear relationship between urban water demand and time. By employing a sliding window approach to analyze the time series of the original data in conjunction with filtering reconstruction, the mean absolute percentage error was reduced to 2.26%, with approximately 92% of the data points exhibiting an error range within 5%. This performance is notably superior to that achieved through single time series analysis and other machine learning methods. The water demand prediction model, developed using the specified algorithm, was implemented in the production scheduling of a water company. As a result, the power consumptions associated with product water pressurization and water supply pressurization were reduced by 5.64% and 4.55%, respectively.
更新日期/Last Update:
2025-04-01