基于DWT-Informer模型的水量预测研究

Research on Water Consumption Prediction Based on DWT-Informer Model

为准确呈现水消耗的变化趋势以及预测未来的用水需求,提出一种基于DWT-Informer模型的用水量预测方法。与传统方法相比,该预测方法具有以下优势:(1)对历史用水量数据进行DWT分解,可以更好地捕捉用水量信号的不同频率成分和变化趋势;(2)Informer模型具有更强的时间序列建模能力和预测能力,可以更准确地预测未来日用水量;(3)采用多头注意力机制构建输入与输出的全局关系,有利于提升参数水平。通过实际日用水量数据进行算例分析,分析结果表明,相较于其他常用预测方法,该文提出的方法在MAE、RMSE、MAPE等指标上均表现优异。

To accurately present the trend of water consumption changes and predict future water demand,a water consumption prediction method based on the DWT-Informer model is proposed.Compared with traditional methods,this prediction method has the following advantages:(1)DWT decomposition of historical water consumption data can better capture the different frequency components and changing trends of water consumption signals;(2)The Informer model has stronger time series modeling and prediction capabilities,which can more accurately predict future daily water consumption;(3)Using a multi-headed attention mechanism to construct a global relationship between input and output is beneficial for improving parameter levels.Example analysis is conducted based on actual daily water consumption data,the results show that compared to other commonly used prediction methods,the method proposed in this paper performs excellently in MAE,RMSE,MAPE and other indicators.