基于小波神经网络的供水预测研究与应用

Research and Application of Water Supply Prediction Based on WaveletNeural Network

针对城镇供水预测不准确的问题,设计开发了一种基于小波变换的时间序列神经网络供水预测系统。首先使用小波变换提取供水流量曲线的规律性,去除流量数据中的高频噪声,并对假期变量和天气温度变量进行非线性处理,对比分析表明模型加入小波变换后,预测准确率显著提高,同时该系统可以计算流量实际值与预测值的实时差值,比较差值和设定阈值判断供水异常情况。经验证,该系统供水预测准确率达到95.9%,可为供水预测和供水异常识别提供决策支持。

Aiming at the problem of inaccurate urban water supply forecasting,a time series neural network water supply forecasting systerm based on wavelet transform was designed and developed.Firstly,the system uses wavelet transform to summarize the regularity of water supply flow curve and remove the high-frequency noise in the flow data.Then,nonlinear processing was carried out on the holiday variable and weather temperature variable.Comparative analysis proves that the forecasting accuracy was significantly improved after the introduction of wavelet transform to the model.At the same time,the system can calculate the real-time difference between the actual flow value and the predicted flow value.When the difference exceeds the set threshold,the water supply is considered abnormal.The actual data verify that the water supply forecasting accuracy of the system reaches 95.9%,which can provide suggestions for final decision for water supply forecasting and water supply anomaly recognition.