A computer-implemented method for modeling hydraulic networks

Abstract

A computer-implemented method for modeling a hydraulic network including pipelines and pressure monitoring points, the method comprising: creating a model including a plurality of nodes and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network, and each edge connects two nodes and represents a combination of pipelines of the network spanning between the pressure monitoring points represented by the two nodes.

Description

Technical Field

[0001] The invention relates to a method of modeling a hydraulic network and to a hydraulic network model. Background Art

[0002] Water utilities use hydraulic network models to understand the behavior of the utility's water distribution network and to perform forecasting work. Common examples of uses include management of large events (such as a rupture that causes a consumer to lose supply), design work for new users of the network, and planning new ways for the network to be configured or controlled for greater efficiency.

[0003] Typically, the network model is very detailed and incorporates a large number of pipes across most streets in urban environments and rural areas, as well as the backbone of the network (i.e., large pipes from reservoirs and other water sources). The model is represented as a mathematical graph containing nodes and edges with various parameters. These parameters typically include water consumption data, residential and commercial characteristics, control information, altitude, and the length, diameter, and roughness of the pipes.

[0004] Determining such parameters, known as model calibration, is an expensive, time-consuming and often manual activity. As a result, water utilities are unable to frequently recalibrate their conventional network models, resulting in network models that are often out-of-date and inaccurate. For example, it is not uncommon for conventional network models to be up to 10 years old. Therefore, there is a need for methods of modeling hydraulic networks that are more dynamic so that the resulting model can be more easily calibrated.

[0005] A known approach is to construct a lower resolution network model from a full network model such as the conventional network model previously described. This is done by applying a simplification process to the full network model. By simplifying the full network model, the number of hydraulic equations representing the model will be reduced and simpler methods can then be applied to calibrate the network. However, the need to have a full network model at the start of the analysis leads to a number of disadvantages, such as the basic requirement for a full network model. This is problematic because not all global water utilities have a full hydraulic network model for the water utility's system, or if the water utility has a full hydraulic network model, such a model may be of low quality or difficult to implement.

[0006] Furthermore, simulation of hydraulic networks is important because the behavior of hydraulic networks changes from day to day. Water utilities currently use strategically placed monitoring elements to obtain a real-time view of the behavior of the water utility's network. A typical configuration is to establish discrete areas in the network, sometimes referred to as District Metering Areas (DMAs), and to install flow meters at the entrances to such areas, so that the water utility can understand the needs of a particular area, such as water consumption and leakage, on a daily basis. However, a disadvantage of this known method is that the real-time monitoring of the network is at a coarser level than the conventional hydraulic network models previously described. For example, a defined area typically contains up to 5,000 features at any location. Therefore, there is a need for a method of modeling a hydraulic network that provides a compromise between the resolution of the model and the dynamics of the model.

[0007] Simultaneously, simulations using full hydraulic network model simulation software are limited by their reliance on the full network model. For example, such simulations are limited by the ability to update parameters in the full network model due to the large number of parameters involved in the full network model and the limited number of sensory points within the network.

[0008] A known alternative for determining the current state within a water network and alerting water utilities to anomalies is noise logging. Noise logging is an acoustic noise measurement technique in which the acoustic noise generated by the fluid in the network is recorded by a noise logging tool. Noise logging can help identify and locate new leaks and breaches based on the noise results. However, noise logging has several disadvantages, including:

[0009] Many noise recorders are needed. That is, noise recorders need to be arranged very densely to locate new leaks. This increases the complexity and cost of the network.

[0010] Noise recorders do not provide an estimate of the size of a leak, i.e. noisy leaks are not correlated with large leaks. This means that noise recorders only provide a simple indication that a leak has occurred; they do not provide detailed information about the leak.

[0011] • Recording the leakage by means of a noise recorder does not directly contribute to the calibration of the network model, since the size of the rupture cannot be estimated.

[0012] It would therefore be desirable to provide a method of modelling a hydraulic network which allows for an improved simulation of the network, allowing for an improved detection of, for example, leaks.

[0013] Finally, pressure regulation and control in hydraulic networks is another aspect that is important for the management of hydraulic networks. It is known that methods of minimizing pressure in a network are performed by controlling valves at critical points (CP) in the network (CP is the point in the network with the lowest pressure) and performing pressure monitoring by building a statistical relationship based on pressure monitoring and flow measurements at the valves to control how the valves need to react (but without building a hydraulic model). As a result, this method has no knowledge of the resistance or demand within the network and a single CP problem must be solved multiple times in order to control multiple points in the area. For completeness, resistance in a network is a parameter related to the amount of energy lost in the water flowing between two points in the network.

[0014] Hope these issues can be resolved. Summary of the invention

[0015] Various aspects of the invention are set out in the independent claims. Optional features of the various aspects are set out in the dependent claims.

[0016] According to one aspect, a method for modeling a hydraulic network is provided, the hydraulic network including pipelines and pressure monitoring points, and the method includes: creating a model including a plurality of nodes and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network, and wherein each edge connects two nodes and represents a combination of pipelines of the network spanning between the pressure monitoring points represented by the two nodes.

[0017] The method may be computer-implementable and may therefore be computer-implemented.

[0018] Advantageously, this results in a compact model of the hydraulic network. Such a compact model places a lighter burden on computing resources by: (a) taking up less memory space when stored, and (b) reducing processor burden. The processor burden is reduced because, due to the compact size of the model, simulation results of a hydraulic network using the model are less processor intensive than, for example, simulation results involving a full network model. As a result, additional technical effects include: reduced computation time; the ability to simulate the hydraulic network on smaller computer systems and / or lower powered computer systems; and the ability to simulate large networks.

[0019] The number of pipes in the hydraulic network can be significantly greater than the number of edges in the model. The number of pressure monitoring points can be equal to the number of nodes in the model. The number of pressure monitoring points can be greater than the number of nodes in the model.

[0020] The pressure monitoring point may be a real-time pressure monitoring point. A device arranged to monitor the fluid pressure in the hydraulic network may be arranged at each of the real-time pressure monitoring points.

[0021] The method may include calculating one or more parameters based on the model. The one or more parameters may be a demand at a node and / or a resistance for an edge. The method may include automatically calculating the one or more parameters.

[0022] The method may include using the model to calculate demands at nodes to identify ruptures at or near pressure monitoring points represented by the nodes.

[0023] The method may include using the model to calculate resistance for the edge in order to identify a blockage in the conduit represented by the edge.

[0024] Calculating the one or more parameters based on the model may include: performing regularization on the model so that the one or more parameters can be calculated. Regularization may be performed on the model when mathematical unknowns exist in the model.

[0025] The method may include calculating a demand at a node. The demand may be a time-dependent parameter. The demand may represent a combined water consumption and / or leakage at or near a pressure monitoring point represented by the node. The method may include calculating the demand at a plurality of nodes in the model. The method may include calculating the demand at each of the nodes in the model.

[0026] The method may include calculating a resistance for one or more of the edges. For each edge, the resistance may be related to the amount of energy lost by a fluid such as water in flowing between pressure monitoring points represented by nodes at the ends of the edge, for example, the resistance may indicate the amount of energy lost by a fluid such as water in flowing between pressure monitoring points represented by nodes at the ends of the edge. The resistance may be calculated for multiple edges in the model. The resistance may be calculated for each of the edges in the model.

[0027] The method may include automatically calibrating the model. The model may be automatically calibrated using a numerical optimization method.

[0028] The method may include calculating a parameter at a node or edge for a first time period and for a second time period. The first time period and the second time period may be different. The first time period and the second time period may be non-overlapping time periods. The first time period and the second time period may be adjacent time periods.

[0029] The method may include recursively updating the model parameters. The method may additionally or alternatively include monitoring the parameter calculation results or estimation results. That is, monitoring the estimated one or more demands and / or one or more resistances. In addition, the method may include generating an alarm to identify changes or changes in the parameter estimation results. The change may be a significant change. The parameter estimation results may be monitored by comparing the parameter estimation results at the node or edge for the first time period with the parameter estimation results at the node or edge for the second time period. When the parameter estimation results are substantially unequal, an alarm may be generated. For example, when the estimation results are not completely equal or exceed the conventional system tolerance.

[0030] When the parameter estimation result for a given node or edge is not substantially equal to the previous estimation result of the same parameter for the same node or edge, an alarm may be generated. When the parameter estimation result for a given node or edge is not exactly equal to the previous estimation result, an alarm may be generated.

[0031] A change in demand at a node may indicate a rupture at or near the pressure monitoring point represented by the node.

[0032] A change in resistance to an edge may indicate a blockage in the conduit represented by the edge.

[0033] One or more parameters may be calculated over a given time period. The time period may be a 24 hour time period. The previous estimate may be an estimate for the immediately preceding time period, such as the immediately preceding 24 hour time period. Alternatively, the previous estimate may be an estimate for a time period previously identified as representing stability at a node and / or edge.

[0034] The model may be a model of a region of a hydraulic network. The region may include one or more fluid inlets. A flow meter may be arranged at each inlet leading to the region. Each flow meter may be configured to measure the flow of the fluid entering the region at the corresponding inlet of the flow meter. A node may exist at each inlet leading to the region.

[0035] One or more parameters may be calculated based at least in part on flow data from the or each flow meter. That is, the flow data may be an input to the model. Additionally or alternatively, one or more parameters may be calculated based at least in part on pressure data from a pressure monitoring point. That is, the pressure data may be an input to the model.

[0036] The demand at the node and / or the resistance at the edge may be calculated according to the model based on data collected at the or each flow meter and data collected at the pressure monitoring points.

[0037] The method may include detecting a rupture in the hydraulic network using the model.A rupture in the hydraulic network may be detected based on a change or variation in estimated demand at nodes of the model.

[0038] Edges may be generated by mathematically reducing existing pipeline topology data (i.e., a representation of the pipeline combination of the network may be generated). Existing pipeline topology data may be mathematically reduced using Kron Reduction and / or Delta-Wye Transformation. Alternatively or additionally, edges may be generated in a manual manner. Alternatively or additionally, statistical methods may be used to generate edges.

[0039] According to another aspect, a computer readable medium comprising instructions executable by a processor to perform the method is provided.

[0040] According to another aspect, a computer program product is provided, the computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method.

[0041] According to another aspect, a model of a hydraulic network is provided, which includes pipelines and pressure monitoring points, and the model includes: a plurality of nodes; and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network; and each edge connects two nodes and represents a combination of pipelines across the network between the pressure monitoring points represented by the two nodes.

[0042] The model may be used to calculate one or more parameters. As noted above, the one or more parameters may be the demand at a node and / or the resistance for an edge.

[0043] The network may include three pressure monitoring points, and / or the model may include three nodes. The network may include five pressure monitoring points, and / or the model may include five nodes.

[0044] According to another aspect, there is provided a use of a model for calculating a demand for a node and / or a resistance for an edge of the model.

[0045] According to another aspect, there is provided a use of a model for identifying a rupture and / or a blockage in a hydraulic network.

[0046] Optional features of each aspect may be optional features of each of the other aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Specific embodiments are described below, by way of example only, and with reference to the accompanying drawings, in which:

[0048] Figure 1 An equivalent network model of a first water distribution area overlaid with a full network model of the area is shown;

[0049] Figure 2 A full network model of the second water distribution area is shown;

[0050] Figure 3 showing a full network model of the second region overlaid with features of an equivalent network model of the second region;

[0051] Figure 4 showing a full network model of the second region overlaid with an equivalent network model of the second region;

[0052] Figure 5 An example of simulation of pressure and flow through a full network model of the second region is shown;

[0053] Figure 6 a graph showing flow demands at a first node and a second node of an equivalent network model of a second area;

[0054] Figure 7 Shown in Figure 4 The location of the simulated rupture on the model;

[0055] Figure 8 showing a graph of flow demand at a first node in view of a rupture;

[0056] Fig. 9 a graph showing flow demand at a second node in view of a rupture; and

[0057] Fig.10 A method for creating an equivalent network model of a hydraulic network is shown.

[0058] Throughout the specification, like reference numerals are used to describe like components. DETAILED DESCRIPTION

[0059] The invention outlined herein, referred to as the Equivalent Network (EN) model, is intended to fill the aforementioned gap between outdated but highly detailed network models and real-time but very coarse network monitoring systems.

[0060] Overview

[0061] In summary, the EN model is an improved method for modeling a specific area of ​​a hydraulic network that allows for improved estimates of fluid flow leaving the area by using information about the fluid flow entering the area and pressure readings throughout the area. Fluid can leave the area in a variety of ways: for example, through consumer demand and usage, background leaks, or emerging ruptures. In the case of a new rupture, the estimate of flow leaving the area can be detected as an anomaly with respect to previous estimates of flow leaving the area at a specific location. By detecting these anomalies, an advantageous EN model can be used to alert a water utility to a new rupture before prior techniques (such as noise recordings that require waiting one or two nights to detect and confirm a rupture) and in a more localized manner than traditional area flow monitoring (where an anomaly in the flow at the entrance to the area only tells the water utility that there is a rupture in the area, not the location in the area where the rupture is located).

[0062] A hydraulic network is a system of interconnected pipes arranged to carry a fluid, such as water. The fluid in the network may be pressurized. A hydraulic network may include at least one inlet arranged to receive a fluid into the system. A water supply network and a water network are examples of hydraulic networks.

[0063] A change in demand that deviates significantly from the expected demand may be considered an anomaly. As noted above, such an anomaly may indicate a rupture in the hydraulic network. Advantageously, the EN model may be used to identify such anomalies by: (a) using the model to estimate the expected (forecasted) demand; (b) using the model to estimate new demand based on actual conditions (e.g., flow data and pressure data); (c) comparing the expected demand to the new demand; and (d) identifying an anomaly if the estimate of the new demand is significantly higher than the estimate of the expected demand.

[0064] Example 1

[0065] Figure 1 An EN model 100 is shown overlaid with a full network model 110 of the water distribution area. The full network model 110 may be a conventional network model of the type described above.

[0066] In the present disclosure, a zone is defined as a discrete area of ​​a water supply network (i.e., a discrete area of ​​a hydraulic network). At the same time, the inlet of a zone is the point at which new water can enter the zone. In the UK, the most commonly formed zones are called District Metering Areas (DMAs). For example, DMAs found in the UK are typically smaller than zones in other countries in Europe. A zone can have more than one inlet. In the UK, it is most common for each DMA to have one inlet.

[0067] return Figure 1, a flow meter 120 is provided at the entrance of the physical area represented by the EN model 100. In order to use the EN model 100 for simulation, flow data at the entrance of the area is required. Typically, the flow data required and used by the EN model 100 is low frequency. In this example, low frequency typically means that measurements are taken every 5 minutes or every 15 minutes (however, in other examples, the frequency can be different). Such flow data is usually available from most water utilities.

[0068] Like the full network model, in which edges represent real pipes and nodes are connection points between pipes, the EN model 100 also includes nodes and edges. Figure 1 There are four nodes in the region: a first node 130, a second node 140, a third node 150, and a fourth node 160. The first node 130 and the second node 140 are connected by a first edge 170. The second node 140 and the third node 150 are connected by a second edge 180. The third node 150 and the fourth node 160 are connected by a fourth edge 190. The third node 150 is located at the entrance of the region.

[0069] The nodes 130, 140, 150, 160 of the EN model 100 represent real-time pressure monitoring points within the hydraulic network. For the numerical optimization required by the EN model 100, highly accurate pressure measurements are required. An example device that provides such accurate pressure measurements is the InflowSense RTM pressure sensing device sold by Inflowmatix Ltd. As described above, Figure 1 The EN model 100 shown in FIG. 1 includes four nodes 130, 140, 150, 160, each represented by an icon of an InflowSense RTM pressure sensing device. The nodes 130, 140, 150, 160 are distributed throughout the area. To monitor customer service levels, water companies typically already have at least one pressure sensor in each area of ​​the water company's area, but the number of sensors per area is increasing, with three sensors being satisfactory. In some arrangements, there may be five sensors.

[0070] Edges 170, 180, 190 in the EN model 100 represent imaginary pipes connecting the nodes 130, 140, 150, 160 of the EN model 100. The imaginary pipes represent a combination of real pipes spanning between any two pressure monitoring points of the network, ie, between any two nodes.

[0071] The use of the EN model 100 to estimate parameters associated with a modeled hydraulic network will now be described.

[0072] Advantageously, the parameters that can be estimated in the EN model 100 on a daily basis are equal to the corresponding parameters in the full network model 110. In more detail, at each node 130, 140, 150, 160 of the EN model 100, a time-dependent parameter, referred to as demand, can be estimated. Demand represents the combined water consumption and leakage at or near a particular node 130, 140, 150, 160, i.e., at or near that point in space. For example, an increase in demand at a node may indicate an actual rupture at or near that node. A parameter referred to as resistance acting on an imaginary pipe, i.e., the edges 170, 180, 190 between nodes, may be estimated. Resistance is related to the amount of energy lost by water flowing between any two monitoring points, i.e., nodes, in the EN model 100. An increase in resistance for an edge may indicate, for example, an actual blockage in one of the pipes represented by the edge. The resistance between two pressure monitoring points is affected by many practical factors, such as: (1) the distance between the two pressure monitoring points, the greater the distance, the greater the resistance; (2) the diameter of one or more pipes between the two pressure monitoring points, the smaller the diameter, the greater the resistance; (3) the roughness of one or more pipes between the two pressure monitoring points, the rougher the one or more pipes, the more fluid energy is lost and therefore the greater the resistance; and (4) the bends in one or more pipes between the two pressure monitoring points, the more bends the fluid must pass through, the more fluid energy is lost and therefore the greater the resistance. Therefore, the EN model 100 can be used to detect and manage ruptures and flows in general in hydraulic networks. The specific advantages of using the EN model 100 to detect and manage ruptures and flows are discussed in detail below.

[0073] Example 2

[0074] To aid the skilled person's understanding, a more detailed example of applying the described EN model to a publicly available full network model will now be provided. The example shows the ability of EN modeling to warn of and locate breaks in a water main network.

[0075] In this example, a publicly available full network model is accessible from the following website: https: / / emps.exeter.ac.uk / engineering / research / cws / resources / benchmarks / design-resiliance-p areto-fronts / data-files / . The full network model is in the EPANET format, which is an open source hydraulic simulation package. More information about models using the EPANET format is provided at the following website, particularly in the user manual that can be downloaded at the website: https: / / www.epa.gov / water-research / epanet. The full network model is called Fossolo, and will be referred to as Fossolo in this disclosure.

[0076] Figure 2 A Fossolo network 200 is shown. The Fossolo network 200 includes a plurality of nodes connected by pipes. The topology of the Fossolo network 200 will now be described.

[0077] The Fossolo network 200 has: one entry node 210, mathematically represented by no; 36 non-entry nodes, mathematically represented by nn; and 58 pipes, mathematically represented by np. As indicated above, the pipes connect the nodes. The single entry node 210, the non-entry nodes, and the pipes together form a highly circular topology.

[0078] The Fossolo network 200 does not include a design day consumer demand pattern. The design day consumer demand pattern is a consumer usage pattern, typically representing a consumer usage pattern within a 24-hour period, and is typically used for design work related to hydraulic networks, such as planning to install new control assets such as valves or understanding the impact of new housing development on the network. The data on which the design day consumer demand pattern is based (or used) is typically very outdated. For example, it is not uncommon for the life of the data to be between 5 and 10 years. Because the Fossolo network 200 does not include a design day consumer demand pattern, in order to generate the EN model, a pattern comprising 96 points (i.e., a day is divided into 15-minute intervals) will be applied. In other examples, a pattern comprising a different number of points may be used. For example, a pattern comprising 48 points (i.e., a day is divided into 30-minute intervals) may be used. Alternatively, a pattern comprising 144 points (i.e., a day is divided into 10-minute intervals) may be used.

[0079] The first stage of EN modeling is to build a simplified form of the topology that contains only nodes that are points of pressure sensors within the network. All inlet nodes must have flow data available. It is common for water utilities to monitor flows at the inlets of network areas.

[0080] Returns the topology structure that is built in a simplified form. Figure 3 A Fossolo network 200 is shown, annotated with circles identifying nodes that are pressure sensor points within the network. Water utilities typically have pressure sensors already placed within the water utility's network and regions. Placing such pressure sensors in an optimal manner is a complex discipline and matter, as water utilities typically do not employ advanced mathematical methods. In this example, for the purpose of generating the EN model, it is assumed that there are three pressure sensors within the network at locations that are typically of interest to water utilities: a first node 310 at the network entrance; a second node 310 at a CP 320; and a third node 330 at a point in the network that represents a key area or consumer.

[0081] Next, the remaining topology of the EN model is established. This can be done in a variety of ways, including one or a combination of the following:

[0082] - Use the recorded pipe topology from the water company, such as the water company's Geographic Information System (GIS), to mathematically simplify the network to a simplified form containing only the nodes of interest. Methods for simplifying the network in this way include Kron Reduction or Delta-Wye Transformation. Guidance on both methods can be found in the electrical engineering literature.

[0083] - If the network is simple enough, the topology is selected manually using network knowledge and domain expertise.

[0084] - Statistically inferring topology based on both low-resolution and high-resolution pressure data, for example, by experimenting with how dynamic pressure changes propagate through the network and how they are accurately detected at various time-synchronized sensors.

[0085] - Use other data sources and / or underlying assumptions about the water utility network. For example, if it is assumed that the water supply network in a particular area of ​​interest: (a) contains no closed valves, and (b) generally follows the street layout of a town or city, then the use of open source datasets such as OpenStreetMap can be used to identify approximate connectivity between sensors. The assumption made in (b) is generally correct because most consumers are located on the street and the network operator must generally be able to drive to locations throughout the network.

[0086] Returning to this example, since in this example the network is relatively basic and the total number of sensors is very low, a simplified topology of the EN model is considered Figure 4 The topology shown in .

[0087] Figure 4 A Fossolo network 200 is shown overlaid with a determined EN model 400. The EN model 400 includes the first node 310, the second node 320, and the third node 330 described previously. In the EN model 400, the first node 310 is connected to the second node 320 by a first edge 410. The second node 320 is connected to the third node 330 by a second edge 420. The third node 330 is connected to the first node 310 by a third edge 430. Each of the edges 410, 420, 430 is a straight line between the nodes connected by the edge. As previously described, the nodes 310, 320, 330 represent pressure sensor points in the system, and each edge 410, 420, 430 represents an imaginary pipe, which itself represents a combination of actual pipes in the network that span between the two nodes connected by the edge.

[0088] The simplified topology of the EN model 400 can be mathematically represented using an incidence matrix. For this Fossolo network 200, and using a notation similar to that in the EPANET User Manual (see above), the incidence matrix for the EN model is:

[0089]

[0090] Next, the topology of EN model 400 can be used within a least squares optimization solver to estimate multiple variables. In more detail, a least squares problem can describe: (1) multiple variables to be estimated, and (2) multiple equations to be applied (with some degree of error due to being calculated by a least squares solver).

[0091] (2) The equations described include principles related to hydraulic conservation. In more detail, first, for each node in the network, mass conservation applies. That is, the flow entering the node must be equal to the flow leaving the node. This is based on the assumption that the mass of the flow is incompressible. This mass conservation can be expressed by the following equation:

[0092]

[0093] In the equation, is the node "demand" and represents unknown network egress traffic due to consumer usage or disruptions in the network, or due to known ingress traffic at regional entry points; and is the unknown flow rate for each pipe.

[0094] As mentioned previously, the demand at a given node is a time-dependent parameter that represents the combined water consumption and leakage at or near that node.

[0095] The unknown demand is the primary variable of interest because, by estimating the unknown demand from the optimization problem, anonymous demand or one or more breaches in the system can be identified and corresponding alerts can be issued.

[0096] Secondly, for each edge in the network (e.g., for each of the edges 410, 420, 430), energy remains conserved. Pecci F, Abraham E, Stoianov II, 2017 show that a quadratic head loss approximation for complex energy conservation relations such as Hazen-Williams or Darcy-Weisbach in water supply networks (Journal of Hydroinformatics, Vol: 19, Pages: 493-506, ISSN: 1464-7141) for optimizing the problem can be estimated with good accuracy for a single pipe using only quadratic equations. Therefore, for the described EN modeling method, a quadratic head loss relation is used to capture the energy conservation throughout the simplified network, i.e., throughout the EN model 400:

[0097] A 12 h+kq 2 =0

[0098] In this quadratic head loss relationship, is the node pressure head, which is formed by adding the known pressure data at each node in the network to the height of the known pressure data relative to some reference, and is the assumed unknown edge resistance.

[0099] The height required to form a node pressure head can be measured by one or a combination of the following methods: using field equipment; obtaining from existing water company records such as a GIS; and / or inferring from pressure data at multiple points in the network during periods of low flow (i.e., low energy loss), such as at night.

[0100] The least squares optimization problem uses the above conservation equations to form a system in which the unknowns are [nt×nn+np×(nt+1)] and the number of equations available is [nt×(np+nn+no), where nt is the length of the time series of pressure and flow data used to form an understanding of the network. As noted above, nn is the number of non-entry nodes; np is the number of pipes; and no is the number of entry nodes.

[0101] Depending on the topology of the network being analyzed, the matrix formed by this system of equations may have full rank, or it may require additional data or a small form of regularization (both of which relate to the resistance of the network).

[0102] In the case where GIS data is available, the GIS data can be used to define the resistance for each edge in the network. If GIS data is not available, for example if the GIS data is missing, and in the case of any other missing data, the example involves "pushing" the resistance in the direction that makes the most logical sense. This is the approach taken in this particular example, where a small regularization is applied that penalizes small resistances. As a result, the demand of the EN model 400 may lose meaning in terms of the absolute value of the demand, however, the demand will still be valuable for detecting anonymous high demands relative to previous estimates.

[0103] The above equations can be used to form A and b in the least squares problem [A×=b]. For a source network (number=1), assuming nt>np, there are more equations than unknowns. The least squares problem for the Fossolo network 200 can then be solved using EPANET, for example, to simulate pressure and flow throughout the network. Example results of this simulation are shown in Figure 5 as shown in .

[0104] Figure 5 A Fossolo network 200 is shown overlaid with two graphs: a first graph 500 shows the pressure simulated at a node in the Fossolo network 200, and a second graph 510 shows the flow simulated at a connection (i.e., pipe) in the Fossolo network 200. Pressure is expressed in meters vs. time; while flow is expressed in liters / second vs. time.

[0105] Next, only the simulated pressures and flows at the sensor points (i.e., nodes 310, 320, 330) are selected as inputs to the EN model 400. These inputs are used to effectively minimize the difference between the observed values ​​and the model variables in a least squares problem. The demands at the second node 320 and the third node 330 are then calculated based on the design day data. The calculated demands are shown in Figure 6 as shown in .

[0106] In more detail, Figure 6 The outlet flow rate (in liters per second) of each of the second node 320 and the third node 330 within 24 hours is shown. That is, Figure 6 The outlet flow rate (in liters per second) during the first day is shown. Figure 6 , the demand at the third node 330 is represented by the line labeled "Node 1," and the demand at the second node 320 is represented by the line labeled "Node 2." For both nodes, the demand, i.e., the outlet flow, is relatively low until around 6 a.m., when a sharp increase in demand is seen. The demand is then seen to be relatively high throughout the day until around 8 p.m., when the demand begins to decline. A small drop in demand is also seen between the hours of approximately 11 a.m. and 3 p.m. Although the overall demand at the second node 320 is lower than the overall demand at the third node 330, the demand curves at the two nodes are similar during the first day. Therefore, overall, Figure 6 A typical consumer pattern is shown in which demand is low at night when people are sleeping, peaks in the morning when people wake up, and peaks again in the evening after people finish work and school.

[0107] To illustrate how the EN model 400 can be used to detect anomalies in demand, such as ruptures, a second day of simulation is performed on the Fossolo network 200. On the second day, a simulated rupture of 1 liter per second is introduced near a critical point at 12 p.m., that is, near the second node 320. Figure 7 Shows coverage in Figure 4 The position of the simulated rupture 700 is closer to the second node 320 than to any one of the first node 310 and the third node 330.

[0108] Figure 8 and Fig. 9 The corresponding predicted demand and actual demand for the third node 330 and the first node 310 in the second day are shown, respectively. The predicted demand is based on the demand of the corresponding node on the first day. The actual demand is the model demand for the second day, which is determined using the EN model 400 in substantially the same manner as described above. Figure 8 and Fig. 9Both show the demand flow in litres / second over a 24 hour period making up the second day.

[0109] If Figure 8 As shown, the demand at the third node 330 on the second day is approximately the same as the demand at the third node 330 on the first day. In other words, according to the model, the rupture event has no effect on the demand at the third node 330. Therefore, Figure 8 It shows that there is no difference between the actual demand and the forecasted demand for the next day.

[0110] However, if Fig. 9 As shown, the demand at the second node 320 on the second day is different from the demand at the second node on the first day. In particular, the demand at the second node 320 on the second day is higher between the hours of approximately 12 p.m. and 12 a.m. compared to the demand at the second node 320 on the first day. Therefore, the rupture is visible at the second node 320. This is because the second node 320 is closer to the rupture location 700.

[0111] method

[0112] Fig.10 A method of modeling a hydraulic network 1000 is shown. Each of the steps of the method is represented by a box. The method 1000 involves creating 1020 a model of a hydraulic network, which includes pipes and pressure monitoring points. In addition, the method 1000 involves representing 1020 each pressure monitoring point in the hydraulic network by a node. Finally, the method 1000 involves representing 1030 a combination of pipes spanned between pressure monitoring points by edges, each edge connecting two nodes.

[0113] The method 1000 may be a computer-implemented method that can be executed, for example, on a processor of a computer. The model created by the method 1000 may be stored in a memory of a computer or on the computer.

[0114] The method 1000 may also include modeling the hydraulic network, for example using the created model, to identify a breach in the hydraulic network.

[0115] The method 1000 may also include modeling the hydraulic network using the created model to calculate the demand at the node and / or the resistance for the edge. The demand at the node may be used to identify a rupture at or near the node.

[0116] Alternatives

[0117] Optionally, the above example may be performed in conjunction with an optimal sensor placement method. Riskily, this maximizes the performance of the described EN modeling method. This is because, as the number of sensors in the network increases, smaller demand estimates are created for each node in the EN model 400, and therefore, identifying anonymous demand becomes simpler relative to predicting demand. However, in practice, the method must also be adapted to work with an existing sensor arrangement that has been selected by the water company responsible for the hydraulic network, as it is not always feasible for the water company to purchase and install additional sensors. Therefore, there are practical limits to optimizing the sensor placement method.

[0118] Optionally, any number of pressure sensors may be used in a region represented by an EN model. Thus, any number of nodes may be present in the EN model. For example, the actual number of pressure sensors may vary depending on the particular hydraulic network being modeled. For example, there may be three pressure sensors in the region. Thus, there may be three nodes in the EN model for the region. Alternatively, there may be five pressure sensors in the region. Thus, there may be five nodes in the EN model for the region.

[0119] Optionally, there can be any number of edges in the EN model.

[0120] An EN model may represent more than one region. For example, an EN model may represent two, optionally adjacent regions.

[0121] The term network in this disclosure refers to a hydraulic network. Throughout this disclosure, the terms network and hydraulic network may be used interchangeably. Additionally, the term network may refer to an example of a hydraulic network, such as a water supply network.

[0122] The results of the EN model can be represented and / or visualized by a full network topology, such as a traditional network topology. Advantageously, this will make more intuitive sense to users of the technology.

[0123] The EN model itself can be represented and / or visualized on a full network topology, such as a traditional network topology. Advantageously, this will make more intuitive sense to users of the technology.

[0124] Each edge can represent an independent pipe between two pressure monitoring points in the hydraulic network.

[0125] benefit

[0126] The described EN modeling approach has many technical advantages, including the following:

[0127] The EN model addresses the shortcomings of the previously described regional level monitoring and full network models. This is because, like full network monitoring, the EN model can be used to provide an up-to-date understanding of the conditions in a particular region; however, unlike full network monitoring, the EN model can be automatically constructed and calibrated multiple times a day. This can be achieved using numerical optimization methods. Therefore, a more accurate and dynamic modeling approach is provided, especially because the parameters of the EN model will have more recent and relevant data than the corresponding parameters of the full network model. The advantage of the EN model is that better spatial resolution is achieved compared to regional level detection. For example, the physical location of a rupture in a region can be identified, rather than simply the presence of a rupture in a region.

[0128] As mentioned previously, the EN model can be used to give real-time hydraulic understanding within the network area, including regionalized understanding of:

[0129] · ruptures, background leaks and unusual consumer demands; and / or

[0130] Restrictions, blockages and improperly throttled valves in the hydraulic network.

[0131] By feeding the parameter estimates of the EN model into a simple anomaly detection method, an alert can be generated to identify significant changes in the parameter estimates. There is a strong motivation for water utilities to quickly identify and locate emerging ruptures, many of which are not visible at ground level. These motivations include financial and environmental. As will be appreciated by those skilled in the art, water is a valuable resource and it is highly desirable to limit the waste of water in the hydraulic network, for example through ruptures. In addition, it is also beneficial from a financial and environmental perspective to identify resistances in the network that are greater than necessary. This is because water utilities must provide sufficient pressure in the system to overcome such resistance, resulting in increased energy use and higher levels of pressure associated with background leakage.

[0132] In addition, EN models can be used to control the network to bring higher efficiency. For example, existing methods for managing pressure within discrete areas involve continuous monitoring of two important points: a pressure reducing valve (PRV) near the entrance of the area, which is used to control the pressure in the area; and a critical point (CP) at which the pressure in the area is the lowest. The monitoring method generally includes: (1) collecting low-resolution pressure data (typically 15-minute resolution data) at both the PRV and the CP; and (2) collecting flow data at the PRV. This data can be used to establish relationships to construct optimized control information. EN models can be used in this particular space, and EN models can also enable more advanced control schemes, where an accurate network model is required for the successful operation of the scheme.

[0133] Furthermore, as previously noted, not all water utilities around the world have a full hydraulic network model for their systems, or, if they do, such a model may be of low quality or difficult to implement. A major advantage of the EN model is that it can be built and calibrated without a full hydraulic network model, and therefore, for example, can be built and calibrated regardless of the particular water utility managing the hydraulic network or the current capabilities of the water utility. This is because the EN model is created starting from sensor data (e.g., data from pressure sensors) and that model is built into the EN model, rather than, for example, starting from a full hydraulic network model and simplifying that full model. This means that there are fewer barriers to creating an EN model, and therefore, for example, the environmental benefits of the EN model described previously can be provided regardless of the (lack of) information available on the hydraulic network.

[0134] Yet another advantage of the EN model is that using the EN model, resistance and demand in the network can be understood. As a result, multiple points in a specific area can be controlled. This means that a single CP problem can be solved multiple times. This is different from the previous description, where the pressure in the network can be minimized only by valve control and pressure monitoring at the CP in a simple approach.

[0135] In general, the EN model provides an improved method for detecting rupture events. This is because the EN model provides a compromise between speed, localization level and burden on computing resources. This is because, based on demand prediction (assuming that the network has a predictable pattern, which is generally applicable to most networks), simple logic can be formed and used to generate alarms (e.g., alarms for rupture events) earlier than when other technologies are used. For example, earlier than when minimum night flow monitoring or acoustic noise recording is used. Therefore, ruptures can be detected (and corrected) quickly, and because simple logic can be used, the burden on computing resources can be managed within an acceptable level. In addition, the alarm can include a more accurate localization form than locating the network only for the entire area. That is, the location of the rupture can be more accurately identified than a simple identification of a given area of ​​the network. In contrast, for example, the minimum night flow monitoring process is only local to the entire area of ​​the network. Therefore, the resulting localization level is higher. In general, the EN modeling method therefore allows for faster and more accurate identification of rupture events, which means that such rupture events can be solved more quickly by engineers, thereby reducing water losses from the hydraulic system.

[0136] Finally, as will be appreciated by those skilled in the art, hydraulic (e.g., water) networks are large in scale and have a very limited number of sensors. As a result, knowing something about every pipe in the network (such as whether a rupture has occurred) is challenging and may seem infeasible to those skilled in the art. The disclosed EN modeling approach addresses this challenge by reducing the size of the network and the problem. Therefore, it is not obvious to a technician that there is enough actual sensor data for some hydraulic networks to accurately solve the problem. For other network topologies, where there is still not enough data to accurately solve the problem, the EN model can be used to make additional assumptions (i.e., regularization) to solve the problem. The reason why the EN model can be used to make such additional assumptions is because, when the EN model is used to calculate demand, the demand does not need to be accurate in an absolute sense; rather, the demand only needs to be accurate relative to a previous estimate. This is because anomalies in demand (which may indicate a rupture in the system) can be identified based on changes in demand estimates rather than absolute values.

[0137] Example layout

[0138] This disclosure includes the subject matter described in the following clauses:

[0139] Item 1: A model of a hydraulic network comprising pipelines and pressure monitoring points, the model comprising: a plurality of nodes; and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network; and each edge connects two nodes and represents a combination of pipelines of the network spanning between the pressure monitoring points represented by the two nodes.

[0140] Clause 2: According to the model of Clause 1, the network includes three pressure monitoring points and the model includes three nodes.

[0141] Clause 3: A model according to clause 1 or clause 2, wherein the pressure monitoring point is a real-time pressure monitoring point.

[0142] Clause 4: A method for modeling a hydraulic network comprising pipelines and pressure monitoring points, the method comprising: creating a model comprising a plurality of nodes and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network, and wherein each edge connects two nodes and represents a combination of pipelines of the network spanning between the pressure monitoring points represented by the two nodes.

[0143] Clause 5: The method of clause 4, wherein the pressure monitoring point is a real-time pressure monitoring point.

[0144] Clause 6: The method according to clause 5, further comprising calculating a time-dependent parameter at a node, referred to as demand, wherein demand represents a combined water consumption and leakage at or near a pressure monitoring point represented by the node.

[0145] Clause 7: A method according to clause 5 or clause 6, the method further comprising calculating a parameter for one or more of the edges, referred to as resistance, wherein, for each edge, the resistance is related to the amount of energy lost by water flowing between pressure monitoring points represented by nodes at the ends of the edge.

[0146] Clause 8: The method of any one of clauses 5 to 7, further comprising: automatically calibrating the model using a numerical optimization method.

[0147] Clause 9: The method of any one of clauses 5 to 8, further comprising: monitoring the parameter estimation results; and generating an alert to identify significant changes in the parameter estimation results.

[0148] The methods described herein may be implemented on a computer-readable medium, which may be a non-transitory computer-readable medium. The computer-readable medium carries computer-readable instructions arranged to be executed on a processor, thereby causing the processor to perform any or all of the methods described herein.

[0149] As used herein, the term "computer-readable medium" refers to any medium that stores data and / or instructions for causing a processor to operate in a particular manner. Such storage media may include non-volatile media and / or volatile media. Non-volatile media may include, for example, optical disks or magnetic disks. Volatile media may include dynamic memory. Exemplary forms of storage media include floppy disks, floppy disks, hard disks, solid-state drives, magnetic tapes or any other magnetic data storage media, CD-ROMs, any other optical data storage media, any physical media with one or more hole patterns, RAM, PROMs, EPROMs, FLASH-EPROMs, NVRAMs, and any other memory chips or cassettes.

[0150] Features of the above embodiments may be combined in any suitable manner. It will be understood that the above description is only a description of specific embodiments by way of example, and that many modifications and variations will be within the capabilities of the skilled person and are intended to be covered by the scope of the appended claims.

Claims

1. A computer-implemented method for modeling a hydraulic network, the hydraulic network comprising pipelines and pressure monitoring points, and the method comprising: Creating a model, the model comprising a plurality of nodes and a plurality of edges connecting the nodes; wherein each node represents a pressure monitoring point within the hydraulic network, and each edge connects two nodes and represents a combination of the pipes of the network spanning between the pressure monitoring points represented by the two nodes; The method comprises: calculating one or more parameters based on the model; the one or more parameters comprising demand for nodes and / or resistance for edges; demand representing a combined water consumption and / or leakage at or near a pressure monitoring point represented by a corresponding node; resistance relating to an amount of energy lost by water flowing between pressure monitoring points represented by nodes at ends of a corresponding edge; The method comprises: calculating a parameter for a first time period and for a second time period; and comparing the calculated parameter results for the first time period with those for the second time period; The method includes identifying a change in parameter calculation results for the first time period and for the second time period and generating an alert indicating the change.

2. The method according to claim 1, wherein: The model is a model of a zone in the hydraulic network and wherein a flow meter is arranged at each inlet to the zone.

3. The method according to claim 1, wherein: The method includes detecting a rupture in the hydraulic network using the model.

4. The method according to claim 3, wherein: Detecting a rupture in the hydraulic network includes detecting a change in an estimated demand at a node of the plurality of nodes.

5. The method according to claim 1, comprising: The model is used to calculate resistance for an edge of the plurality of edges to identify a blockage in the conduit represented by the edge.

6. The method according to claim 1, wherein: The pressure monitoring point is a real-time pressure monitoring point.

7. The method according to claim 1, comprising: The model is automatically calibrated using numerical optimization methods.

8. The method according to claim 1, comprising: Demand at a node among the plurality of nodes is calculated.

9. The method according to claim 1, comprising: A drag is calculated for an edge in the plurality of edges.

10. The method according to claim 1, wherein: Calculating the one or more parameters includes performing regularization on the model.

11. The method according to claim 1, comprising: The model parameters are recursively updated.

12. The method according to claim 1, wherein: When the parameter is a demand at a node of the plurality of nodes, the change indicates a rupture at or near the pressure monitoring point of the hydraulic network represented by the node.

13. The method according to claim 1, wherein: When the parameter is a resistance for an edge of the plurality of edges, the change indicates a blockage in the conduit of the hydraulic network represented by the edge.

14. The method according to claim 1, wherein: The one or more parameters are calculated based at least in part on the flow data and / or the pressure data.

15. The method according to claim 1, wherein: The plurality of edges are generated by one or more of: mathematically reducing existing pipeline topology data; manually generating the edges; and / or statistically generating the edges.

16. A computer-readable medium comprising instructions executable by a processor to perform the method according to any one of claims 1 to 15.

Images