| 注册
首页|期刊导航|净水技术|基于图片特征提取的水泵能耗分析及预测

基于图片特征提取的水泵能耗分析及预测

乔庆 谢小风 刘孟如 方婷婷 史俊 阮应君

净水技术2023,Vol.42Issue(z2):7-16,10.
净水技术2023,Vol.42Issue(z2):7-16,10.DOI:10.15890/j.cnki.jsjs.2023.s2.002

基于图片特征提取的水泵能耗分析及预测

Analysis and Prediction of Pump Energy Consumption Based on Picture Feature Extraction

乔庆 1谢小风 2刘孟如 3方婷婷 2史俊 4阮应君3

作者信息

  • 1. 同济大学中意可持续发展中心,上海 200092
  • 2. 上海城投水务<集团>有限公司,上海 200002
  • 3. 同济大学机械与能源工程学院,上海 200090
  • 4. 同济大学环境科学与工程学院,上海 200092
  • 折叠

摘要

Abstract

This paper proposed a multiple linear regression prediction model for water pump energy consumption based on image feature extraction,which was to improve the accuracy of the analysis of factors affecting the energy consumption of water pumps and the precision of the prediction model.A water pump system installed in a residential community in Shanghai was chosen as the research object.Firstly,the collected historical operating data of the water pump system were preprocessed,and the one-dimensional time series were transformed into two-dimensional gramian angular field(GAF)images.Then,14 parameters were screened using a method based on the correlation between time series and pictures to obtain the corresponding energy consumption features.Finally,based on the results of feature selection,the optimal predictive model was studied using a multiple linear regression model and different methods for determining correlation.The results showed that compared with two time series feature selection methods,the image-based feature selection method exhibits better adaptability.When the number of input features was 4,the predictive model in the testing set increased by 0.87%and 3.36%in the coefficient of determination(R2),and decreased by 10.38%and 28.26%in the mean squared error(MSE)and 6.11%and 24.57%in the mean absolute error(MAE),respectively.The conclusion of this study provides new ideas for the selection of pump system energy consumption features and serves as a theoretical reference for the analysis and prediction of pump system energy consumption in the Shanghai region.

关键词

水泵能耗预测/图片特征提取/相关性判定/多元线性回归拟合/图片生成

Key words

pump energy consumption forecast/picture feature extraction/correlation determination/multiple linear regression fitting/picture generation

分类

建筑与水利

引用本文复制引用

乔庆,谢小风,刘孟如,方婷婷,史俊,阮应君..基于图片特征提取的水泵能耗分析及预测[J].净水技术,2023,42(z2):7-16,10.

基金项目

上海城投科研项目:建筑加压调蓄设施的能效评估及碳减排建设模式研究(CS.JS.002G.02B.2015-0) (CS.JS.002G.02B.2015-0)

净水技术

OACSTPCD

1009-0177

访问量0
|
下载量0
段落导航相关论文