Short-Term Water Demand Forecast Based on Deep Learning Method

Guo, Guancheng; Liu, Shuming; Wu, Yipeng; Li, Junyu; Ren, Zhou; Zhu, Xiaoyun

doi:10.1061/(asce)wr.1943-5452.0000992

Cited by 118 publications

(62 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As previously stated, the ARIMA, SVR and RF models are selected to enable the performance comparison with the LSTM based models. The ARIMA is chosen due to its wide applications in both the academic and industry fields, representing a standard urban water demand prediction model (Guo et al, 2018). The SVR and RF models are selected because they are advanced data-driven models that have shown great merits for urban water demand forecasts (Bai et al, 2015, Chen et al, 2017, and hence it is interested to demonstrate whether the LSTM based model (also a type of data-driven model) can outperform the SVR and RF models or not (this comparison has not been done in the area of the urban water demand prediction).…”

Section: Short-term Urban Water Demand Prediction Modelsmentioning

confidence: 99%

“…Medium-term forecasts often predict demands at a monthly or yearly resolution, and these predictions are mainly used to develop strategies for water usages (Ghiassi et al, 2008). Short-term forecasts at hourly or daily resolutions are generally employed to enable the effective operations of water treatment plants or pumping stations, typically aimed to provide sufficient demands for urban users with the lowest operation cost (Guo et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, a type of recurrent neural networks named as the long short-term memory (LSTM) based model has been emerged as an important prediction tool (Guo et al, 2018). Compared to traditional ANNs, the LSTM based model is better suited for time-series predictions as they possess the ability to preserve previous information through learning time series data, thereby improving the accuracy of predictions (Mikolov et al, 2010.…”

Section: Introductionmentioning

confidence: 99%

“…To our best knowledge, only limited studies have been undertaken so far to apply the LSTM based models to predict short-term urban water demands. Guo et al (2018) have made the first attempt to implement the LSTM method for urban water demand predictions. In the study of Guo et al (2018), the performance of the LSTM based model has been compared with ARIMA and ANNs based on data with 15-minute resolution, and results showed that the LTSM based models exhibited better capacity than the other two methods in predicting accurate water demands.…”

Section: Introductionmentioning

confidence: 99%

“…Given that the LSTM has only been investigated in Guo et al (2018), there is therefore a lack of sufficient case study application experience as well as comprehensive understanding on its performance in dealing with short-term urban water demand forecasts. These include how the LSTM based models perform (i) when handling urban water demand predictions with various time resolutions as only 15-minute resolution data were considered in Guo et al (2018), (ii) when predicting inflection data points that have abrupt changes relative to their corresponding nerbouring demand values, as well as data with a relatively high uncertainty level, (iii) when comparing with other advanced data-driven models such as SVR and RF models, in addition to the traditional ARIMA model considered in Guo et al (2018), and (iv) when predicting data with a 24-hour time resolution with the aid of external covariates (such as temperature and rainfall). The present case study paper aims to provide additional and improved knowledge as well as evaluations regarding the LSTM' performance in predicting short-term urban water demands, thereby enabling the wider up-takes of the LSTM based models for real-world applications.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Hourly and Daily Urban Water Demand Predictions Using a Long Short-Term Memory Based Model

Li¹,

Zheng

Tao

et al. 2020

J. Water Resour. Plann. Manage.

View full text Add to dashboard Cite

This case study uses a long short-term memory (LSTM) based model to predict short-term urban water demands for the Hefei City of China. The performance of the LSTM based model is compared with autoregressive integrated moving average (ARIMA) model, the support vector regression (SVR) model and the random forests (RF) model based on data with time resolutions ranging from 15-minute to 24-hour. Additionally, this paper investigates the performance of the LSTM based model in predicting multiple successive data points. Results show that the LSTM based model can offer predictions with improved accuracy than the other models when dealing with data with high time resolutions, data points with abrupt changes and data of a relatively high uncertainty level. It is also observed that the LSTM based model exhibit the best performance in predicting multiple successive water demands with high time resolutions. In addition, the inclusion of external parameters (e.g., temperature) cannot enhance the performance of the LSTM based model, but it can improve ARIMAX's prediction ability (ARIMAX is the ARIMA with variables). These obtained insights based on the Hefei case study provide additional and improved knowledge as well as evaluations regarding the LSTM based models used for short-term urban water demand forecasting, thereby enabling their wider take-ups in practical applications.

show abstract

Section: Short-term Urban Water Demand Prediction Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Hourly and Daily Urban Water Demand Predictions Using a Long Short-Term Memory Based Model

Li¹,

Zheng

Tao

et al. 2020

J. Water Resour. Plann. Manage.

View full text Add to dashboard Cite

show abstract

A hybrid model for monthly water demand prediction: A case study of Austin, Texas

Matthews

Han

2020

AWWA Water Science

View full text Add to dashboard Cite

Water demand prediction is an essential prerequisite for urban water resource allocation and pipeline network management. This article uses a hybrid model to improve the accuracy of predicting water demand. The model combines a simulated annealing (SA) algorithm, genetic cross factor (GCF), fruit fly optimization algorithm (FFOA), and support vector machine (SVM) to predict monthly water demand. The monthly water consumption data from 2012 to 2017 in Austin, Texas, are used to develop and test the model. The prediction results are compared with other intelligent algorithms, such as the hybrid model combination particle swarm optimizer with SVM and back-propagation neural network. The results show that the mean absolute percentage error of FFOA-GCFSA-SVM is the smallest, and it is reduced by 16% and 51.17%, respectively, compared with the other two models. Moreover, the influences of kernel function and input variables on prediction results are analyzed, and the effects of sliding window length, data classification, and forecasting period on prediction results are discussed. We concluded that (1) for the city of Austin, the model with a Gaussian kernel function and historical data period of 6 months has higher prediction accuracy;(2) the monthly water demand in Austin is mainly influenced by the local population; and (3) data classification by season can reduce the interference of temperature on prediction. K E Y W O R D Sfruit fly optimization algorithm, genetic cross factor, monthly, simulated annealing algorithm, support vector machine, water demand prediction

show abstract

Prediction of Water Consumption in 31 Provinces of China Based on FGM(1,1) Model

Qiao

Yang

2022

CLEAN Soil Air Water

View full text Add to dashboard Cite

Human production activities and social development cannot be separated from the indispensable water resources. In recent years, with the increasing speed of China's economic development, the consumption of water resources has gradually become the focus of attention of the Chinese government and the public. Based on the original data of water use of 31 provinces (municipalities and autonomous regions) in China from 2014 to 2018, a fractional order accumulative gray prediction model (FGM(1,1)) is established to predict the water use of 31 provinces (municipalities and autonomous regions) in China in the short term, and the applicability of the model is verified by the mean absolute percentage error (MAPE) value. According to the MAPE value, the data obtained by the FGM(1,1) model is relatively accurate. The prediction results show that in case of little change in total water consumption, agricultural water and industrial water use show a downward trend, while domestic water use shows an upward trend. The government's water management policy for urban residents needs to be adjusted, and the publicity work to improve the water-saving awareness of urban residents needs to be improved.

show abstract

Short-Term Water Demand Forecast Based on Deep Learning Method

Cited by 118 publications

References 32 publications

Hourly and Daily Urban Water Demand Predictions Using a Long Short-Term Memory Based Model

Hourly and Daily Urban Water Demand Predictions Using a Long Short-Term Memory Based Model

A hybrid model for monthly water demand prediction: A case study of Austin, Texas

Prediction of Water Consumption in 31 Provinces of China Based on FGM(1,1) Model

Contact Info

Product

Resources

About