The present invention generally relates to the field of monitoring fluid supply networks including leakage detection in fluid supply networks. Pipeline networks are the most economic and safest mode of transportation for fluids like water, oil, ga'ses and other fluid products. As a means of long-distance transport, pipelines have to fulfill high demands of safety, reliability and efficiency. If properly maintained, pipelines can last indefinitely without leaks. The leaks in the pipeline network can be caused by various reasons: damage from nearby excavation equipment, corrosion of pipes, accidents, earth movement, cyclic load, related to changes in the network operation mode etc . Such a system of interconnected pipes that carries a pressurized fluid such as water, oil, gases and other fluid products is called hydraulic supply network. When monitoring hydraulic supply networks, one often faces the task of monitoring states of fluid supply networks including leakage detection and leakage localization. If there is a problem in the network, it is very important to troubleshoot the defect in short time. Timely localization of the problem allows reducing cost of repairing and possible liquid losses on the network. Consequences of the leakage can be very destructive. Any fluid supply network, including a water supply system, typically includes: fluid sources or fluid storage facilities such as reservoirs, tanks, etc.; pressurizing components such as pumping stations, pumps, etc.; a network of pipes for transportation and distribution of fluid to the consumers. Further the fluid supply network is considered on an example of a water supply network.

A water supply system or water supply network also belongs to the hydraulic supply networks which provide water supply to different types of consumers. The water in the supply network is maintained at positive pressure to ensure that water reaches all parts of the network, that a sufficient flow is available at every take-off point, i.e. at every consumer, and to ensure that untreated water in the ground cannot enter the network. The water is typically pressurized by the pressurizing components .

Different types of pipes are used in the pipe network of the water supply network. In general, the pipes can be classified in two categories depending on purpose: main pipes or transportation pipes, that are mainly long pipes located underground with large diameters of, for example, 300 - 700 mm, but can be of giant diameters of more than 3m, moving pressurized water from the water storage facilities into the town or district of the town. distribution pipes, that are pipes with small diameters of, for example, 80 - 300 mm, used to take the water from the main pipes to the consumers, which may be private houses as whole or each apartment individually, or industrial, commercial or institution establishments, and other usage points such as fire hydrants.

The topology of the water supply network is well known by the utilities companies who service the network. It means that further characteristics of the water supply network are well known: structure and arrangements of pipes, diameters of pipes, pipes lengths, location of sensors of different types, pipe roughness, vertical topology of pipes, etc.

By now water has become one of the most important goods in the 21st century. However, sometimes considerable water losses occur in water supply networks.

The term "water loss" is generally adopted to indicate the difference between the overall amount of water supplied into the network and the sum of the water volumes corresponding to the consumers' consumption recorded by flow meters installed on consumers' nodes.

These water losses can be divided into two groups: the apparent losses, e.g. unrecorded water volumes used for public functions, such as cleaning of roads and urban areas, irrigation of green spaces, operation of public fountains, fire-extinguishing service, which consist of water volumes actually consumed but not accounted for, and the real losses, that are caused by damages that may have occurred to the network pipes or by the deterioration of the pipe junctions or the hydraulic devices. Real losses are the physical losses of water from the water supply system, also referred as "water leakages".

These losses put a strain on water supply and inflate the management cost for the water utilities since they represent water that is extracted and treated but never reaches the consumers .

In many cases, minor water leakages deriving from the inefficient hydraulic seal of junctions or from small cracks on pipes may lie hidden for a long time, sometimes for months or even years. Major leakages can be easily observed when significant damages to the pipes occur, as they usually result in large amounts of water erupting from ground or flowing in the consumer properties. The proven method around the world to reduce leakage from the water supply system is to proactively find the leaks before they appear at the surface. This can be achieved by monitoring the network and has the benefits of reducing the time the leaks are running, and wasting water. The supply network monitoring can be performed using the local leak detection methods by using various equipment, such as acoustic equipment, thermograph, ground penetration radar, etc. In most cases such method requires a lot of labor since experts with the respective equipment should trace the whole supply network to find the leakage.

Also there are global methods of the supply network monitoring that are based on gathering a lot of data about the network condition such as nodal pressure and pipe flows and analyzing them.

According to international and national standards, the best practice method for monitoring a water supply network is to sectorize it into district metered areas (DMA) . A DMA is an area with strict boundaries within the water supply network with measured inflow into this discrete area. This technique was first introduced at the beginning of the 90s. However the DMA' s approach has its disadvantages. Firstly, dividing the water supply network into smaller areas comes with a cost - the cost of an area survey, installation design, flow meter and chamber installation etc., is substantial, particularly if small areas (say less than 1000 consumers) are chosen. Since the main pipe diameters are big (typically more than 300 mm) , flow meters and valves to be installed on this main pipes are big and therefore expensive as well.

Also, creating DMAs requires the 'permanent' closure of many boundary valves - and, because of area supply arrangements and network characteristics (such as topography and low system pressures) some networks are hydraulically difficult to divide into single-feed DMAs without disadvantaging consumers .

In general, the existing ways of the fluid supply network monitoring are complex, time consuming, costly, and require very sophisticated and expensive equipment. In the light of the foregoing discussion, it is evident that there is a strong need of an easy and convenient determining a fluid supply network state.

The object is solved by a method for determining a fluid supply network state as defined in claim 1, and by a system for determining a fluid supply network state as defined in claim 8..

Consequently, the present invention provides a method for determining a state of a fluid supply network, where the fluid supply network comprises a network of pipes for delivering a fluid to consumers, wherein the network of pipes comprises a plurality of pipes. The fluid supply network also comprises at least one sensor located on a pipe of the plurality of pipes. The at least one sensor is adapted to measure at least one parameter of a fluid condition in the pipe of the fluid supply network where the at least one sensor is located, wherein measuring such at least one parameter is able to provide an observation over change of a fluid supply network state. Moreover the at least one sensor is adapted to provide sensor data, which comprises information about the measured parameter.

According to the present invention the method comprises at least a training mode and at least a test mode.

Within the training mode, a plurality of sensor reference data sets for a plurality of predefined states of the fluid supply network is calculated. Each sensor reference data set comprises a plurality of sensor data that would be provided by the at least one sensor in case the fluid supply network would be in that particular predefined state which corresponds to the respective sensor reference data set. Thus one sensor reference data set corresponds to one predefined state, and different sensor reference data sets of the plurality of the sensor reference data sets correspond to different predefined states of the plurality of predefined states of the fluid supply network.

And further on a classifier is trained by using the plurality of calculated sensor reference data sets for the plurality of predefined states of the fluid supply network, such that the trained classifier is adapted to select and provide a particular state of the fluid supply network as a response to a sensor data set provided to the classifier. While the sensor data set comprises the sensor data provided by the at least one sensor.

Within the test mode an actual sensor data set is collected wherein the actual sensor data set comprises the actual sensor data provided by the at least one sensor in a particular point in time for which an actual state of the fluid supply network should be determined.

After that the judgment about an actual state of the fluid supply network is made. Such judgment is based on the collected actual sensor data set: the collected actual sensor data set is provided to the trained classifier and the trained classifier provides an actual state of the fluid supply network (1) as a response to the collected actual sensor data set provided to the trained classifier.

Furthermore the present invention provides a system for determining an actual state of a fluid supply network. According to the present invention the system comprises at least one sensor wherein the at least one sensor is located on a pipe of the plurality of pipes. The at least one sensor is adapted to measure at least one parameter of a fluid condition in the pipe of the fluid supply network where the at least one sensor is located, wherein measuring such at least one parameter is able to provide an observation over change of a fluid supply network. Moreover the at least one sensor is adapted to provide sensor data, which comprises information about the measured parameter. In addition to that the system comprises a control unit configured to determine a fluid supply network state in accordance with method of any one of the claims 1 - 7.

Therefore there is a fluid supply network with the system for determining an actual state of a fluid supply network. The present invention is based on the insight that using a hydraulic model of the fluid supply network that represents the real fluid supply network it is possible to model different predefined states of the fluid supply network and to calculate the sensor reference data sets, i.e. calculate what the value would be measured by each sensor installed on pipes in the fluid supply network in case the fluid supply network would be in that particular predefined state and such sensor reference data sets will be unique combination of sensor data for the predefined state of the fluid supply network: and to train the classifier by using the calculated sensor reference data sets such that the trained classifier can provide an actual state of the fluid supply network as a response to the collected actual sensor data set provided to the trained classifier.

Therefore a judgment about the actual state of the fluid supply network can be made by using the trained classifier. Such judgment is based on knowledge of the predefined states for which the plurality of sensor reference data sets was calculated.

In other words as soon as the actual sensor data set that corresponds to an actual state of the fluid supply network to be determined, is provided to the trained classifier, the trained classifier as a response selects and provides an particular state of the fluid supply network out of the plurality of predefined states that corresponds to the actual sensor data set.

Actually the plurality of the sensor reference data can be calculated for different predefined states of the fluid supply network, e.g. ^' the states of the fluid supply network with different leakages of different types and sizes, in different locations, with different combinations of such leakages; the states of the fluid supply network with different fluid pressure provided by at least one pressurizing component, wherein the at least one supply network pressurizing component provides a predefined positive pressure in the network of pipes ; the states of the fluid supply network with different consumptions by the consumers, the state of the fluid supply network for a particular point in time, etc. The more sensor reference data sets are calculated for the more predefined states of the fluid supply network, the more precise the actual state of the fluid supply network can be determined.

As soon as the sensor reference data sets are calculated for the predefined states with or without leakages, the actual state of the fluid supply network can comprise information about leakages. Therefore leakages can be easily suspected and localized as well.

Thus, the present invention is proposed to provide a new method and a system for determining a fluid supply network state .

Further embodiments of the present invention are subject of the further sub-claims and of the following description, referring to the drawings . In a possible embodiment in addition to the plurality of sensor reference data sets, historical sensor data sets are calculated as well.

To calculate the historical sensor data sets, historical sensor data are collected by taking a plurality of measurements of the parameters measured by the at least one sensor for a given period of time in past. Wherein different historical sensor data provided by the at least one sensor corresponds to different point in time within the given period of time in past. After that plurality of historical sensor data sets are calculated. Each historical sensor data set comprises, at least one historical averaged sensor data which is an average of those historical sensor data collected within the given period of time that corresponds to the same point in time within a day. Each historical sensor data set corresponds to a particular point in time, and different historical sensor data sets correspond to different particular points in time within a day.

Further on at the step of calculating the plurality of sensor reference data sets for the plurality of predefined states of the fluid supply network, each sensor reference data set comprises the plurality of sensor data, that would be provided by the at least one sensor in case the fluid supply network would be in that particular predefined state which corresponds to the respective sensor reference data set, wherein such sensor data is reduced by the historical averaged sensor data of the respective historical sensor data sets for the respective sensor.

Therefore the classifier is trained by using the plurality of the calculated plurality of sensor reference data sets wherein historical behavior of the fluid supply network is taken into account. Further, within the test mode, at the step of collecting an actual sensor data set, the actual sensor data set comprises the sensor data provided by the at least one sensor in a particular point in time for which a state of the fluid supply network should be determined, reduced by the historical sensor data set for the respective point in time for the respective sensor.

Therefore the judgment about the actual state of the fluid supply network is made based on the collected actual sensor data where the historical behavior of the fluid supply network is taken into account. This embodiment allows taking into account historical average behavior of the fluid supply network. This feature allows training the classifier in a better way, therefore determining the actual state of the fluid supply network more precisely.

In enhanced embodiment within the training mode at the step of calculating the plurality of sensor reference data sets, the plurality of sensor reference data sets comprises at least one no- leakage sensor reference data sets which comprises a plurality of sensor data that would be measured by the at least one sensor in case the fluid supply network would be in a corresponding predefined state of the fluid supply network without any leakages in the fluid supply ne work, a plurality of leakage sensor reference data sets wherein each leakage sensor reference data set comprises a plurality of sensor data that would be measured by the at least one sensor in case the fluid supply network would be in a corresponding predefined state of the fluid supply network with at least one leakage. '

The different leakage sensor reference data sets correspond to different leakage states of the fluid supply network, wherein different leakage states correspond to states of the fluid supply network with leakages in different locations and / or of different parameters.

After training the classifier by using such plurality of leakage and no- leakage sensor reference data sets, the judgment about an actual state of the fluid supply network can be done . In enhanced embodiment a judgment about a location and/or parameter of the at least one leakage can be made. Such judgment is based on knowledge of the predefined states for which the plurality of sensor reference data sets were calculated. Therefore the leakages in the fluid supply network can be suspected and localized.

In other embodiment the at least one parameter of the fluid condition measured by at least one sensor is fluid pressure. The pressure of the fluid in the pipe of the fluid supply network is sensitive to the different changes of condition of the fluid supply network, for example to leakages in the fluid supply network. In addition to that pressure sensors are relatively cheap in comparison with, for example, flow meters especially if the flow meters are installed on the main pipes .

In another embodiment the at least one sensor is installed on a distribution pipe. As it was mentioned above the network of pipes for delivering a fluid to consumers comprises at least one main pipe to transport fluid from a source into the fluid supply network, and at least one distribution pipe, wherein the at least one distribution pipe is adapted to transport fluid from the main pipe to a consumer fluidly connected to the distribution pipe.

The distribution pipes are pipes with small diameter (e.g. less than 300 mm) to carry the pressurized fluid from the main pipe to the consumers . The pressure sensors that are installed on the distribution pipes are relatively cheap in comparison with the ones that installed on the main pipes. In addition to that the installation process of pressure sensors on a distribution pipe is not connected with the interruption of fluid supply to the consumers of the fluid supply network.

Therefore using pressure sensors installed on the distribution pipes makes this method more cost effective in comparison with the case when pressure sensors are installed on the main pipe . In other possible embodiment the fluid supply network is a water supply network.

For a more complete understanding of the present invention and advantages thereof, reference is now made to the following description taken in accompanying drawings . The invention is explained in more detail below using exemplary embodiments which are specified in the schematic figures of the drawings, in which:

Fig. 1 shows a block diagram of a fluid supply network; Fig. 2 shows a flow diagram of an embodiment of a method for determining a fluid supply network state according to the present invention;

Fig. 3 shows a flow diagram of another embodiment of a method for determining a fluid supply network state according to the present invention;

Fig. 4 shows a block diagram of a system for determining a fluid supply network state according to the present invention.

Various embodiments are described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purpose of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more embodiments. It may be evident that such embodiments may be practice without these specific details.

The invention relates to a system 13 for determining a fluid supply network state.

FIG 1 illustrates a water supply network 1 with a water source (not shown in FIG 1) , a supply network pressurizing component 2, main pipes 3 and distribution pipes 4. The fluid pressurized by the supply network pressurizing component 2 is being transported through main pipes 3. Distribution pipes 4 are used to connect the consumers 5 with the main pipes 3.

The main pipes 3 and distribution pipes 4 might be equipped with sensors 6. Such sensors 6 can be installed directly in the main pipes 3, or in the distribution pipes 4. Such sensors 6 can measure different parameters of a fluid condition in the pipe of the fluid supply network 1 where such sensors 6 are located. Each sensor 6 can measure one parameter or plurality of them. Also several sensors measuring at least one parameter can be installed in the same place in the pipe 3 , 4.

Such parameters measured by the at least one sensor 6 can be pressure of the fluid in the pipes 3, 4, pressure of the fluid provided by the supply network pressurizing component 2, temperature, consumptions, fluid quality parameters such as fluid chemical composition, turbidity, etc.

The sensor can be a pressure sensor, a flow meter and etc.

However the observability of fluid supply network state should be achieved by measuring such parameters . In other words in case the state of the fluid supply network 1 is changing the value of such parameters are also changing. Therefore it is possible to determine a state of the fluid supply network 1 by measuring such parameters.

For example, measuring pressure in the fluid supply network it is possible to determine emergency of the fluid leakage. Since emergency of the fluid leakage somewhere in the fluid supply network 1 will bring to the pressure drop in the pipe 3, 4 where the leakage is appeared and can be detected by the sensors 6 installed on the pipes 3, 4. As it is mentioned above said distribution pipes 4 are pipes with small diameter (e.g. less than 300 mm) to carry the pressurized fluid from the main pipe 3 to the consumers 5. The sensors 6 that are installed on the distribution pipes 4 are relatively cheap in comparison with the ones that are installed on the main pipes 3. In addition to that the installation process of said sensors 6 on the distribution pipes is not connected with the interruption of fluid supply to the consumers 5. The more sensors 6 are installed in the fluid supply network 1 the more precise the state of the fluid supply network 1 can be described.

The topology of the fluid supply network 1 is well known by the utilities companies who service the fluid supply network 1. It means that further characteristics of the fluid supply network 1 are well known: structure and arrangements of pipes 3, 4, diameters of pipes 3, 4, pipes lengths, location of sensors 6, etc. Also other information about the fluid supply network parameters, e.g. the pipeline roughness or vertical topology of the fluid supply network 1 is available or can be modelled / calculated.

Having all these information describe the fluid supply network 1 it is possible to create a hydraulic model of the fluid supply network 1. Obviously the more data about the fluid supply network 1 is available, the more representative the hydraulic model of the fluid supply network 1 can be created.

Model of the fluid supply network 1 must be representative, i.e. represent the real fluid supply network as accurately as possible. The more accurately the fluid supply network 1 is modelled, the more precise the state of the fluid supply network can be determined later.

FIG 2 shows a flow diagram of an embodiment of a method for determining a fluid supply network state. The method comprises at least a training mode I and at least a test mode II. The training mode I can be processed also in off-line regime. At step 7 a plurality of sensor reference data sets for a plurality of predefined states of the fluid supply network 1 are calculated.

Each sensor reference data set comprises a plurality of sensor data that would be provided by the at least one sensor in case the fluid supply network would be in that particular predefined state which corresponds to the respective sensor reference data set.

Since the hydraulic model of the fluid supply system is available, different predefined states of the fluid supply network 1 can be modelled. Therefore the sensor data for each sensor 6 installed in the fluid supply network 1 for the particular predefined state of the fluid supply network 1 can be calculated, i.e. the sensor data that would be provided by each sensor 6 in case the fluid supply network 1 would be in the predefined state.

The sensor reference data sets can be calculated for different predefined states of the fluid supply network 1, for example: for states with different initial parameters, for example, for states with different pressure provided by the supply network pressurizing component 2, and / or state with different consumptions provided by different consumers 5, and for different combinations of such initial parameters, etc. for states with at least one leakage in different locations,' of different sizes, in different times during of day, and for different combinations of such leakages, etc; for states with different initial parameters and / or different leakages present in the fluid supply network 1 for the particular point in time within a day.

The more predefined states of the fluid supply network 1 are modelled, the more precise the actual state can be determined on later steps . The predefined states for which the sensor reference data sets should be calculated are defined by the experts .

In the enhanced embodiment of the method the sensor reference data sets for the fluid supply network states with or without leakages can be calculated.

In particularly at least one no- leakage sensor reference data set should be calculated. The at least one no-leakage sensor reference data set comprises a plurality of sensor data that would be measured by the at least one sensor 6 in case the fluid supply network 1 would be in a corresponding predefined state of the fluid supply network 1 without any leakages in the fluid supply network 1.

There might be more than one no- leakage sensor reference data set, since such no- leakage sensor reference data sets can be calculated, for example, for different points in time and / or for different initial parameters, for example, for different pressure provided by the supply network pressurizing component 2, etc.

In addition to the at least one no-leakage reference data set, a plurality of leakage sensor reference data sets should be calculated. Each such calculated leakage sensor reference data set comprises a plurality of sensor data that would be measured by the at least one sensor 6 in case the fluid supply network 1 would be in a corresponding predefined state of the fluid supply network l with at least one leakage. One sensor reference data set corresponds to one predefined state, and different sensor reference data sets of the plurality of sensor reference data sets correspond to different predefined states of the plurality of predefined states of the fluid supply network.

The different leakage sensor reference data sets correspond to different leakage states of the fluid supply network 1, wherein different leakage states correspond to states of the fluid supply network 1 with leakages in different locations, and / or of different parameters, including of different sizes and / or of different types (unauthorized consumption, holes, etc . ) .

The leakage sensor reference data sets can be calculated for different combinations of leakages. Leakages can be simulated in different ways. The more different leakages and its combinations in different locations in different points in time within a day are modeled, the more precise the leakage can be localized on the later steps. The sensor reference data set might have one data point or more than one data point . The number of data points in the sensor reference data sets depends on the number of sensors 6 installed in the fluid supply network 1 and for which the calculation are performed, and on number of parameters measured by each sensor 6. In preferable case the sensor reference data set should comprise the number of data points at least equal to that number of sensors 6 installed in the fluid supply network 1. In other words the sensor data should be calculated for all sensors 6 installed in the fluid supply network 1.

Each sensor 6 is adapted to provide at least one point of the sensor data. In case sensor 6 is adapted to measure several parameters of the fluid condition on the fluid supply network 1 the sensor 6 can provide several point of sensor data. Since the parameters measured by the at least one sensor 6 are able to provide an observation over the change of a fluid supply network state, the plurality of calculated sensor reference data sets describe such plurality of predefined states of the fluid supply network 1. At step 8 a classifier is trained by using the plurality of calculated sensor reference data sets for the plurality of predefined states of the fluid supply network 1. For example well-known decision tree can be used as a classifier. The training of the classifier is done in such way that the trained classifier is adapted to select and provide a particular state of the fluid supply network as a response to a sensor data set provided to the classifier, wherein the sensor data set comprises the sensor data provided by the at least one sensor.

Within the test mode at step 9 an actual sensor data set is collected, wherein the actual sensor data set comprises the actual sensor data provided by the at least one sensor 6 in a particular point in time for which a state of the fluid supply network 1 should be determined. Therefore as a result of the step 9 the actual sensor data set that describes the actual state of the fluid supply network 1 is collected and available for further determination of what this actual state is.

At step 10 the judgment about the actual state of the fluid supply network 1 is made. The judgment is based on the collected actual sensor data: collected actual sensor data set is provided to the trained classifier, and the trained classifier provides an actual state of the fluid supply network (1) as a response to the collected actual sensor data set provided to the trained classifier. The trained classifier selects the actual state out of the plurality of the predefined states that corresponds to the collected · actual sensor data set.

In the enhanced embodiment of the method, in case the at least one no-leakage reference data set and the plurality of leakage reference data sets are calculated during the step 7, the judgment about the state of the fluid supply network might include a judgment about a location and/or parameter of the at least one leakage .

Such judgment is based on knowledge of the predefined states for which the plurality of sensor reference data sets was calculated and on the fact that the measured parameters and therefore sensor reference data sets provides observation over the change of the fluid supply network state.

FIG 3 shows a flow diagram of an embodiment of a method for determining a fluid supply network state. In addition to the previously described step of calculating the plurality of sensor reference data sets 7 for the plurality of predefined states of the fluid supply network 1, at step 11 historical sensor data with the at least one sensor 6 are collected by taking a plurality of measurements of the parameters measured by the at least one sensor 6 for a given period of time m past, wherein different historical sensor data provided by the at least one sensor corresponds to different point in time.

The measurements of the parameters by the at least one sensor 6 should be done for the given period of time in past with a given time step. The given period of time in past can be for example 1 month, 3 months, 6 months, or 1 year. The given time step can be for example 1 minute, 5 minutes, 15 minutes.

The value of the given time step can be limited by the available equipment, for example by the at least one sensor

6. The given period of time and the given time step are defined by experts.

Further at step 12 the plurality of historical sensor data sets are calculated. Since there are the plurality of the historical sensor data collected within the given period of time with a given time step, it is possible to select all historical sensor data that were collected at the same point in time within a day during the given period of time for the particular sensor. Therefore a historical averaged sensor data for the particular sensor and for the same point of time within a day can be calculated by averaging all historical sensor data that were collected at the same point in time within a day during the given period of time for the particular sensor. Consequently it is possible to calculate the historical sensor data wherein each historical sensor data set comprises at least one historical averaged sensor data which is an average of those collected historical sensor data that corresponds to the same point in time within a day during the given period of time.

Each historical sensor data set corresponds to a particular point in time, and different historical sensor data sets correspond to different particular points in time. Each historical sensor data set might have more than one data point. The number of data points in the historical sensor data sets depends on the number of sensors 6 installed in the fluid supply network and a number of parameters measured by the each sensor 6. The more historical data sets calculated the more precise the real fluid supply network 1 is described.

The number of the historical sensor data sets depends on the given time step with which the historical sensor data were collected. As a result of this step 12 there are available a plurality of the historical sensor data sets that describes the averaged state of the fluid supply network 1 at the particular point in time within a day as the fluid supply network 1 had in the past within the given period of time. Since the historical sensor data are aggregated and averaged, and taking into account the assumption that the fluid supply network 1 is usually on the normal state and not standard / abnormal behavior of the fluid supply network 1 is rather rare case, calculating average for all historical sensor data received within the given period of time in the same point in time within a day allows to eliminate any abnormal sensor data. Further at the step of calculating the plurality of sensor reference data sets (7) for the plurality of predefined states of the fluid supply network 1, while calculating each sensor reference data set, that comprises the plurality of sensor data that would be provided by the at least one sensor in case the fluid supply network would be in that particular predefined state which corresponds to the respective sensor reference data set, sensor data should be reduced by the historical averaged sensor data of the respective historical sensor data sets for the respective sensor 6.

Moreover such reduction should be performed with due regard for the particular point of time the historical data sets and the sensor reference data sets are calculated. As it was mentioned above the sensor reference data set are calculated for the different predefined states wherein the predefined state includes as well the information about the particular point in time the respective sensor reference data set is calculated for. In addition to that the historical sensor data sets are also calculated for the particular points in time within a day.

Therefore within this step 7 of the current embodiment of the method the sensor data of the sensor reference data set calculated for the particular point in time should be reduced by the historical averaged sensor data of the respective historical sensor data set calculated for the same point in time as the sensor reference data set is calculated for.

The reduction should be done for the each sensor 6 installed in the fluid supply network 1. In other words the sensor data of the sensor reference data set for the particular point in time should be reduced by the historical averaged data of the respective historical reference data set calculated for the same sensor 6. After that at the step 8 the classifier should be trained by using plurality of the calculated sensor reference data sets as it was described above .

Further within the test mode II at the step of collecting an actual sensor data set 9 for which a state of the fluid supply network 1 should be determined, the actual sensor data provided by the at least one sensor 6 in a particular point in time should be reduced by the historical average sensor data of the historical sensor data set for the respective point in time for the respective sensor 6.

After that at the step 10 the judgment about the actual state of the fluid supply network 1 is made based on such actual sensor data set as it was described above .

In a possible embodiment of the method the at least one parameter of a fluid condition in the pipe 3, 4 of the fluid supply network 1 measured by the at least one sensor 6 is fluid pressure.

In another possible embodiment when the network of pipes for delivering a fluid to consumers 5 comprises at least one main pipe 3 to transport fluid from a source into the fluid supply network 1, and at least one distribution pipe 4, wherein the at least one distribution pipe 4 is adapted to transport fluid from the main pipe 3 to a consumer 5 fluidly connected to the distribution pipe (4), the at least one sensor 6 is installed on a distribution pipe 4.

In a possible embodiment the fluid supply network is a water supply network.

FIG 4 shows a system 13 for determining a fluid supply network state according to the present invention. The system 13 as it was described above on the FIG 1 comprises at least one sensor 6 wherein each pressure sensor 6 is adapted to be located on a pipe 3 , 4 of the network of pipes . Each pressure sensor 6 is adapted to to make measurement of at least one parameter of a fluid condition in the pipe 3, 4 of the fluid supply network 1 where the at least one sensor 6 is located, wherein measuring such at least one parameter is able to provide an observation over change of a fluid supply network state, and to provide sensor data, which comprises information about the measured parameter,

The system 13 also comprises a control unit 14 configured to determine a fluid supply network state in accordance with a method of any one of the claims 1 to 7.

While the invention has been illustrated and described in detail with the help of preferred embodiment, the invention is not limited to the disclosed examples. Other variations can be deducted by those skilled in the art without leaving the scope of protection of the claimed invention.

Reference numerals

1 - fluid supply network

2 - supply network pressurizing component

3 - main pipe

4 - distribution pipe

5 - consumer

6 -sensor

7 - 12 - method steps

13 - system

14 - control unit