DE102019213530A1 - Method and device for operating a water distribution system - Google Patents

Abstract

The invention relates to a method for operating a water distribution system (1), with the following steps:
- Provision (S1, S2) of operating data and / or demand data for the water distribution system (1);
- Determination (S3) of a manipulated variable specification (S) for operating a pumping station (3) with the aid of an operating model as a function of the operating data and / or the demand data;
- Operating (S5) at least one pump of the pumping station (3) as a function of the manipulated variable specification (S) in order to provide a pump power; characterized in that the operating model comprises a data-based trainable model, in particular a Gaussian process model, which is trained to provide the manipulated variable specification (S) to be provided as a function of the operating data and / or the demand data in such a way that pressure fluctuations at the tapping points of the pipeline network are reduced.

Description

Technical area

The invention relates to water distribution systems, and in particular to measures for an energy- and water-consumption-efficient operation of a water distribution system.

Technical background

Drinking water is distributed to households or production facilities in a city or town via branched water pipe systems. A water supply system usually includes a water supply station, a water storage tank, one or more pumping stations or pressure boosting stations and a network with water pipes and taps, usually in the connected households and production facilities. In order to ensure a trouble-free supply of drinking water, it is necessary that there is always a water pressure that ensures that water can be drawn from the taps. At the same time, this water pressure must not exceed a maximum so that there is no damage to the lines and / or energy or water losses (pressure reducing valves).

Disclosure of the invention

According to the invention, a method for operating a water distribution system according to claim 1 and a device for operating a water distribution system and a water distribution system according to the independent claims are provided.

Further refinements are given in the dependent claims.

• - Bereitstellen von Betriebsdaten und/oder Bedarfsdaten für das Wasserverteilungssystem ;
• - Ermitteln einer Stellgrößenangabe zum Betreiben einer Pumpstation mithilfe eines Betriebsmodells abhängig von den Betriebsdaten und/oder den Bedarfsdaten;
• - Betreiben mindestens einer Pumpe der Pumpstation abhängig von der Stellgrößenangabe, um eine Pumpleistung bereitzustellen;

dadurch gekennzeichnet, dass das Betriebsmodell ein datenbasiertes trainierbares Modell, insbesondere ein Gaußprozessmodell, umfasst, das so trainiert ist, um die Stellgrößenangabe abhängig von den Betriebsdaten und/oder den Bedarfsdaten so bereitzustellen, dass Druckschwankungen an Zapfstellen des Leitungsnetzes reduziert werden.According to a first aspect, a method for operating a water distribution system is provided, comprising the following steps:

• - Provision of operating data and / or demand data for the water distribution system;
• Determination of a manipulated variable specification for operating a pumping station with the aid of an operating model as a function of the operating data and / or the demand data;
• - Operating at least one pump of the pumping station depending on the manipulated variable information in order to provide a pump output;

characterized in that the operating model comprises a data-based trainable model, in particular a Gaussian process model, which is trained to provide the manipulated variable information depending on the operating data and / or the demand data in such a way that pressure fluctuations at the tapping points of the pipeline network are reduced.

To ensure that there is sufficient pressure in a water distribution system at all times, pumping stations are conventionally provided on the one hand, which can convey a sufficient amount of water into the pipeline network in order to maintain the necessary pipeline pressure and flow even when there is a high water withdrawal. On the other hand, one or more pressure reducing valves are provided which limit the pressure in the pipe network by draining water.

Pumping water, which is drained via the pressure reducing valves, into the pipe network against a high pressure is of no use for the provision of water at the draw-off points. The energy loss resulting therefrom due to the pump power to be applied is already very high in conventional water distribution systems, and it would therefore be desirable to operate water distribution systems in such a way that the energy loss is reduced.

The water distribution system is essentially operated by controlling the pumping power of at least one pumping station. The pumping station enables a quantity of water to be conveyed in order to provide a specific pressure in the pipe network of the water distribution system. There are usually different pressure zones in water distribution systems, with a pressure of around 2 bar and 4 to 6 bar at transition points usually being provided at the tapping points. An overpressure above 10 bar, for example, should be avoided in order to avoid damage to the water pipe system (pipes) or other facilities. The water pressure in the pressure zones is regulated via the pressure reducing valves. 90% of the energy consumption in a water distribution system is accounted for by pumping, with the energy loss caused by draining water through the pressure reducing valves being around 30 to 40%.

Furthermore, depending on the dispensing behavior in the household, dynamic pressure fluctuations arise in the line network, which cannot be sufficiently compensated by previous devices, since these devices cannot be adapted to changes in the system.

The above method provides for the use of a prediction system that processes the operating data and demand data with the aid of an operating model in order to provide the best possible pump performance. This pump power can thus predictively take into account the expected water demand in a water distribution system, so that high water withdrawals do not lead to the pressure falling below the required minimum at one point in the water distribution system.

The operating model can be designed as a trainable, data-based model, in particular as a Gaussian process model, which uses the available operating data and demand data to determine a manipulated variable, in particular as a currently set pump output of a pumping station.

By using the demand data, the operating model becomes predictive. The demand data contain data that correlates with the water demand, such as time of day data, weekday and season information (calendar data), which can be taken into account. The operating model can increase the pumping capacity and vice versa in anticipation of increased water withdrawal from the water distribution system. In this way, a pressure drop below a predetermined minimum pressure at a specific point in the water distribution system can be avoided. Furthermore, the frequency of an overpressure or a drainage of water via the pressure reducing valves in the event of a reduction in the water withdrawal can be reduced by predictive throttling of the pump output in order to avoid draining water via a pressure reducing valve.

By using the operating model, on the one hand, the pressure fluctuations within the water distribution system can be reduced and, on the other hand, falling below a minimum pressure can be prevented and the frequency of draining water via the pressure reducing valves can be reduced.

Furthermore, the at least one pump can continue to be operated at the operating point specified by the operating data and / or the demand data as a function of an uncertainty statement regarding the manipulated variable specification.

It can be provided that the operating data include pressures and / or flow rates and / or water temperatures at at least one sensor position in a line network of the water distribution system.

• - die Tageszeit, den Wochentag und/oder die Jahreszeit,
• - der Wetterdaten, insbesondere der Tages- und/oder Nachttemperatur und/oder einer Wolkensituation und/oder einer Regenwahrscheinlichkeit,
• - eine Angabe über anstehende Ereignisse, die Einflüsse auf die Wasserentnahme aus dem Leitungsnetz haben.

Furthermore, the demand data can include one or more of the following information about external conditions:

• - the time of day, the day of the week and / or the season,
• - the weather data, in particular the day and / or night temperature and / or a cloud situation and / or a probability of rain,
• - an indication of upcoming events that have an impact on water withdrawal from the pipeline network.

Furthermore, the data-based trainable model can be trained to provide the manipulated variable information depending on the operating data and / or the demand data so that a required minimum pressure is not undershot at a point in the water distribution system and / or that a predetermined maximum pressure is not exceeded.

• - Bereitstellen von Betriebsdaten und/oder Bedarfsdaten für das Wasserverteilungssystem ;
• - Ermitteln einer Stellgrößenangabe zum Betreiben einer Pumpstation mithilfe eines Betriebsmodells abhängig von den Betriebsdaten und/oder den Bedarfsdaten;
• - Betreiben mindestens einer Pumpe der Pumpstation abhängig von der Stellgrößenangabe, um eine Pumpleistung bereitzustellen;

wobei das Betriebsmodell ein datenbasiertes trainierbares Modell, insbesondere ein Gaußprozessmodell, umfasst, das so trainiert ist, um die Stellgrößenangabe abhängig von den Betriebsdaten und/oder den Bedarfsdaten so bereitzustellen, dass Druckschwankungen an Zapfstellen des Leitungsnetzes reduziert werden.According to a further aspect, a device, in particular a control unit, is provided for operating a water distribution system, the device being designed to:

• - Provision of operating data and / or demand data for the water distribution system;
• Determination of a manipulated variable specification for operating a pumping station with the aid of an operating model as a function of the operating data and / or the demand data;
• - Operating at least one pump of the pumping station depending on the manipulated variable information in order to provide a pump output;

wherein the operating model comprises a data-based trainable model, in particular a Gaussian process model, which is trained to provide the manipulated variable information depending on the operating data and / or the demand data in such a way that pressure fluctuations at tapping points of the pipeline network are reduced.

According to a further aspect, a water distribution system is provided with a line network, a pumping station for providing water in the line network and the above device for controlling the pumping station for providing the pump power.

Figure list

• 1 eine schematische Darstellung eines beispielhaften Wasserverteilungssystems;
• 2 ein Verfahren zum Betreiben eines Wasserverteilungssystems der 1; und
• 3 ein Verfahren zum Trainieren eines Betriebsmodells für ein Wasserverteilungssystem der 1.

Embodiments are explained in more detail below with reference to the accompanying drawings. Show it:

• 1 a schematic representation of an exemplary water distribution system;
• 2 a method for operating a water distribution system of the 1 ; and
• 3 a method for training an operating model for a water distribution system of 1 .

Description of embodiments

1 shows a water distribution system 1 with a water supply station 2 for the provision of drinking water from a water reservoir or a water treatment, a pumping station 3 for pumping drinking water under a predetermined pressure, a pipe network 4th for the distribution of drinking water with water pipes 41 and taps 5 for the withdrawal of drinking water by consumers. The line network also has pressure reducing valves 6th to the pressure in the relevant water pipe at certain points in the pipe network 41 to limit to a predetermined maximum pressure. The pipeline network shown is only an example of existing water distribution networks and can be expanded as required in terms of its size and type of branches. The topology of the line network 4th can be diverse, and it can be a conventional line network branching into branches, a ring line network or a mesh line network.

The water distribution system 1 is operated by a pump in the pumping station 3 is controlled with a certain pump power, so that drinking water from the water supply station 2 in the pipeline network 4th the water distribution system 1 is promoted. The drinking water should be in the pipeline network 4th be provided under increased pressure so that at the tapping points 5 the drinking water can be withdrawn at a pressure that does not fall below a minimum pressure.

In the pipeline network 4th are still condition sensors 7th arranged the information on the in the line network 4th existing pressures and / or flow rates at the sensor positions of the condition sensors 7th specify.

For pipeline networks, nominal line pressures of 2 bar are usually required at the tapping points 5 and 4th to 6th bar is optimal at transition points, whereby line pressures of over 10 bar should be avoided for reasons of line protection. For controlling the pressure distribution in the pipe network 4th becomes the pump of the pumping station 3 controlled with a predetermined pump power which has a safety tolerance in order to ensure that the pressure fluctuations are minimized and the minimum pressures are not undershot regardless of the dispensing behavior of consumers. The pump output then causes the line pressures to be regulated with the pressure reducing valves 5 that are used to set the nominal line pressure for drinking water from the mains 4th Drain into the sewer system.

To protect the pipeline network 4th are pressure reducing valves 6th provided distributed, the drinking water from the mains at an excess pressure above the specified target line pressure 4th drain. The distribution of the pressure reducing valves 6th should be provided in such a way that in no area of the pipeline network 4th Line pressures above the specified nominal line pressures can arise. However, the pressure reducing valves are distributed 6th empirically, so that, depending on the dynamics of the dispensing behavior, pressure fluctuations in the pipeline network 4th may occur.

Depending on the tapping behavior, ie depending on the water drawn from the taps 5 , it comes in the pipeline network 4th to time-varying pressure distributions, which may lead to pressure fluctuations that are disruptive for a consumer. In addition, areas in the line network can be due to dynamic effects 4th temporarily have water pressures below a predetermined minimum pressure, so that water can be withdrawn from certain draw-off points 5 is deteriorated, ie has a lower flow rate or is no longer possible under certain circumstances.

To avoid this, the pressure fluctuations in the pipe network 4th , especially the pressure fluctuations at the tapping points 5 , can be reduced by increasing the pump output, however this can lead to more drinking water through the pressure reducing valves 6th which leads to a loss of energy and water.

To operate the water distribution system 1 with regard to energy consumption and pressure fluctuations at the draw-off points 5 to improve becomes a control unit 10 provided in such a way that they can operate the water distribution system using an operating model 1 predictive controls. For this purpose, the operating model is designed in such a way that it determines a manipulated variable specification S as input variables as a function of operating data and demand data.

The operating model is designed as a data-based trainable model, in particular as a Gaussian process model. A Gaussian process model represents a universal function approximator. In general, a Gaussian process model represents a function whose function values can only be modeled with probabilities due to incomplete information. With the help of functions of the expected values, variances and covariances, a Gaussian process describes the function values as a continuum of correlated random variables in the form of a normal distribution. This enables a regression even with input variables with errors or tolerances and stochastically distributed model variables.

The manipulated variable specification S is used to set the pump output of the pump. This can be done directly by entering the manipulated variable specification S the Pump output corresponds, or indirectly, by specifying a setpoint pressure for a line position immediately downstream of the pump, to which the pump output is regulated.

The operating data include pressures and / or flow rates from the state sensors 7th in the pipeline network 4th be measured. In addition, water temperatures can be measured at the sensor positions of the condition sensors 7th recorded and taken into account in the operating data.

• - die Tageszeit, den Wochentag und/oder die Jahreszeit (in Form einer Angabe des Monats oder des Kalendertags),
• - der Wetterdaten, wie z.B. der Tages und Nachttemperatur, Wolkensituation und dergleichen, einer Regenwahrscheinlichkeit,
• - eine Angabe über anstehende Ereignisse, wie beispielsweise von Festen, Großveranstaltungen und dergleichen, die Einflüsse auf die Wasserentnahme aus dem Leitungsnetz 4 haben.

The demand data can include one or more of the following information about external conditions:

• - the time of day, the day of the week and / or the season (in the form of an indication of the month or calendar day),
• - the weather data, such as day and night temperature, cloud situation and the like, a rain probability,
• - an indication of upcoming events, such as festivals, major events and the like, the influences on water abstraction from the pipeline network 4th to have.

In this way, fluctuations in water consumption can be assigned on the basis of the demand data. For example, higher water consumption can be assigned to warm weather.

By taking into account demand data, time, weather and event-dependent predictions can be carried out so that the pump output can be adapted via the manipulated variable specification S in accordance with the expected water demand. For example, the water pressure in the pipe network 4th before a high water withdrawal is expected to be increased within the permissible range (without exceeding a specified maximum pressure) in order to avoid an excessive drop in the water pressure at the time of the high water withdrawal.

In addition, with a state-based setting of the pump output, the short-term fluctuations can be avoided by monitoring the pressures and flow rates in the pipeline network 4th are taken into account and lead directly to an intervention in the pumping capacity.

Operation of the control unit 10 takes place by cyclically calculating the manipulated variable specification S by the operating model. A cycle time can, for example, be in the range from 0.1 to 20 s, preferably between 1 and 10 s and in particular 1 s. Alternatively, the operating model can also be formed by adding a number of partial operating models to it. The partial operating models can, for example, process the operating data and the demand data separately and add the corresponding partial pump outputs to a total pump output. This has the advantage that fewer data points are required to train the partial operating models.

The operating data are used to regulate the pressure in the pipeline network 4th and the demand data for providing a sufficient amount of water at the tapping points 5 . For this purpose, it can also be provided that the partial operating model processing the operating data is designed as a conventional control or regulation and only the predictive partial operating model that processes the demand data is provided as a data-based trainable model.

In conjunction with the flow chart of the 2 becomes a method of operating the water distribution system 1 described. The method can be designed as an algorithm in a microprocessor or some other data processing device as software and / or in hardware.

In step S1 are operating data on pressures and / or flow rates and, if applicable, the water temperature in the water distribution system 1 provided.

In step S2 demand data is provided, which provides information about the time of day, day of the week, season, upcoming events, weather conditions and other data affecting water consumption.

In step S3 the operating model is evaluated based on the operating data and the demand data and a manipulated variable specification S for controlling the pump of the pumping station 3 determined. The evaluation can take place, for example, by determining an expected value of the operating model designed as a Gaussian process model.

In step S4 an uncertainty specification of the manipulated variable specification S is determined at the evaluation point (operating data and demand data) of the operating model. When using a Gaussian process model for the evaluation point under consideration, the uncertainty specification or a variance can be determined in a simple manner known per se with the aid of the covariance matrix.

In step S5 the pump power is set directly or indirectly (by pressure control) based on the manipulated variable specification S and the uncertainty specification. The uncertainty information can lead to an increase in the pump power to be set in order to take into account an uncertainty in the model prediction of the operating model.

Then becomes step S1 jumped back to a cyclical operation of the water distribution system 1 to ensure.

In the flow chart of the 3 a method for training the operating model is presented.

In step S11 the relevant input variables are selected from the operating data and demand data.

Then in step S12 Training data is generated which assign the selected or relevant input variables to the most optimal manipulated variable information possible. For this purpose, the measurement is based on a specific withdrawal situation occurring at the moment and a suitable, optimal control variable specification is determined, e.g. by manual or automatic variation of the manipulated variable specification S with simultaneous monitoring of the continuity of the water outflow from the open taps and compliance with the pressure and flow conditions within the pipeline network (with the aid of the condition sensors).

In particular, the operating model can then be trained based on a pressure at the reference point or on a pump power which leads to a sufficient pressure or flow rate at all open tapping points. This can be done by adaptively adjusting the pump output during a specific dispensing profile. In this way, measurement points are obtained that can be used to train the operating model, which can be designed as a Gaussian process model. For the mapping of a predictive behavior, the effects of an operating point defined by operating data and demand data on water supply can be determined for a period in the future. This period of time can be between 0.5 and 30 seconds, for example. In this way dynamic effects can be mapped.

In step S13 the Gaussian process model is trained based on the determined training data and is thus available for use in operating the water distribution system.

If partial operating models are provided with regard to the operating data and with regard to the demand data, the partial operating model can be recorded with regard to the demand data of the behavior model, different times of the day, days of the week, seasons, events and the like flow rates at the reference point or pressures at the reference point, with the recording at a given pump capacity can be done. Due to fluctuations in the pressure at the reference point, the different withdrawal behaviors can be determined and used in a corresponding additive contribution to the target pressure at the reference point or the pumping capacity with the aid of the behavior model, which is also designed as a Gaussian process model.

Alternatively, topographical data such as line length, line diameter, line material, line age, type of line network and the like can also be taken into account as input variables for the operating model, so that operating models that have been determined for other line networks can also be used. The transferability of the invention described here to other line networks (e.g. branched, ring or mesh line networks) is made possible by the sequencing of these line networks in simply structured networks (as exemplified in 1 shown) feasible.

Claims (10)

Method for operating a water distribution system (1), comprising the following steps: - providing (S1, S2) operating data and / or demand data for the water distribution system (1); - Determination (S3) of a manipulated variable specification (S) for operating a pumping station (3) with the aid of an operating model as a function of the operating data and / or the demand data; - Operating (S5) at least one pump of the pumping station (3) as a function of the manipulated variable specification (S) in order to provide a pump output; characterized in that the operating model comprises a data-based trainable model, in particular a Gaussian process model, which is trained to provide the manipulated variable information (S) as a function of the operating data and / or the demand data in such a way that pressure fluctuations at the tapping points of the pipeline network are reduced. Procedure according to Claim 1 wherein the at least one pump continues to be operated as a function of an uncertainty information about the expected value of the manipulated variable information (S) at the operating point specified by the operating data and / or the demand data. Method according to one of the Claims 1 and 2 , the operating data including pressures and / or flow rates and / or water temperatures at at least one sensor position in a line network (4) of the water distribution system (1). Method according to one of the Claims 1 to 3 , where the demand data includes one or more of the following information about external conditions: the time of day, the day of the week and / or the season, - the weather data, in particular the daytime and / or nighttime temperature and / or the cloud situation and / or a probability of rain, - an indication of upcoming events that have an impact on water withdrawal from the pipeline network . Method according to one of the Claims 1 to 4th , the operating model being trained to provide a manipulated variable information (S) predictively as a function of the demand data, in order to take into account an expected water withdrawal state. Method according to one of the Claims 1 to 5 , the data-based trainable model being trained to provide the manipulated variable information (S) as a function of the operating data and / or the demand data in such a way that a required minimum pressure is not undershot at a point in the water distribution system (1) and / or that a predetermined maximum pressure is not exceeded is exceeded. Device, in particular a control unit (10), for operating a water distribution system (1), the device being designed to: - provide operating data and / or demand data for the water distribution system (1); - Determination of a manipulated variable specification (S) for operating a pumping station (3) with the aid of an operating model as a function of the operating data and / or the demand data; - Operating at least one pump of the pumping station (3) as a function of the manipulated variable specification (S) in order to provide a pump output; characterized in that the operating model comprises a data-based trainable model, in particular a Gaussian process model, which is trained to provide the manipulated variable information (S) as a function of the operating data and / or the demand data in such a way that pressure fluctuations at the tapping points of the pipeline network are reduced. Water distribution system (1) with a pipe network (4), a pumping station (3) for providing water in the pipe network (4) and a device according to Claim 7 to control the pumping station to provide the pumping power. Computer program which is set up to carry out all steps of a method according to one of the Claims 1 to 8th to execute. Electronic storage medium on which a computer program according to Claim 11 is stored.

Images