Inappropriate scheduling plans can result in additional economic losses and the safety of water distribution network (WDN). Optimizing manual experience based scheduling plans can help water utilities rationally allocate water plants and pump stations, ensuring the safety, stability, and economy of the water supply system. However, there is a lack of real-time, rational, and optimized scheduling methods. To address this, we proposed a novel intelligent scheduling framework based on deep learning. In this framework, two neural network models, multi-heads convolutional gated recurrent unit network (MH-CGRU) and multi-head gated recurrent unit network (MH-GRU), can effectively extract key features from the WDNs. Operating data were used as decision variables to predict and generate scheduling orders for water plants and pump stations, respectively. The rationality of the orders is verified by combining a high precision online hydraulic model and the evaluation of the operational status of the WDNs. This system has been deployed in a real WDN and put into practical application. From June to November of 2022, the total adoption rate of all orders reached 96.29%, with the average deviation between predicted and actual control targets being less than 5%, and energy consumption decreased by 3.05% compared to the previous year.

  • Proposed an optimized, real-time, and secure intelligent control method for water supply networks based on deep learning algorithms.

  • Presented a data evaluation approach for selecting high-quality samples from monitoring data in the water supply system.

  • Developed an intelligent verification mechanism that combines a high-precision hydraulic model with scheduling orders for improved control reliability.

The scheduling of water distribution networks (WDNs) is an essential component of urban water supply system operation and management for ensuring hydraulic and water quality safety, as well as energy reduction in the WDNs. On the one hand, with urbanization, the structure and scale of the water supply network become more and more complex, and the pump scheduling that relies on traditional manual experience gradually begins to face challenges. On the other hand, 70–80% of energy consumption stems from pumping stations' transmission and distribution (C. U. W. Supply, D. Association 2021), the overall operating efficiency of pumps in WDNs ranges from 50 to 75%, and the energy consumption of water transmission and distribution units is usually above 370 km3/MPa, far exceeding the industry's optimization target value of 350 km3/MPa (C. U. W. S. Association 2005), which means that there is a large space for energy consumption optimization.

In order to improve the efficiency of WDNs, numerous studies have attempted to address this issue. The methods developed in these studies can be roughly divided into two categories: rule-based control and optimization algorithm-based methods (Giustolisi et al. 2013). Rule-based control methods typically involve setting thresholds to trigger the operation of pumps. The simplest rule-based control is called fixed trigger level (FTLs), which involves setting two water levels for controlling the pump's operation. The pump is activated when the water level exceeds the on-trigger level, and it is turned off when the water level falls below the off-trigger level (Alvisi & Franchini 2017). Reduced fixed trigger levels (RFTLs) are improved methods based on FTLs, which distinguish between peak and off-peak periods. The method optimizes the on-trigger level during off-peak periods and the off-trigger level during peak periods. It tries to achieve a beneficial condition by ensuring that the water level in the tank reaches the highest acceptable point at the end of the off-peak period and the lowest acceptable point at the end of the peak period (Alvisi & Franchini 2016; Creaco et al. 2016; Marchi et al. 2016). Compared to FTLs and RFTLs methods, the time-variable triggering method for pump control can provide an optimal arrangement between non-peak and peak power consumption periods. This method primarily calculates the triggering level patterns by solving a multi-objective optimization problem to minimize energy consumption and the number of pump station switches (Housh & Salomons 2019; Quintiliani & Creaco 2019). The advantage of these methods is that they can be used for real-time decision-making; however, such real-time decisions only respond to the current operating state without considering the overall optimality of the network operation over a period of time.

The optimization algorithm-based method usually refers to the method that uses an optimization algorithm to decide whether the pump should be on or off at a certain time. Linear programming (LP) is a classical method that can handle optimization problems with constrained objective functions and is suitable for optimization problems with the decision variables of WDNs (Giacomello et al. 2013; Puleoa et al. 2014); however, because the WDNs do not satisfy the assumption of a linear system, the optimization effect is poor. Nonlinear programming (NLP) takes into consideration the nonlinear hydraulics of the WDNs and can be applied to practical scenarios; however, the computation time grows exponentially with the number of decision variables, limiting its application to large WDNs (El Mouatasim 2012; Skworcow et al. 2014). The mixed-integer nonlinear (MINLP) programming approach is more computationally efficient than the nonlinear planning method for this situation, but the issue of computational time consumption remains (Samani & Zanganeh 2010; Fooladivanda & Taylor 2015; Costa et al. 2016; Khatavkar & Mays 2017). Meta-heuristic algorithms (such as genetic algorithms (Costa et al. 2010; Moreira & Ramos 2013; Odan et al. 2015; Makaremi et al. 2017), ant colony optimization (Hashemi et al. 2014; Babaei et al. 2015), and particle swarm optimization (Rajabpour et al. 2015a, 2015b; Rajabpour & Talebbeydokhti 2020), etc.) have been proven to be capable of handling large-scale problems and can be applied to WDNs in the real world. These methods are applicable to both continuous and discrete variables and can find global parallel optimization. Nevertheless, for real-time scheduling decisions, the following problems remain: first, the model calculation takes a long time, and it cannot achieve high-precision real-time response when applied to large WDNs; second, such algorithms frequently only consider optimality, without considering the safety of scheduling operations and the overall optimality of scheduling.

In recent years, deep learning algorithms have gained increasing attention as data-driven modeling tools for applications such as computer vision (LeCun et al. 2010; Leibe et al. 2010), natural language processing (Young et al. 2018; Otter et al. 2020), and speech recognition (Dahl et al. 2011; Nassif et al. 2019). For the pump scheduling problem in the WDNs, deep learning-related research is still very limited. Studziński & Ziółkowski (2020) propose a method using neural networks instead of hydraulic models, which can accelerate the model solution. There are also many studies applying deep reinforcement learning in pump station scheduling to optimize the reward function by controlling agents interacting with hydraulic models (Xu et al. 2021; Donâncio et al. 2022; Hu et al. 2023). However, a common problem in these studies is that the optimality and reliability of the scheduling orders depend on the simulation accuracy of the hydraulic model. Different from the above studies, this study automatically extracts temporal and spatial features from real water distribution network (WDN) monitoring data based on deep learning and establishes a mapping relationship between the operational status of the WDN and the scheduling orders.

In general, the main obstacles to the application of optimization and scheduling technologies for the WDNs are the inability to balance real-time decision-making, seek optimization capabilities, and ensure water supply safety. To address these technical bottlenecks, this paper innovatively implements a real-time intelligent scheduling system based on deep learning algorithms and applies it to a large-scale water supply network in Shanghai, with the following features:

  • Real-time capability: The system captures monitoring data from the water supply system and generates scheduling instructions through a pre-trained deep learning scheduling model at a frequency of every 5 min.

  • Optimality: The training data are optimized for model performance using various data evaluation indicators, and the results demonstrate improved energy-saving levels when compared to historical energy consumption.

  • Safety: The system integrates a high-precision hydraulic network model to assist in scheduling decisions and filters out unreasonable orders.

Framework description

This section provides an overview of the intelligent scheduling system framework, which mainly consists of two parts: the offline training process and the online scheduling process. The overall framework structure can be summarized in Figure 1.
Figure 1

Intelligent scheduling system structure diagram.

Figure 1

Intelligent scheduling system structure diagram.

Close modal

The first phase is the offline training process, which includes historical data collection, data preprocessing, data optimization, and ultimately generating a training dataset to train water plant pressure prediction models, pump station pressure prediction models, and water level prediction models.

The second phase is the online scheduling process, which involves receiving real-time monitoring data of the WDNs and inputting it into the deep learning model to obtain future water plant pressure, pump station pressure, and water level predictions. Then, these prediction results are used to trigger corresponding scheduling orders. The system then combines these orders with a high-precision hydraulic model simulation to validate the reasonableness and optimization of the orders. Finally, the orders are pushed to the operator for manual decision-making.

Data and case description

Water distribution system description

The study area is in the eastern portion of Shanghai's Qingpu District, as shown in Figure 2. This region is served by one water plant with a daily capacity of 200,000 m3. There are nine pumps in the water plant, three of which are variable speed pumps, and the rest are fixed speed pumps, with a single pump flow of 1,670 m3/h, a head of 39 m, and a power of 355 kW.
Figure 2

Schematic diagram of the water distribution network in the study area.

Figure 2

Schematic diagram of the water distribution network in the study area.

Close modal

In addition, there are two pump stations in the area, XinFengXi and HuaXiang (hereinafter referred to as HX and XFX). The reservoir capacity of HX reservoir is 15,000 m3, and there are five pumps in the pumping station. The reservoir capacity of XFX pump station is 7,500 m3, and there are five pumps in the pump station. In terms of SCADA monitoring data, many SCADA monitoring points are arranged in the area. The types of monitoring points are mainly divided into two categories: flow monitoring points and pressure monitoring points. There are 24 flow meters in the area, which record the pressure and flow data of the main pipeline with the diameter of DN 500–1,000; 27 pressure monitoring points, which record the flow and pressure data of the pressure measurement nodes.

Data description and preprocessing

In this study, a supervisory control and data acquisition (SCADA) monitoring system with a time resolution of 5 min was employed to collect the operation status of the WDNs in the study area between 2019 and 2021. The dataset comprises three parts: water plant data, pump station data, and monitoring point data. The water plant data include the outflow data and the pressure data; pump station data encompass inflow and outflow, pump station pressure, and tank water levels; while monitoring point data primarily consist of pressure data at the monitoring points and flow data of the main pipelines of the WDNs. To ensure the quality of the dataset, the smoothing Z-score algorithm was used to detect outliers in the data, as shown in the following equation.
formula
(1)
where x is the point value, and are the mean and standard deviations of the historical window data. In this study, the duration of the length of window was set at 10,000 and the threshold was set at 3.5. Additionally, the time weighted k-nearest neighbors regression (TW-KNNR) algorithm was applied to impute missing values in the raw data, the specific formula is shown in the following equation.
formula
(2)
where is the interpolation value, represents the observation point with the same period as the missing values, and represents the time distance between and missing value.
Lastly, in order to eliminate the influence of the input data scale, this study employed the Min–Max normalization method to scale the dataset to the range [0, 1], as illustrated in the following equation.
formula
(3)
where is the normalized data, x represents the original data, is the minimum value of the data, and is the maximum value of the data.

Data evaluation and training dataset construction

In deep learning algorithms, the model's understanding of the system properties of the learning object is directly related to the selection of the training dataset (Shokri & Shmatikov 2015). In the historical monitoring data of WDNs, differences in experience among operators can also lead to variations in the scheduling performance of the WDNs. To extract excellent scheduling experience from historical monitoring data, it is crucial to evaluate the quality of the WDNs' operation status. Therefore, this study proposes a data evaluation method, including pressure satisfaction and historical state similarity. This method evaluates the WDNs operation status from the perspectives of safety, rationality, and optimization, and selects data samples representing excellent scheduling experience as the training dataset.

The pressure satisfaction is used to filter data samples from historical data that meet the water supply pressure requirements and yet have lower energy consumption. This method uses an exponential decay function, which implies that samples with higher scores will have a water supply pressure closer to the minimum requirement, and correspondingly, their energy consumption will be lower.
formula
(4)
where represents the pressure at time t, represents the minimum pressure requirement, which is set to 165 kPa for the peak hours (6:00–10:00 a.m. and 16:00–19:00 p.m.) and 180 kPa for the off-peak hours, and C is a constant controlling the rate of decay, and is usually set to the standard deviation of the data distribution in order to ensure that scores between different samples are discriminatory and that a sufficient number of samples can have high scores, here taken as 5.
The historical state similarity is mainly used to evaluate the reliability of the current control conditions based on the similarity between the current WDNs operation status and the historical WDNs operation status, the calculation method is as follows:
formula
(5)

For each moment, the operation status of the WDNs St is represented as a vector that includes water plant pressure and flow (WPPt WPFt), pump station pressure and flow (), and monitoring point pressure (). Discretize each component into n state intervals using f, where represents the indicator function and calculates the proportion of historical data with the same operating status as the current WDNs in the total historical data. Then, normalize the values to a range between 0 and 1.

After obtaining the above scores, the final score S is given according to the following weighted average method.
formula
(6)
where and are the weights, and the appropriate values need to be determined experimentally.

Deep learning model for scheduling

The working principle of the deep learning-based WDNs scheduling model is based on model predictive control (MPC), in which the deep learning model predicts the future changes in the scheduling decision variables and generates corresponding orders based on the predictions (Camacho & Bordons 2007; Katz et al. 2020). In the specific application scenario of this research, the main control variables of the WDNs include water plant pressure, pump station pressure, and tank water level. Firstly, for the two different scheduling facilities (water plants and pump stations), two neural network structures, MH-CGRU and MH-GRU, are designed separately for predicting the target control variables.

Additionally, thanks to the proportional integral derivative (PID) negative feedback control device (Johnson & Moradi 2005), water plants can directly adjust the operation frequency of water pumps by setting the control pressure; for pump stations, there are mainly two types of pumps: booster-type pumps and reservoir-type pumps. Booster-type pumps work by increasing the water pressure flowing through them, directly improving the water pressure in the connected pipes, and their on–off status can be reflected by the changes in the outlet pressure. Reservoir-type pumps operate by transporting water from a reservoir to the WDNs, and their on-off status can be reflected by changes in the reservoir water level. Therefore, to obtain the control orders for the pump station, it is necessary to first predict the future outlet pressure and reservoir water level that need to be achieved, and then map the prediction results into specific control orders.

Prediction model for water plant operation

When predicting water plant pressure, it is necessary to consider the overall operation status of the WDNs and extract temporal and spatial features from it. For water plant pressure prediction, a network structure as shown in Figure 3(a) is adopted, which includes a one-dimensional convolution layer (Conv1D) (Chen et al. 2022) and a gated recurrent unit network layer (GRU) (Chung et al. 2014).
Figure 3

Flow chart of the structure of the neural network: (a) Multi-head CGRU model for water plant pressure prediction and (b) multi-head GRU model for pump station pressure and water level prediction.

Figure 3

Flow chart of the structure of the neural network: (a) Multi-head CGRU model for water plant pressure prediction and (b) multi-head GRU model for pump station pressure and water level prediction.

Close modal

In the proposed MH-CGRU neural network, a multi-head structure with different model inputs is employed. Pump station outlet pressure, main pipeline pressure, and monitoring point pressure in the past 3 h are fed into head 1 to perceive the current environmental state, while pump station outlet pressure, main pipeline pressure, monitoring point pressure, water plant outflow, and water plant pressure in the past 7 days' concurrent data are fed into head 2 for learning the periodic variation patterns of the WDN operation state. In each head, two Conv1D layers are designed to extract spatial correlations among the monitoring data, and a GRU layer is developed to model long-term temporal dependencies. Ultimately, the encoded features are concatenated and input into a fully connected layer, generating the target control pressure for the water plant in the next hour.

Prediction model for water pump station

Pump stations are typically used for local pressurization to ensure the water supply pressure at the most disadvantageous consumption node (MDCN). Therefore, when predicting the pump station outlet pressure and tank water level, the relationship between the pressure at the MDCN and the pump station's attributes (such as tank water level, outlet pressure, and flow) needs to be considered. In this section, a network called MH-GRU that takes these factors into account is designed for predicting the pump station outlet pressure and water level, with the network structure illustrated in Figure 3(b).

Similar to MH-CGRU, the input of MH-GRU is also divided into two parts: the past 3 h of input data is used to perceive the current network operation state, while the data from the same period 7 days before are used for learning the periodic patterns of the prediction targets. However, the MH-GRU network does not include Conv1D blocks. This is attributable to the fact that for water plants that need to supply water to the entire region, the model must consider the spatial correlations of global input features, while for pump stations mainly used for local pressurization, the pressure of the MDCN and the station's own state are the primary considerations.

After obtaining the predicted values for the pump station outlet pressure and tank water level , there needs to be a method to convert these predictions into control signals. Therefore, this study implements a threshold-triggered method to convert the predicted pump station pressure and tank water level into pump station control orders. The main principle of this triggering method is to set specific trigger thresholds for different control scenarios, and jointly consider the predicted variable value and the change in the predicted variable (i.e., ). This is because capturing the timing of pump station switching requires considering both the current state of the predicted variable and its transient changes.

Figure 4(a) and 4(b) illustrates the specific threshold control process for a booster-type pump: The pump-on order is triggered when the predicted pressure simultaneously meets both the and threshold trigger conditions. On the other hand, the pump-off command is triggered when the predicted pressure simultaneously meets the and control conditions, as summarized in the following equation.
formula
(7)
where is booster-type pump control signal in the next 5-min (0: pump-off, 1: pump-on), and is the specific threshold for pump-on triggering, and are the specific threshold for pump-off triggering, and are the pump station outlet pressure prediction and its differential value, respectively.
Figure 4

Logic diagram of the booster-type pump start/stop order trigger: (a) absolute predicted pressure change with marked orders trigger conditions; (b) predicted pressure differential change with marked orders trigger conditions.

Figure 4

Logic diagram of the booster-type pump start/stop order trigger: (a) absolute predicted pressure change with marked orders trigger conditions; (b) predicted pressure differential change with marked orders trigger conditions.

Close modal
Figure 5(a) and 5(b) illustrates the threshold triggering process for reservoir-type pumps, which differs from pressure-boosting pumps in that they also include water-filling operations through valves. Specifically, when the predicted reservoir water level meets conditions and , the inflow valve opens; when the predicted reservoir water level meets conditions and , the reservoir-type pump starts, and under other conditions, the reservoir-type pump is turned off.
formula
(8)
where is the reservoir-type pump control signal (0: pump-off and valve-off, 1: pump-on, −1: valve-on); and are the specific threshold for pump-on triggering, and is the specific threshold for valve-on triggering, and are the specific threshold for valve-off triggering, and are the specific threshold for pump-off triggering, and are the tank water level prediction and its differential value, respectively.
Figure 5

Logic diagram of the reservoir-type pump start/stop order trigger: (a) absolute predicted water level change with marked orders trigger conditions; (b) predicted water level differential change with marked orders trigger conditions.

Figure 5

Logic diagram of the reservoir-type pump start/stop order trigger: (a) absolute predicted water level change with marked orders trigger conditions; (b) predicted water level differential change with marked orders trigger conditions.

Close modal

Model validation

In the practical application of intelligent scheduling systems for WDNs, it is crucial to ensure the rationality of control orders from a programming perspective. To evaluate control orders without actual execution, hydraulic simulation is clearly the best choice. Therefore, in this section, an evaluation method for scheduling schemes based on hydraulic model simulation is proposed and combined with the evaluation indicators of the operating status of the WDNs, the rationality of orders is assessed and automatically filtered through the changes before and after simulation.

Real-time hydraulic modeling of the WDNs

Firstly, a hydraulic model was established based on EPANET 2.0, shown in Figure 2 (Rossman 2000). The hydraulic model, after calibration, has a mean absolute error (MAE) of less than 2 m for the pressure of the scheduled facility (water plant and pump station) and pressure monitoring points, and a relative error of less than 10% for the scheduled facility flow rate. The specific errors can be found in Appendix A.

An online hydraulic model calculation interface was then developed based on WNTR. This calculation interface accepts real-time water monitoring data of the WDNs and control orders generated by the neural network model as inputs and performs a single-step simulation to obtain the operating status of the WDNs after the execution of the control orders, and the specific input and output items are shown in Table A.5 and Table A.6.

In this way, for each order generated by the neural network, the program automatically collects the required real-time monitoring data of the WDNs and returns the corresponding operating status of the WDNs after execution. It is also worth noting that in order to minimize inaccuracies in the delayed simulation of the hydraulic model, the results returned for any output variable can be adjusted as follows:
formula
(9)

In this formula, represents the observed data at time t, while and represent the simulated data at times t and , respectively. This is because the error in the simulation result comes from the simulation error of and the simulation error of , here the simulation error of is eliminated by replacing the simulated a1 value with .

Orders filtering

For practical scheduling operations, it is essential to verify the rationality of scheduling orders before executing them. In Section 2.3, a data evaluation method has been proposed to assess the safety, rationality, and optimality of the operation status of the WDNs. A simple intuition is that we hope that after executing the scheduling orders predicted by the neural network, the operation status of the WDNs can show better performance. Therefore, the gain of the scheduling orders for the WDNs can be evaluated by comparing the scores of the WDNs' operation status before and after the execution of the orders. However, one problem is that the model is unable to know the status of the system after the execution of a scheduling order before it is executed. Thanks to the implementation of the high-precision real-time hydraulic model interface in Section 2.5.1, the operation status of the WDNs before and after the execution of the orders can be accurately simulated (the calculation accuracy of the real-time hydraulic model interface has been verified by historical data testing, as shown in Appendix B for details). Suppose the score of WDNs' operation status before the execution of the orders is , and the score after the execution is , with the allowed maximum score drop being . If the score of the WDNs' operation status drops more than the allowed maximum score drop after the execution of the orders, the orders will be filtered out, as shown in the following equation.
formula
(10)
where the maximum allowed score drop is set to 10.

Variable frequency control

Due to the presence of variable frequency pumps in the pumping station, in order to more accurately optimize the scheduling performance, it is not only necessary to predict the on/off status of each pump but also to set the optimal frequency for each variable frequency pump. In this study, the intelligent scheduling system uses the neural network model to predict the pump switching signals and combines the online hydraulic model to simulate the operation status of the WDNs after the order execution under different frequency conditions. Finally, based on the evaluation of the operation status of the WDNs, the frequency with the highest score is selected.

The specific optimization strategy can be shown in Figure 6. It considers opening the pump when the pressure at the MDCN is 168 kPa. By simulating the operation status of the WDNs after the execution of the order at 35, 40, 45, and 50 Hz, it can be concluded that the best operating performance is achieved at 45 Hz. Therefore, the order to improve the frequency to 45 Hz is executed. Similarly, considering closing the pump when the pressure of MDCN is at 191 kPa, it can be concluded that the best operating performance is achieved at 35 Hz. Therefore, the order to decrease the frequency to 35 Hz is executed.
Figure 6

By using the online hydraulic model to simulate the expected pressure changes in the MDCN and the operation status score of the WDNs after execution under different frequency settings: (a) pump-on and (b) pump-off.

Figure 6

By using the online hydraulic model to simulate the expected pressure changes in the MDCN and the operation status score of the WDNs after execution under different frequency settings: (a) pump-on and (b) pump-off.

Close modal

Apart from the pump's switching signals, during the scheduling operation, the order trigger of the variable frequency pump is subject to negative feedback control from the pressure at the MDCN When it exceeds or falls below , the aforementioned operation is repeated to simulate the operation state of the WDN under different frequencies, thereby generating the optimal control frequency.

From the aforementioned, it is clear that the frequency control method relies on the real-time calculation of the intelligent scheduling model, the multi-conditions hydraulic simulation of the online hydraulic model, and the evaluation of the operation status, which can further optimize the scheduling performance based on the generation of pump-on/off signals and is a crucial component of pump station scheduling. It may enrich the control methods of the intelligent scheduling system under various operating conditions and make the pressure regulation of pipeline networks more stable and reasonable.

Optimization of data evaluation weights

In consideration of the data evaluation method mentioned in section 2.3, one of the challenges is to find an optimal combination of weights to calculate scores and filter data, such that the model trained has the best performance on the validation dataset. To address this, we designed a set of experiments using four combinations of weights to filter the data. Then, we tested the performance of the trained model on the validation dataset (from 23 May 2021 to 31 December 2021), and the specific results are shown below. It should be noted that from 20 July 2022 to 11 August 2022, and 14 September 2022 to 25 October 2022, the system was temporarily suspended and switched to manual scheduling due to pump cutting and saltwater intrusion, respectively.

From Table 1, it can be observed that the performance of the model improves as the weight of pressure satisfaction increases. The best performance is achieved with the weight combination of (0.6, 0.4), which results in the lowest RMSE, MAE, and MAPE. This suggests that the pressure satisfaction plays a significant role in filtering suitable data samples for training the model.

Table 1

Performance of different combinations of data evaluation weights on the validation dataset (23 May 2021–31 December 2021)

Weight combination ()RMSEMAEMAPE
(0.2, 0.8) 19.10 17.38 7.58 
(0.4, 0.6) 14.96 11.04 4.79 
(0.6, 0.4) 10.45 7.92 3.32 
(0.8, 0.8) 10.49 8.06 3.37 
Weight combination ()RMSEMAEMAPE
(0.2, 0.8) 19.10 17.38 7.58 
(0.4, 0.6) 14.96 11.04 4.79 
(0.6, 0.4) 10.45 7.92 3.32 
(0.8, 0.8) 10.49 8.06 3.37 

Scheduling system operation performance

In order to accurately evaluate the actual performance of the intelligent scheduling system, we continuously collected scheduling order data from June 2022 to November 2022 and analyzed the adoption rate, evaluation scores, and energy consumption per unit.

Order adoption rate

The order adoption rate is the ratio of the actual number of executed orders to the number of orders pushed to the operator by the intelligent scheduling system. A higher adoption rate means that professionals have a higher appreciation of the reasonability and reliability of the orders generated by the intelligent scheduling system. The data of the order adoption rate from June 2022 until November 2022 have been summarized in Table 2.

Table 2

Order adoption rate statistics during system operation from June 2022 to November 2022

Number of orders generatedNumber of orders executedOrders adoption rate (%)
Water plant 224 223 99.55 
Pump station 1,446 1,385 95.78 
Total 1,670 1,608 96.29 
Number of orders generatedNumber of orders executedOrders adoption rate (%)
Water plant 224 223 99.55 
Pump station 1,446 1,385 95.78 
Total 1,670 1,608 96.29 

The result shows that the total adoption rate of all the orders generated by the intelligent scheduling system is 96.29%, including 99.55% for water plant orders and 95.29% for pumping station orders, which indicates that the orders generated by the intelligent scheduling system are reasonable and can meet the water supply demand of the service area most of the time. The very few scheduling orders that were not adopted were mainly false triggers due to unstable monitoring data. The control algorithm should be further modified in the future to avoid the influence of monitoring noise on scheduling orders.

Model evaluation score

As introduced in Appendix B, the evaluation score of the scheduling orders is obtained by assessing the WDN operation status after executing the scheduling orders, thereby reflecting the effects of the scheduling orders. It includes four indicators: pressure compliance rate, energy consumption reduction rate, pressure fluctuation rate, and order accuracy, comprehensively considering the rationality, optimization, stability, and accuracy of the scheduling orders. During the actual operation of the scheduling system, the WDNs operation status data are collected for calculating the evaluation score, and its specific data distribution is shown in Figure 7.
Figure 7

Model evaluation scores during system operation (June 2022–November 2022). The color of the points represents the evaluation level of the corresponding scheduling orders, where the gray points represent the evaluation of manual scheduling orders during the downtime of the intelligent scheduling system.

Figure 7

Model evaluation scores during system operation (June 2022–November 2022). The color of the points represents the evaluation level of the corresponding scheduling orders, where the gray points represent the evaluation of manual scheduling orders during the downtime of the intelligent scheduling system.

Close modal

According to the statistics of the evaluation scores for the scheduling orders, 89% of the scores after executing the scheduling orders are greater than 75, which are considered ‘excellent’ level. It can be found that the low-scoring periods of scheduling scoring orders are mainly concentrated in the period when the intelligent scheduling system is shut down, while the scheduling scores are higher in other periods, which indicates that compared with manual scheduling, the scheduling orders generated by the intelligent scheduling system have a better performance in terms of rationality, optimization, stability, and accuracy.

Unit energy consumption

The level of energy consumption reduction is an important indicator for evaluating whether the orders of an intelligent scheduling system perform better than manual scheduling. Therefore, in this study, the unit energy consumption data from June 2022 to November 2022 was compared with the same period of the previous year. The specific results are shown in Figure 8.
Figure 8

Unit energy consumption data comparison during system operation (June 2022–November 2022).

Figure 8

Unit energy consumption data comparison during system operation (June 2022–November 2022).

Close modal

In Figure 8, the dark blue line represents the change in unit energy consumption data in 2022, the light blue line represents the change in unit energy consumption data in 2021, and the portion checked in red represents the period when the intelligent scheduling system is out of operation due to water system maintenance and the impact of salty tides. The results show that the unit energy consumption in 2021 is higher than the unit energy consumption in 2022 during the time when the intelligent scheduling system is in operation, and the overall unit energy consumption reduces by 3.05% on average compared to the historical period. Meanwhile, comparing the energy consumption data in 2022, it can be found that the energy consumption of the WDNs increases noticeably when the intelligent scheduling system is switched to manual scheduling, which indicates that the scheduling orders generated by the intelligent scheduling system can effectively reduce the energy consumption of the WDNs compared with manual scheduling.

This paper proposes a novel real-time intelligent scheduling system for the large-scale WDNs based on deep learning models. The workflow of the real-time intelligent scheduling system includes: (1) screening samples from historical data through the data evaluation method; (2) constructing two novel deep learning neural networks(MH-CGRU and MH-GRU) to generate scheduling orders by constructing the mapping relationship between the operating status of the WDNs and the decision-making variables of the pump station; and (3) combining with the hydraulic model to further simulate the execution performance of the scheduling orders to ensure its reasonableness.

The proposed framework is successfully applied to the real-time scheduling in the Qingdong region of Shanghai, and more than 97% of the system-generated scheduling orders are adopted by manual operators during the actual operation period, which demonstrates the reasonableness of the generated orders. In addition, the framework has proven to have good tracking performance, with a tracking error of less than 3%. More importantly, the energy consumption of the water supply system was reduced compared to the manual scheduling data of the previous year, which demonstrates the optimization of the intelligent scheduling system.

In future work, reinforcement learning algorithms can be further incorporated to construct intelligent agents that fully interact with the hydraulic model and design reward functions in conjunction with the existing scheduling evaluation mechanism to expand the application of the scheduling system under various abnormal conditions (e.g., pumps shutdown and water supply accidents).

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.

Alvisi
S.
&
Franchini
M.
2016
A methodology for pumping control based on time variable trigger levels
.
Procedia Engineering
162
,
365
372
.
Alvisi
S.
&
Franchini
M.
2017
A robust approach based on time variable trigger levels for pump control
.
Journal of Hydroinformatics
19
(
6
),
811
822
.
C. U. W. S. Association
.
2005
Development Plan for Technological Progress of Urban Water Supply Industry in 2010 and Visionary Goals for 2020
.
China Architecture & Building Press, Beijing
.
C. U. W. Supply, D. Association
.
2021
Urban Water Industry Development Plan Outline for 2035
.
China Architecture & Building Press, Beijing
.
Camacho
E. F.
&
Bordons
C.
2007
Model predictive controllers
. In:
Model Predictive Control
(Camacho, E. F. & Bordons, C.).
Springer
,
London
, pp.
13
30
.
Chen
L.
,
Yan
H.
,
Yan
J.
,
Wang
J.
,
Tao
T.
,
Xin
K.
,
Li
S.
,
Pu
Z.
&
Qiu
J.
2022
Short-term water demand forecast based on automatic feature extraction by one-dimensional convolution
.
Journal of Hydrology
606
,
127440
.
Chung
J.
,
Gulcehre
C.
,
Cho
K.
&
Bengio
Y.
2014
Empirical evaluation of gated recurrent neural networks on sequence modeling, arXiv preprint arXiv:1412.3555
.
Costa
L. H. M.
,
de Castro
M. A. H.
&
Ramos
H. M.
2010
Use of hybrid genetic algorithms for optimized operation of water supply systems
.
Engenharia Sanitaria E Ambiental
15
,
187
196
.
Costa
L. H. M.
,
de Athayde Prata
B.
,
Ramos
H. M.
&
de Castro
M. A. H.
2016
A branch-and-bound algorithm for optimal pump scheduling in water distribution networks
.
Water Resources Management
30
,
1037
1052
.
Creaco
E.
,
Lanfranchi
E.
,
Chiesa
C.
,
Fantozzi
M.
,
Carrettini
C.
&
Franchini
M.
2016
Optimisation of leakage and energy in the Abbiategrasso district
.
Civil Engineering and Environmental Systems
33
(
1
),
22
34
.
Dahl
G.
,
Yu
D.
,
Deng
L.
&
Acero
A.
2011
Context-dependent DBN-HMMs in large vocabulary continuous speech recognition
. In:
Proc. ICASSP
.
Donâncio
H.
,
Vercouter
L.
&
Roclawski
H.
2022
The pump scheduling problem: A real-world scenario for reinforcement learning, arXiv preprint arXiv:2210.11111
.
El Mouatasim
A.
2012
Boolean integer nonlinear programming for water multireservoir operation
.
Journal of Water Resources Planning and Management
138
(
2
),
176
181
.
Fooladivanda
D.
&
Taylor
J. A.
2015
Optimal pump scheduling and water flow in water distribution networks
. In:
2015 54th IEEE Conference on Decision and Control (CDC)
. pp.
5265
5271
.
Giacomello
C.
,
Kapelan
Z.
&
Nicolini
M.
2013
Fast hybrid optimization method for effective pump scheduling
.
Journal of Water Resources Planning and Management
139
,
175
183
.
Giustolisi
O.
,
Laucelli
D.
&
Berardi
L.
2013
Operational optimization: Water losses versus energy costs
.
Journal of Hydraulic Engineering
139
(
4
),
410
423
.
American Society of Civil Engineers
.
Housh
M.
&
Salomons
E.
2019
Optimal dynamic pump triggers for cost saving and robust water distribution system operations
.
Journal of Water Resources Planning and Management
145
(
2
),
04018095
.
Johnson
M. A.
&
Moradi
M. H.
2005
PID Control
.
Springer, London
.
Katz
J.
,
Pappas
I.
,
Avraamidou
S.
&
Pistikopoulos
E. N.
2020
Integrating deep learning and explicit MPC for advanced process control
. In:
2020 American Control Conference (ACC)
.
IEEE
, pp.
3559
3564
.
Khatavkar
P.
&
Mays
L. W.
2017
Model for optimal operation of water distribution pumps with uncertain demand patterns
.
Water Resources Management
31
,
3867
3880
.
LeCun
Y.
,
Kavukcuoglu
K.
&
Farabet
C.
2010
Convolutional networks and applications in vision
. In:
Proceedings of 2010 IEEE International Symposium on Circuits and Systems
.
IEEE
, pp.
253
256
.
Leibe
B.
,
Mitzel
D.
,
Weyand
T.
&
Floros
G.
2010
Current Topics in Computer Vision and Machine Learning
.
Marchi
A.
,
Simpson
A. R.
&
Lambert
M. F.
2016
Optimization of pump operation using rule-based controls in EPANET2: New ETTAR toolkit and correction of energy computation
.
Journal of Water Resources Planning and Management
142
(
7
),
04016012
.
Moreira
D.
&
Ramos
H. M.
2013
Energy cost optimization in a water supply system case study
.
Journal of Energy
2013
,
1
9
.
Nassif
A. B.
,
Shahin
I.
,
Attili
I.
,
Azzeh
M.
&
Shaalan
K.
2019
Speech recognition using deep neural networks: A systematic review
.
IEEE Access
7
,
19143
19165
.
Odan
F. K.
,
Reis
L. F. R.
&
Kapelan
Z.
2015
Real-time multiobjective optimization of operation of water supply systems
.
Journal of Water Resources Planning and Management
141
,
04015011
04015011
.
Otter
D. W.
,
Medina
J. R.
&
Kalita
J. K.
2020
A survey of the usages of deep learning for natural language processing
.
IEEE Transactions on Neural Networks and Learning Systems
32
(
2
),
604
624
.
Puleoa
V.
,
Morleyb
M. S.
,
Frenic
G.
&
Savi
D. A.
2014
Multi-Stage Linear Programming Optimization for Pump Scheduling
. Elsevier, London.
Rajabpour
R.
&
Talebbeydokhti
N.
2020
Developing a new algorithm (G-JPSO) for optimal control of pumps in water distribution networks
. Scientia Iranica 27, 68–79.
Rajabpour
R.
,
Talebbeydokhti
N.
&
Ahmadi
M.
2015a
Using jumping particle swarm optimization for optimal operation of pump in water distribution networks
.
Journal of Civil and Environmental Engineering
9
,
1200
1204
.
Rajabpour
R.
,
Talebbeydokhti
N.
&
Ahmadi
M.
2015b
Developing new algorithm and its application on optimal control of pumps in water distribution network
.
Journal of Civil and Environmental Engineering
9
,
1238
1241
.
Rossman
L. A.
2000
Epanet 2: Users Manual
. USEPA, Cincinnati, OH. https://nepis.epa.gov/Adobe/PDF/P1007WWU.pdf
Samani
H. M.
&
Zanganeh
A.
2010
Optimisation of water networks using linear programming
. In:
Proceedings of the Institution of Civil Engineers-Water Management
, Vol.
163
(Vali, H. M. & Zanganeh, A.).
Thomas Telford Ltd, Chichester
, pp.
475
485
.
Shokri
R.
&
Shmatikov
V.
2015
Privacy-preserving deep learning
. In:
Proceedings of the 22nd ACM SIGSAC Conference on Computer and Communications Security
. pp.
1310
1321
.
Skworcow
P.
,
Paluszczyszyn
D.
&
Ulanicki
B.
2014
Pump schedules optimisation with pressure aspects in complex large-scale water distribution systems
.
Drinking Water Engineering and Science
7
(
1
),
53
62
.
Young
T.
,
Hazarika
D.
,
Poria
S.
&
Cambria
E.
2018
Recent trends in deep learning based natural language processing
.
IEEE Computational Intelligence Magazine
13
(
3
),
55
75
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).

Supplementary data