基于图深度学习的供水管网短期需水量预测研究

林昱道; 陶涛; 信昆仑; 蒲政衡; 陈磊

doi:10.13205/j.hjgc.202304021

基于图深度学习的供水管网短期需水量预测研究

doi: 10.13205/j.hjgc.202304021

同济大学环境科学与工程学院, 上海 200092

详细信息

作者简介:
林昱道(1998-),男,硕士研究生,主要研究方向为给排水管网智能调度。linyudao2010@hotmail.com

通讯作者:
陶涛(1974-),女,教授,主要研究方向为城市水资源与给排水管网设计运行最优化。taotao@tongji.edu.cn

计量
- 文章访问数: 324
- HTML全文浏览量: 33
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2022-04-21
- 网络出版日期: 2023-05-26
- 刊出日期: 2023-04-01

GRAPH DEEP LEARNING: APPLICATION ON SHORT-TERM WATER DEMAND FORECASTING FOR WATER DISTRIBUTION NETWORK

College of Environmental Science and Engineering, Tongji University, Shanghai 200092, China

摘要

摘要: 供水系统短期需水量预测是供水管网优化调度的基础,预测的准确性对供水调度有重要影响。图神经网络是一种新型神经网络架构,通过学习对象的拓扑特征优化学习效果。利用供水管网的拓扑特征搭建了图波网络(Graph Wavenet)模型,用于学习供水管网各监测点的空间拓扑关系及其需水量时间序列关系,实现了更准确的供水管网短期需水量预测。使用真实管网历史数据训练Graph Wave Net模型,未来5 min预测的最小百分比误差达到1.28%,平均百分比误差达到1.34%。实验成功地将给水管网拓扑结构用于优化深度学习模型效果,使用图卷积网络模型有效提升了供水管网短期需水量预测精度。并对图神经网络在供水管网领域的应用前景进行了简要探讨。
- 需水量预测 /
- 供水系统 /
- 图神经网络 /
- 时间序列预测
Abstract: Water demand forecasting is the basis of the optimization of water networks, and its accuracy is important for the following work. Graph neural network, an emerging architect, takes advantage of the topological information to enhance the learning results, which means the topology of the water network can an input feature of the graph-based neural network model and lift the prediction accuracy. This work tried to utilize Graph Wavenet model in water demand forecasting and make use of the topological feature of the water network. After utilizing real-world history data in the training process, the average 5 min prediction MAPE error result was 1.28%. Besides, this experiment showed that Graph Wavenet raised the prediction accuracy, compared to WaveNet without graph convolution operation, confirming the contribution of the graph convolution mechanism. This work also talked about future applications of GNN on water distribution networks.
- water demand forecast /
- water distribution network /
- graph neural network /
- time series forecasting

HTML全文

参考文献(18)

[1]	HOWE C W, LINAWEAVER F P JR. The impact of price on residential water demand and its relation to system design and price structure[J]. Water Resources Research, 1967, 3(1): 13-32.
[2]	JAIN A, VARSHNEY A K, JOSHI U C. Short-term water demand forecast modelling at IIT Kanpur using artificial neural networks[J]. Water Resources Management, 2001, 15(5): 299-321.
[3]	ADAMOWSKI J, FUNG C H, PRASHER S O, et al. Comparison of multiple linear and nonlinear regression, autoregressive integrated moving average, artificial neural network, and wavelet artificial neural network methods for urban water demand forecasting in Montreal, Canada[J]. Water Resources Research, 2012, 48(1): W01528.
[4]	BRAUN M, BERNARD T, PILLER O, et al. 24-hours demand forecasting based on SARIMA and support vector machines[J]. Procedia Engineering, 2014, 89(1): 926-933.
[5]	XU Y B, ZHANG J, LONG Z Q, et al. Hourly urban water demand forecasting using the continuous deep belief echo state network[J]. Water, 2019, 11(2): 351-362.
[6]	GLOROT X, BORDES A, BENGIO Y. Domain adaptation for large-scale sentiment classification: a deep learning approach[C]//ICML, 2011.
[7]	GOUWS S. Deep unsupervised feature learning for natural language processing[C]//Proceedings of the NAACL HLT 2012 Student Research Workshop, 2012: 48-53.
[8]	SCHMIDHUBER J. Deep learning in neural networks: an overview[J]. Neural Networks, 2015, 61: 85-117.
[9]	SUN Y. Deep learning face representation by joint identification-verification[D]. Hong Kong: The Chinese University of Hong Kong, 2015.
[10]	GUO G C, LIU S M, WU Y P, et al. Short-term water demand forecast based on deep learning method[J]. Journal of Water Resources Planning and Management, 2018, 144(12): 04018076.
[11]	ABDELNASSER A, RASHAD M, HUSSEIN S. A two-layer water demand prediction system in urban areas based on micro-services and LSTM neural networks[J]. IEEE Access, 2020, 8: 147647-147661.
[12]	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[C]// Proceedings of the 5th International Conference on Learning Representations, Toulon, France, 2017.
[13]	ZHOU X, LIU S M, XU W R, et al. Bridging hydraulics and graph signal processing: a new perspective to estimate water distribution network pressures[J]. Water Research, 2022, 217: 118416.
[14]	FU M L, RONG K Z, HUANG Y Y, et al. Graph neural network for integrated water network partitioning and dynamic district metered areas. Scientific Reports, 2022, 12(1): 19466.
[15]	VERMAAK J, BOTHA E C. Recurrent neural networks for short-term load forecasting[J]. IEEE Transactions on Power System, 1998, 13(1), 126-132.
[16]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural computation, 1997, 9(8): 1735-1780.
[17]	OORD A, DIELEMAN S, ZEN H, et al. Wavenet: A generative model for raw audio[C]//Proceedings of 9th ISCA Workshop on Speech Synthesis Workshop, Sunnyvale, USA, 2016.
[18]	WU Z, PAN S, LONG G, et al. Graph wavenet for deep spatial-temporal graph modeling[C]//Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, Macao, China, 2019.