Performance Assessment for Short-Term Water Demand Forecasting Models on Distinctive Water Uses in Korea

Round 1

Reviewer 1 Report

This paper assesses the performance of four short-term demand forecast models for different water uses in a Korean example network. It compares how forecasted values from ARIMA, RBF-ANN, QMMP+ and LSTM methods compare to observed values according to different metrics (residual, root mean squared error, normalized root mean square error, Nash-Sutcliffe efficiency and Pearson correlation coefficient) for different water uses (domestic, restaurant, church, etc.)

This contribution has potential to be published in the journal, but revision is required in order to clarify the paper. My comments are summarized below:

General comments

1. End-use level. The “end-use level” term has misguided me. The idea of end-uses can be understood as associated with the final use of water per microcomponent (e.g. bathroom tap, shower, washing machine – see Blokker et al. 2010; Blokker et al, 2011) rather than with the type of water use (e.g. residential, industrial). I still need some clarification on what is exactly metered in this case study (see rest of comments), but Table 1 makes me think that the paper centres the analysis on distinctive uses (domestic, restaurant, church, etc.). It will help the reader to revise the title and/or clearly explain in the Introduction and the Abstract what is the scope. Also, it would be useful if they could mention the type/s of water demand that have traditionally been forecasted with the methods presented in the Introduction.
2. Short/medium/long-term forecasting. The authors explain that short-term forecasting focuses in 24-48h but, is there any reference on the approximate limits of short/medium/long-term forecasting? If so, it would be useful to include it in the Introduction. This temporal framework is important to assess the consistency of the observed time series, the sampling rate (1 hour) and the applied methods. How would having higher temporal resolution measurements (e.g. few minutes) affect the performance? Higher temporal resolution might help, for example, to better characterize the peak coefficient (Tricarico et al, 2007; Gargano et al, 2017).
3. Other options. This work concludes that short-term forecast models only use time and amount of water consumption, but the authors say that it may advisable to consider other factors that influence demand (like weather, customer behaviour, etc.) in the process. Stochastic demand models are gaining importance to accurately simulate water demands (see Creaco et al, 2017 for references). It may be a good idea to mention them in the final discussion.

Other comments

1. Abstract (page 1). It is too long and it makes it difficult for the reader to identify the objective and methodology that is then presented in the paper. Please revise.
2. Keywords (page 1). On-site sodium hypochlorite generator is included as keyword, but I have seen no mention to this topic in the paper. Please revise.
3. Introduction (pages 2-3). In page 2, statistical models (paragraph 2) and machine learning models (paragraph 3) are revised. Should paragraph 4 in page 2 be presented as related to hybrid models? Four methods are applied in this work, and it would be useful to see more clearly where they belong and why they have been selected.
4. Study process (pages 3-4). From my understanding, the preprocessing of data consists in imputing both missing values and outliers. Then, 70% of the data are used as training data and 30% for validation. This process is applied to tune in each of the four forecast models used in this work. Once these models have been adjusted, they are used to forecast 24-h ahead demand. But what do 70 and 30% refer to over the studied time series (1 January 2018-1 January 2019)? Please clarify how you treat the measurement time series to get the 24-h ahead forecast in the study area.
5. Methodology (pages 4-7). When presenting each of the four methods, the authors often refer to Matlab functions. Would it be possible for the authors to share the code used in this work to process the data? It would definitely help the reader to replicate the analysis and apply it to other metered systems.
6. Performance assessment (page 7). There may be a typo in Eq. 17. I think that the denominator in NSE should correspond to the difference between the observed value and the mean of observed values. Please revise this equation and notation all the way through the paper.
7. Study area (page 8). I am sorry, but I am not sure I understand the layout of this system and how it is metered. From my understanding, your study area is “block 112”, an area populated with 17000 people and 958 customers, and there are 527 meters (i.e. AMIs) operative in the system. You later mention that the total number of water consumers is 527. Does this mean you only monitor part of block 112? You also divide 527 meters in household use (387), general use (138) and bathroom use (2), but these uses do not correspond to the type of uses you identify in Table 1. Please clarify. Figure 2 should be improved.
8. Results and discussion (pages 11-15). If I understand well, you have 527 water meters in the case study area, but you only present forecasted values for 10 of them. Would it not be possible to somehow show the individual performance of all of them? Why have these 10 been selected?
9. Results and discussion (pages 11-15). I think it would be more intuitive to include the type of water use in Figure 6 as well (for example in brackets next to the AMI No.). In page 11 you comment this figure referring to the water use, and this detail will help the reading. Also, in page 11 you mention that one QMMP+ model produced results that are most similar to the actual water consumption, but according to Table 2 there are actually two (110013012 and 110016860).

References

Blokker, E.J.M., Vreeburg, J.H.G. and van Dijk, J.C. (2010) Simulating residential water demand with a stochastic end-use model. Journal of Water Resources Planning and Management, 136(1), 19-26.

Blokker, E.J.M., Pieterse-Quirijns, E.J., Vreeburg, J.H.G. and van Dijk, J.C. (2011) Simulating nonresidential water demand with a stochastic end-use model. Journal of Water Resources Planning and Management, 137(6), 511-520.

Creaco, E., Blokker, M. and Buchberger, S. (2017) Models for generating household water demand pulses: Literature review and comparison. Journal of Water Resources Planning and Management, 143(6), 04017013.

Gargano, R., Tricarico, C., Granata, F., Santopietro, S. and de Marinis, G. (2017) Probabilistic models for the peak residential water demand. Water, 9, 417, doi:10.3390/w9060417

Tricarico, C., de Marinis, G., Gargano, R. and Leopardi, A. (2007) Peak residential water demand. Water Management, 160, 115-121.

Author Response

Thank you very much for the invaluable suggestions and questions.

Point 1-End-use level: The “end-use level” term has misguided me. The idea of end-uses can be understood as associated with the final use of water per microcomponent (e.g., bathroom tap, shower, washing machine – see Blokker et al. 2010; Blokker et al, 2011) rather than with the type of water use (e.g., residential, industrial). I still need some clarification on what is exactly metered in this case study (see rest of comments), but Table 1 makes me think that the paper centres the analysis on distinctive uses (domestic, restaurant, church, etc.). It will help the reader to revise the title and/or clearly explain in the Introduction and the Abstract what is the scope. Also, it would be useful if they could mention the type/s of water demand that have traditionally been forecasted with the methods presented in the Introduction.

 

Response 1: As the reviewer mentioned, we have revised the title of the paper to "Performance Assessment for Short-Term Water Demand Forecasting Models on Distinctive Water Uses in Korea." Also, the end-uses in the text have been replaced to suit the context.

 

Point 2-Short/medium/long-term forecasting: The authors explain that short-term forecasting focuses in 24-48h but, is there any reference on the approximate limits of short/medium/long-term forecasting? If so, it would be useful to include it in the Introduction. This temporal framework is important to assess the consistency of the observed time series, the sampling rate (1 hour) and the applied methods. How would having higher temporal resolution measurements (e.g., few minutes) affect the performance? Higher temporal resolution might help, for example, to better characterize the peak coefficient (Tricarico et al, 2007; Gargano et al, 2017).

Response 2: Taking into account what the reviewer mentioned, we have revised the text.

According to Tiwari and Adamowski [4,5], there is no general rule for forecasting water demand, but it can be classified into short-term (hourly, daily, weekly), medium-term (up to 24-month), and long-term (annual, decadal). In general, water supply managers refer to short-term water demand forecast for one day or up to several weeks based on experiences to manage the system, including pumps and valves, efficiently. As water is supplied from a purification plant to a distribution reservoir using pumps, the operating costs can be reduced if the work is performed at night time when power rates are low. In particular, if the short-term (24–48 h) water demand is forecasted, efficient pump scheduling can guarantee stable water availability at a distribution reservoir [6,7]. Hence, an accurate forecast of short-term water demand is required for the efficient management of a water supply system and the reduction of operating costs and energy [8]. Also, the higher measuring time step can lead significant reductions of the peak demand [9], so higher temporal resolution might help forecast water demand. Accordingly, several studies have been conducted in this regard, but there is a lack of research on assessing the performance of short-term water demand forecasting at the types of water uses with higher temporal resolution.

 

Point 3-Other options: This work concludes that short-term forecast models only use time and amount of water consumption, but the authors say that it may advisable to consider other factors that influence demand (like weather, customer behaviour, etc.) in the process. Stochastic demand models are gaining importance to accurately simulate water demands (see Creaco et al, 2017 for references). It may be a good idea to mention them in the final discussion.

Response 3: We have revised the text as below.

Recently, the water demand forecasting models with probabilistic components are becoming more important to accurately forecast water demand [56]. In particular, the SWG environment, information that can differentiate the characteristics of consumers can be obtained in addition to the data on direct information, such as the amount of water consumption and usage time. Therefore, forecasting methods reflecting the propensity and characteristics of consumers can be considered. Specifically, a water supply system can be managed more precisely if various factors, such as water usage time, the amount of water consumption, purpose, the number of household members, occupation, and weather (precipitation, humidity, temperature), and customers behaviour, etc., are applied for forecasting the water demand.

Point 4-Abstract (page 1): It is too long and it makes it difficult for the reader to identify the objective and methodology that is then presented in the paper. Please revise.

Response 4: We have revised the text as below.

It is crucial to forecast the water demand accurately for supplying water efficiently and stably in a water supply system. In particular, accurately forecasting short-term water demand helps in saving energy and reducing operating costs. With the introduction of the Smart Water Grid (SWG) in a water supply system, the amount of water consumption is obtained in real time through smart meter, which can be used for forecasting the short-term water demand. The models widely used for water demand forecasting include ARIMA, ANN, QMMP+, and LSTM. However, there is a lack of research on assessing the performance of models and forecasting the short-term water demand in SWG demonstration plant. Therefore, in this study, the short-term water demand was forecasted for each model using the data collected from smart meter, and the performance of each model was assessed. The Smart Water Grid Research Group installed smart meter in the block 112 located in YeongJong Island, Incheon, and the actual data used for operating the SWG demonstration plant were adopted. The performance of the model was assessed by using the residual, RMSE, NRMSE, NSE, and PCC as indices. As a result of water demand forecasting, it is difficult to forecast water demand only by time and water consumption. Therefore, as the short-term water demand forecasting models using only time and the amount of water consumption have limitations in reflecting the characteristics of consumers, a water supply system can be managed more precisely if other factors (weather, customer behavior, etc.) influencing the water demand are applied.

 

Point 5-Keywords (page 1): On-site sodium hypochlorite generator is included as keyword, but I have seen no mention to this topic in the paper. Please revise.

Response 5: As the reviewer said, we removed the unnecessary keyword ‘on-site sodium hypochlorite generator’.

Point 6-Introduction (pages 2-3): In page 2, statistical models (paragraph 2) and machine learning models (paragraph 3) are revised. Should paragraph 4 in page 2 be presented as related to hybrid models? Four methods are applied in this work, and it would be useful to see more clearly where they belong and why they have been selected.

Response 6: We have revised the text as below.

Short-term water demand forecasting can be conducted mainly using statistical, Machine Learning (ML), Hybrid, and Deep Neural Network (DNN) model. Firstly, the following studies have applied statistical models. Kofinas et al. [10] forecasted the water demand for cities using the AutoRegressive Integrated Moving Average (ARIMA) model. Zhou et al. [11] forecasted the daily water demand using the AutoRegressive (AR) model and Fourier series by applying the weather parameters (maximum temperature, precipitation, evapotranspiration) of the city of Melbourne, Australia. Wong et al. [9] applied the trend, seasonality, weather regression, and calendar effect (holiday) to analyze the correlation when forecasting the daily water demand of Hong Kong. Hutton and Kapelan [7] diagnosed and reduced the forecasting error using the repeated Bayesian likelihood model, and Do et al. [13] proposed a particle-filter-based model as the statistical model for real-time water demand forecasting [14,15]. In the past, a linear regression model has been widely applied, as it is relatively simple [13]; however, as the changes in water demand are nonlinear and cannot be accurately forecasted with linear regression methods [16], studies have shown that nonlinear regression methods are better than linear regression methods for forecasting the water demand of cities [17]. In addition, Quevedo et al. [18] applied the Seasonal ARIMA (SARIMA) model and the exponential smoothing model using time and daily periods to compare the water demand forecasting results; accordingly, they proved that the exponential smoothing model and the SARIMA model are superior in forecasting the water demand based on time and daily periods, respectively.

Secondly, as studies using ML models, Chang et al. [19] used the Radial Basis Function-Artificial Neural Network (RBF-ANN) model to forecast the water demand. Moreover, Braun et al. [16] used the Support Vector Machine (SVM) and SARIMA models. Further-more, Brentan et al. [20] used the model in which SVM and adaptive Fourier series are combined to forecast the water demand, and proved that better forecasting results were produced than when the SVM model was used alone. Moreover, Candelieri [21] combined the SVM and clustering technique to forecast the water demand in Milan, Italy. An Artificial Neural Network (ANN) model [22-27], Random Forest model [28], Extreme Learning Machine model [17], and Multi Evolutionary ANN model [26], which are all ML models, have been reported to be superior to statistical models.

Thirdly, the studies using Hybrid models that combine the statistical model and the ML model are as follows. Rangel et al. [29] used the concept of daily water consumption pattern predicted based on Nearest Neighbor (NN) node estimation, which is a non-parametric method; Cheifetz et al. [30] estimated the water demand pattern for each day of the week using the hourly water consumption data based on the Fourier regression mixture model. Particularly, Farias et al. [31] applied the NN classification, which is an ML model, and a calendar effect based on quantitative and qualitative information, and proposed the Qualitative Multi-Model Predictor Plus (QMMP+) model for estimating water demand patterns based on the moving average (MA). Here, better results were produced when the total daily water demand was forecasted using the SARIMA model applied with a sliding window [26], whereas the hourly water demand was distributed with the NN model according to a calendar effect for daily patterns and compared with the ANN model.

Lastly, recent studies using DNN model focused on water demand forecasting [32,33], power demand forecasting [34-36], tourism flow forecasting [37], airline demand forecasting [38], and sales demand forecasting [39]. In particular, Li and Cao [37] reported that the forecast accuracy of a Long Short-Term Memory (LSTM) model [40] is higher than that of the ARIMA model for forecasting tourism flows.

Therefore, this study selected ARIMA as a statistical model, RBF-ANN as an ANN model, QMMP+ as a hybrid model, and LSTM as a DNN model for short-term water demand forecasting. As the input data for forecasting water demand, the hourly water consumption data collected at the types of water uses of an SWG demonstration plant of the block 112 located in YeongJong Island, Incheon were used. For assessing the water demand forecasting performance of each model, the forecasted value was comparatively analyzed against the 24-h (one day) observed value. The results of this study can be used as index data for applying a short-term water demand forecasting model to an actual water supply system to which an SWG is applied.

 

Point 7-Study process (pages 3-4): From my understanding, the pre-processing of data consists in imputing both missing values and outliers. Then, 70% of the data are used as training data and 30% for validation. This process is applied to tune in each of the four forecast models used in this work. Once these models have been adjusted, they are used to forecast 24-h ahead demand. But what do 70 and 30% refer to over the studied time series (1 January 2018-1 January 2019)? Please clarify how you treat the measurement time series to get the 24-h ahead forecast in the study area.

Response 7: The training set and test set were divided by time sequence. This is the same as the one-year data (hourly based) used by Lopez Farias divided by 70 and 30 percent, so we mentioned it as a reference and we also revied text.

Lopez Farias, R.; Puig, V.; Rodriguez Rangel, H.; Flores, J.J. Multi-model prediction for demand forecast in water distribution networks. Energies 2018, 11, 660.

In this study, among the water consumption data of one year measured at one-hour intervals, 70% of the data were used as the training dataset over time sequence, whereas the remaining 30% were used as the validation dataset [31].

Imputation method for missing and outliers:

First, to find outliers, the local minima and the variables exceeding the maximum discharge capacity per unit time with respect to a pipe diameter are determined and designated as missing in the accumulated water consumption data. For imputing the missing values, the lagged k-NN is applied to find the lagged time and the training dataset to find the k-NN value; then, for integrating the water consumption patterns within the variable, the Fourier transform is approximated and the estimates of the two methods are averaged.

 

Point 8-Methodology (pages 4-7): When presenting each of the four methods, the authors often refer to Matlab functions. Would it be possible for the authors to share the code used in this work to process the data? It would definitely help the reader to replicate the analysis and apply it to other metered systems.

Response 8: We would share the material through the Github website below.

Supplementary Materials: The code as supplementary material is found at

https://github.com/koo00v/water-demand-forecasting.

Point 9-Performance assessment (page 7): There may be a typo in Eq. 17. I think that the denominator in NSE should correspond to the difference between the observed value and the mean of observed values. Please revise this equation and notation all the way through the paper.

Response 9: We have corrected the wrong term in the equation 17 out by the reviewer.

 

Point10-Study area (page 8): I am sorry, but I am not sure I understand the layout of this system and how it is metered. From my understanding, your study area is “block 112”, an area populated with 17000 people and 958 customers, and there are 527 meters (i.e. AMIs) operative in the system. You later mention that the total number of water consumers is 527. Does this mean you only monitor part of block 112? You also divide 527 meters in household use (387), general use (138) and bathroom use (2), but these uses do not correspond to the type of uses you identify in Table 1. Please clarify. Figure 2 should be improved.

Response 10: As the reviewer asked, the only 528 AMIs out of 958 customers have smart meters installed. This is because it cannot be installed without the consent of the customer, resulting in unmeasured customer's water consumption. And we improved figure 2.

And we revised text also.

The SWGRG [1] installed ultrasonic smart water meter in the block to 527 customers who agreed to install it for the first time in Korea, and operated the SWG demonstration plant from 2017 to 2019 (Figure 2).

 

Point 11-Results and discussion (pages 11-15): If I understand well, you have 527 water meters in the case study area, but you only present forecasted values for 10 of them. Would it not be possible to somehow show the individual performance of all of them? Why have these 10 been selected?

Response 11: In this study, 10 customers were selected without water demand forecasting for all customers.

10 AMIs were selected for forecasting water demand for which the missing values were set to 10% or less to minimize the influence of errors in the simulation results. More-over, for a comparison with the forecasted values, the missing data were not included in the 24-h observed values of the validation dataset; the pipe diameters ranged from 15 mm to 32 mm, and the purpose of use included domestic, restaurant, church, supermarket, senior community center, and laundry. Table 1 lists the details of the AMI dataset. The Total Water Demand (TWD) being forecasted ranges from 0.634 m3/day at the minimum to 12.206 m3/day at the maximum. Generally, the amount of water consumption increases as the pipe diameter increases, and the water consumption was the highest for restaurants, churches, and laundries compared with households.

 

Point 12-Results and discussion (pages 11-15): I think it would be more intuitive to include the type of water use in Figure 6 as well (for example in brackets next to the AMI No.). In page 11 you comment this figure referring to the water use, and this detail will help the reading. Also, in page 11 you mention that one QMMP+ model produced results that are most similar to the actual water consumption, but according to Table 2 there are actually two (110013012 and 110016860).

Response 12: We included the type of water use in Figure 6 and revised text also as reviewer mentioned.

Daily water demand forecasting is also important in terms of stably supplying water in addition to water consumption patterns; the differences in the forecasted daily water demand of each AMI are presented in Table 2 by model. Among the 10 AMIs, two ARIMA models (110013044, 110016799), one RBF-ANN model (110013629), one QMMP+ model (110013012, 110016860), and five LSTM models (110013004, 110013074, 110016389, 110018932) produced the results that are most similar to the actual water consumption but the results were underestimated (Table 2).

 

a) 110012984 (restaurant) -32mm	(b) 110013004 (domestic) -15mm
(c) 110013012 (domestic) -15mm	(d) 110013044 (church)-25mm
(e) 110013074 (Laundry) -25mm	(f) 110013629 (mart) -25mm
(g) 110016389 (pre-primary) -15mm	(h) 110016799 (restaurant) -15mm
(i) 110016860 (senior-citizen center) -32mm	(j) 110018932 (domestic)-15mm

 

Author Response File: Author Response.pdf

Reviewer 2 Report

In the manuscript the important issue of the performance assessment for short-term water demand forecasting models at an end-use level. The study area concerns block 112 located in Unseo-dong and Unbuk-dong of  YeongJong Island, Incheon, to which water is supplied through a single submarine pipeline from GongChon Water Purification Plant in Incheon. For assessing the short-term water demand forecasting of each model, the residual, RMSE, NRMSE, NSE, and PCC were selected as the assessment indices in this study. Remarks: Lack of comparative comments with other territorial areas. Please follow the literature review with a clear and concise state-of-the-art analysis. This should clearly show the knowledge gaps identified and link them to your paper goals. Please reason both the novelty and the relevance of your paper goals. In the second section concerning methodology, all borrowed elements and parts have to be properly summarised and references to the original works provided. How were the ARIMA model orders selected? The methodology should be clarified. In the second section, the authors briefly describe the workflow that they have followed, but they do not explain how they have treated the original data and/or performed the statistical analysis. As makes it more available for water managers. The Author should mention in conclusions, if obtained results will constitute a practical and quick tool and if they were consulted with practitioners.

Author Response

Thank you very much for the invaluable suggestions.

Point 1-Remarks: Lack of comparative comments with other territorial areas. Please follow the literature review with a clear and concise state-of-the-art analysis. This should clearly show the knowledge gaps identified and link them to your paper goals. Please reason both the novelty and the relevance of your paper goals.

Response 1: We revised text as below.

The study area is the block 112, which is located in Unseo-dong and Unbuk-dong of YeongJong Island, Incheon, to which water is supplied through a single submarine pipe-line from GongChon Water Purification Plant. And Incheon International Airport, the hub of Northeast Asia, is located, it is the best place to apply ICT technology in response to the water crisis.

 

Point 2-Methodology: In the second section concerning methodology, all borrowed elements and parts have to be properly summarised and references to the original works provided. How were the ARIMA model orders selected? The methodology should be clarified.

Response 2: As the reviewer mentioned us, we revised the text for four models and mentioned the reference code in the supplementary material. Thank you very much.

 

 

Point 3-Study process: In the second section, the authors briefly describe the workflow that they have followed, but they do not explain how they have treated the original data and/or performed the statistical analysis. As makes it more available for water managers.

Response 3: We mentioned and explained pre-processing as below.

Imputation method for missing and outliers:

First, to find outliers, the local minima and the variables exceeding the maximum discharge capacity per unit time with respect to a pipe diameter are determined and designated as missing in the accumulated water consumption data. For imputing the missing values, the lagged k-NN is applied to find the lagged time and the training dataset to find the k-NN value; then, for integrating the water consumption patterns within the variable, the Fourier transform is approximated and the estimates of the two methods are averaged.

 

Point 4-Conclusions: The Author should mention in conclusions, if obtained results will constitute a practical and quick tool and if they were consulted with practitioners.

Response 4: This study is currently in the process of forecasting water demand based on water consumption data collected through the demonstrative operation of living lab. Thank you.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper presents a performance assessment process of various models used for short-term water demand forecasting in a DMA in Korea. The paper is well-written and presented. 

My only comment is to discuss the results of this study with the results of other studies. For example, what are the advantage and disadvantages of the models used in this study and if these results agree or not with the results from other studies. 

Author Response

Thank you very much for the invaluable question.

Point 1: My only comment is to discuss the results of this study with the results of other studies. For example, what are the advantage and disadvantages of the models used in this study and if these results agree or not with the results from other studies.

 

 

Response 1: In this study, we selected four models that can represent statistical, ANN, hybrid, and DNN models used to forecast water demand. And the advantages and disadvantages of each model are described below and whether they are consistent with the results of other studies.

Advantages

ARIMA model: It can derive a model from past observations and error terms, but it is suitable for short-term forecasting because it weights more observations close to the most recent time point.

RBF-ANN model: It is very powerful for numerical calculations, so fast learning is possible without updating weights by iterative calculations.

QMMP+: A hybrid of a statistical model and an ML model, it can reflect the estimated water use patterns and daily water use.

LSTM: It solved the vanishing gradient problem of RNN models.

Disadvantages

ARIMA Model: It cannot reflect the periodic characteristics of time series data, so it is difficult to apply when raw time-series data show seasonality.

RBF-ANN model: All nodes in the hidden layer must compute the RBF function for the input sample vector during classification.

QMMP+ and LSTM: It is difficult to estimate raw time-series data with irregular periods.

As a result of this study, like other studies, the ANN model and DNN model's demand forecasting results are better than the statistical model. In particular, the QMMP+ forecasted results are similar to the results of Lopez Farias et al.

However, all models we selected concluded that under or overestimation of peak runoff and peak runoff time inconsistent are insufficient as a demand forecasting model based on the results of performance assessment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The article concerns concerns forecasting the demand for water. It may be interesting for readers of Sustainability. In general, this manuscript is well organized and written, with comprehensive literature review, detailing the framework approach of the study, clearly stated methodology and nicely presented findings.

The only important aspect, not addressed by the Authors, is to reduce water leaks. This is one of the main purposes of metering the water supply network. Can the Authors explain how this research and type of metering could be used in this aspect?

 

Author Response

Thank you very much for the invaluable question.

Point 1: The only important aspect, not addressed by the Authors, is to reduce water leaks. This is one of the main purposes of metering the water supply network. Can the Authors explain how this research and type of metering could be used in this aspect?

 Response 1: The most fundamental cause of water leakage is the deterioration of the pipeline. Replacing an aging pipeline is a reliable way to reduce leakage, but it is difficult to find where the leakage occurs. Therefore, detecting the leak point in the short time will be a countermeasure to reduce the leak. In this study, water demand can be forecasted in advance by collecting water consumption data for each customer through a smart water meter. If hydraulic analysis of WDNs is performed using the forecasted water demand as input data, the water pressure of each point in WDNs can be predicted. The water pressure meters are installed at various points in the WDNs and the water pressure is compared in real time with the calculated water pressure by hydraulic analysis, it can be seen that there is a leak when the actual water pressure is significantly lower than the calculated water pressure.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have answered most of my comments and have revised the paper accordingly. I still have some remaining concerns:

1. Page 8. I now better understand the case study. However, I still cannot see what is the relationship between household use (387) - general use (138) - bathroom use (2) and the water uses that are later identified in the 10 AMIs (restaurant, domestic, church, etc.).
2. Page 11. The authors mention that two ARIMA models (110013044, 110016799), one RBF-ANN model (110013629), one QMMP+ model (110013012, 110016860) and five LSTM models (110013004, 110013074, 110016389, 110018932) produced the results that are most similar to the actual water consumption. However, the number of AMIs indicated in brackets do not support the two/one/one/five structure. Please revise.
3. Please revise English language throughout the text. For example in page 7, I am not sure about the meaning and implications of this sentence. “And Incheon International Airport, the hub of Northeast Asia, is located, it is the best place to apply ICT technology in response to the water crisis”.

Author Response

Thank you very much for the invaluable suggestions and questions.

Point 1- Page 8: I now better understand the case study. However, I still cannot see what is the relationship between household use (387) - general use (138) - bathroom use (2) and the water uses that are later identified in the 10 AMIs (restaurant, domestic, church, etc.).

Response 1: We deleted the sentence isn't related to the classification (restaurant, domestic, church, etc.) mentioned later.

 

Point 2-Page 11: The authors mention that two ARIMA models (110013044, 110016799), one RBF-ANN model (110013629), one QMMP+ model (110013012, 110016860) and five LSTM models (110013004, 110013074, 110016389, 110018932) produced the results that are most similar to the actual water consumption. However, the number of AMIs indicated in brackets do not support the two/one/one/five structure. Please revise.

Response 2: We modified the structure as follow.

two ARIMA models (110013044, 110016799), one RBF-ANN model (110013629), two QMMP+ model (110013012, 110016860) and four LSTM models (110013004, 110013074, 110016389, 110018932)

                                                     

Point 3: Please revise English language throughout the text. For example, in page 7, I am not sure about the meaning and implications of this sentence. “And Incheon International Airport, the hub of Northeast Asia, is located, it is the best place to apply ICT technology in response to the water crisis”.

Response 3: We revised the text pointed out by the reviewer.

The study area is the block 112 (Unbuk-dong, Unseo-dong) in YeongJong Island, where Incheon International Airport, a hub of Northeast Asia, is located. In this area, water is supplied through a single submarine pipeline, making the place optimal for responses to water crises.

Author Response File: Author Response.pdf

Reviewer 2 Report

Please present a discussion concerning the reliability and validity of your research in the paper. Add the DOI number to the references.

Author Response

Thank you very much for the invaluable suggestions and questions.

Point 1: Please present a discussion concerning the reliability and validity of your research in the paper.

Response 1: In this study, models currently used in many demand forecasting fields were used, and the water demand forecasting were also simulated according to the previous studies, so the reliability of the research is considered to be the same as the previous studies.

However, there are few studies comparing water demand forecasting models using water consumption data for each use collected in SWG living lab, so we think the research is feasible.

Accordingly, we have modified conclusions a little more.

 

Point 2: Add the DOI number to the references.

Response 2: We included DOI in the references.

 

Author Response File: Author Response.pdf