双闭环自适应调参法在城镇供水需水量预测机器学习模型中应用研究

Study on the Application of Dual Closed-Loop Adaptive Parameter Tuning Method in Machine Learning Models for Urban Water Supply Demand Forecasting

针对城市供水企业在日供水量预测方面的需求,提出并设计了一种基于自动控制双闭环调节原理和自适应机器学习的水量预测模型参数调整方法。该方法创新性地引入了日变化系数和15min变化系数,在超出预设误差阈值时启动内外环控制流程调整输出偏差。通过对广东省某自来水公司实际供水数据的分析,采用双闭环自适应调参法建立的水量预测模型在预测日供水量变化时相较于传统预测方法预测精度提升了15%,模型预测结果稳定性也得到一定程度的提高。结果表明,该模型能有效提高供水量预测的准确性和稳定性,为城市供水调度和水资源节约提供了一个先进实用工具,在城镇智慧水务发展中具有重要的应用推广价值。

To address the demand for full-day water demand forecasting in water supply enterprises,a parameter tuning method for water demand forecasting based on the dual closed-loop control theory and adaptive machine learning is designed and developed.The method innovatively introduces the ratio of daily and 15-minute variation coefficients,and triggers the dual closed-loop control processes to adjust output deviations when the preset error threshold is exceeded.Through the analysis of actual water supply data from a water company in Guangdong Province,the application of this parameter tuning method has achieved a 15%increase in prediction accuracy and approximately a 1.2%improvement in model stability compared to traditional forecasting methods in forecasting daily water supply.These results suggest that the model can effectively enhance the accuracy and stability of water supply forecasting,providing a more precise tool for water supply scheduling and resource management,and holds significant practical application value in the development of urban smart water management.