Method for improved control of a technical system by means of a technical component

Abstract

The present invention relates to a method (100) for improved control of a technical system (40) by means of a technical component (30), having the following steps: - detecting (102) first data (32) relating to the technical component (30) by means of a data control unit (50), wherein the first data (32) are used to determine a characteristic operating behavior of the technical component (30); - transmitting (104) the first data (32) via a first interface (57) to a first cloud-based data processing unit (70); - generating (106) second data (33) by processing the first data (32) in the first cloud-based data processing unit (70); - transmitting (108) the generated second data (33) via a second interface (75) to a second cloud-based data processing unit (80); - processing (110) the second data (33) using a data-based algorithm (84) stored in the second cloud-based data processing unit (80), in order to generate third data (34); - transmitting (112) the generated third data (34) via a fourth interface (87) to the data control unit (50); and - at least partially replacing (114) the preset first data (32) relating to the technical component (30) with the generated third data (34) in order to thus improve the control of the technical system (40) by means of the technical component (30).

Description

[0001] Description

[0002] Title:

[0003] Method for improved control of a technical system by means of a technical component

[0004] The present invention relates to a method for improved control of a technical system by means of a technical component.

[0005] State of the art

[0006] In a fuel cell, electrical energy is generated from hydrogen and oxygen. To ensure safe and robust operation of the fuel cell, a control unit (FCCU) continuously evaluates a multitude of sensor values. Based on these sensor values ​​and a suitably selected and applied algorithm, the actuators are then activated, which positively influence the physical behavior of the fuel cell.

[0007] For example, a hydrogen injection valve (HGI) is appropriately supplied with current to refill the required amount of hydrogen into the fuel cell's anode. This must be done as precisely as possible, since too much hydrogen unnecessarily increases the pressure in the anode and thus leads to mechanical stress on the membrane. Too little hydrogen, on the other hand, leads to too low a pressure, which also places mechanical stress on the membrane and can also cause cell-damaging hydrogen depletion at the catalytic layer.

[0008] To optimally adjust the existing control and regulation algorithm in the control unit, the HGI valve is measured and the characteristic data is stored in the control unit in the form of a mathematical model. This mathematical model allows both the feedforward control and the closed-loop control to be adapted to the specific system, allowing the hydrogen supply to be adjusted quickly and robustly.

[0009] In this regard, DE 10 2021 203 798 A1 discloses, by way of example, a method for controlling a fuel cell, in which the control device is connected to a network in order to exchange corresponding control commands for controlling an operating behavior of the fuel cell.

[0010] Disclosure of the invention

[0011] It is therefore the object of the present invention to provide a solution to operate a fuel cell system efficiently and in an improved manner.

[0012] This object is achieved by a method for improved control of a technical system by means of a technical component having the features of the independent claim.

[0013] According to a first aspect, the disclosure relates to a method for improved control of a technical system by means of a technical component.

[0014] In a first step, initial data of the technical component is collected by a data control unit, whereby the first data are used to determine a characteristic operating behavior of the technical component.

[0015] In a second step, the first data is transmitted via a first interface to a first cloud-based data processing unit.

[0016] In a third step, second data is generated by processing the first data in the first cloud-based data processing unit. In a fourth step, the generated second data is transmitted via a second interface to a second cloud-based data processing unit.

[0017] In a fifth step, the second data is processed by applying a data-based algorithm stored in the second cloud-based data processing unit to generate third data.

[0018] In a sixth step, the generated third data is transmitted to the data control unit via a fourth interface.

[0019] In a seventh step, the preset first data of the technical component are at least partially replaced by the generated third data in order to improve the control of the technical system with the technical component. This at least partial replacement can be carried out, for example, by the data control unit. Thus, for the control of the technical component,

[0020] • to the extent that third party data is available, that third party data, and

[0021] • To the extent that no third data is available, the originally recorded first data is used. For example, processed values, variables, and / or parameters can be generated as third data only for a subset of the total values, variables, and / or parameters recorded and / or preset in the form of the first data.

[0022] A fundamental idea of ​​the present invention is that the control unit (FCCU) for controlling the fuel cell uses the data required to determine the HGI

[0023] Data characteristic of the behavior is first recorded. This recorded data is then transmitted to a computer network (cloud) via a suitable data connection. There, this data is processed using a suitable algorithm to determine or optimize the parameters of the mathematical model that describes the characteristic behavior of the HGI valve. These adjusted parameters are then sent back to the control unit via another interface, where they are used to control and pre-control the HGI valve instead of the previously set or recorded values.

[0024] The method according to the invention offers the following advantages:

[0025] Any change in the behavior of the HGI valve is detected, and the control is automatically adjusted in the control unit. This ensures consistent control quality over the entire runtime.

[0026] By processing the collected data in the cloud, significantly more memory and computing power is available than with an algorithm in the control unit. This also enables the use of complex optimization algorithms. Furthermore, parameters can be manually approved by an experienced engineer at an early stage of development, eliminating the need for complex monitoring algorithms in the early stages of development.

[0027] Furthermore, the method according to the invention and the corresponding system architecture for implementing this method can significantly accelerate the commissioning of new systems. Furthermore, drift behavior of a component can be detected and visualized during the testing phase, as optimization into a mathematical behavior model results in a very strong compression of the acquired data. Last but not least, component drift can be detected and compensated for in systems after the start of production, thus increasing the service life of the product. In the event of drift outside the specification, an error or deviation from a preset normal state can also be quickly detected, and an appropriate reaction or measure to correct the detected deviation can be initiated.

[0028] Furthermore, the method according to the invention, or with a corresponding architecture, can also be applied to a large fleet of vehicles, since the required computing capacity can be dynamically scaled as needed via the virtual data processing units. One possible embodiment of the method provides that the first data are measurable variables that are influenced by the technical component and / or influence the technical component, so that the first data describe the characteristic operating behavior of the technical component. This provides the advantage that the operating behavior of the technical component is mapped in an application-oriented manner and dependent on the respective area of ​​application.

[0029] One possible embodiment of the method provides for the initial data to represent parameters of a mathematical model of the technical component that describes the characteristic operating behavior of the technical component. This has the advantage of mapping the operating behavior of the technical component in an application-oriented manner and dependent on the respective application area.

[0030] One possible embodiment of the method provides for the processing of the initial data to include at least one of the following measures: error and completeness checking, interference removal, and data format conversion. This provides the advantage of creating the most accurate possible representation of the actual operating behavior of the technical component, allowing for efficient control.

[0031] One possible embodiment of the method provides that the processing of the second data into third data involves providing adapted or optimized parameters for the mathematical model of the technical component from the second data in the form of the third data. This achieves the advantage of creating the most accurate possible representation of the actual operating behavior of the technical component.

[0032] One possible embodiment of the method provides for the processing of the second data into third data using a virtual data processing unit installed in the second cloud-based data processing unit, which receives the second data from the first cloud-based data processing unit and accesses the data-based algorithm to process the second data. This provides the advantage of enabling efficient and data security-compliant data processing. Furthermore, this architecture also allows the method to be applied to a large fleet of vehicles, since the required computing capacity can be dynamically scaled as needed via the virtual data processing units.

[0033] One possible embodiment of the method provides that the virtual data processing unit can be connected to an output unit via a third interface in order to output the generated third data. This provides the advantage that the generated data can be easily made available for verification.

[0034] One possible embodiment of the method provides for the technical system to be designed as a fuel cell or a fuel cell stack, which is controlled via a hydrogen injection valve (HGI) designed as a technical component. This offers the advantage of always achieving optimal and application-adapted control of the fuel cell or fuel cell stack.

[0035] According to a second aspect, the disclosure relates to a computer program containing machine-readable instructions which, when executed on one or more computers and / or computer instances, cause the computer or computer instances to carry out the method according to the invention.

[0036] According to a third aspect, the disclosure relates to a machine-readable data carrier and / or download product comprising the computer program.

[0037] According to a fourth aspect, the disclosure relates to one or more computers and / or compute instances with the computer program, and / or with the machine-readable data carrier and / or the download product. Further measures improving the invention are presented in more detail below, together with the description of the preferred embodiments of the invention, with reference to figures.

[0038] Examples of implementation

[0039] It shows:

[0040] Figure 1 Schematic flow diagram of the method 100 for improved control of a technical system 40 by means of a technical component 30 according to an embodiment of the present invention; and

[0041] Figure 2 Schematic architecture for implementing the method 100 according to an embodiment of the present invention.

[0042] Figure 1 shows a schematic flow diagram of the method 100 for improved control of a technical system 40 by means of a technical component 30.

[0043] The technical system 40 can preferably and exemplarily be designed as a fuel cell or as a fuel cell stack, which is controlled via a hydrogen injection valve (HGI) designed as a technical component 30. However, the present invention is applicable to different fields of application and a wide variety of technical applications.

[0044] In step 102, first data 32 of the technical component 30 is acquired by a data control unit 50, wherein a characteristic operating behavior of the technical component 30 is determined using the first data 32.

[0045] Optionally, the measurable quantities that are influenced by the HGI and / or influence the HGI are transmitted. Optionally, the first data 32 are measurable quantities that are influenced by the technical component 30 and / or influence the technical component 30, so that the first data 32 describe the characteristic operating behavior of the technical component 30.

[0046] In step 104, the first data 32 are transmitted via a first interface 57 to a first cloud-based data processing unit 70 (see Fig. 2).

[0047] In step 106, second data 33 is generated by processing the first data 32 in the first cloud-based data processing unit 70.

[0048] Optionally, the processing of the first data 32 may include at least one of the following measures: error and completeness checking, cleaning of interference signals, data format conversion.

[0049] In step 108, the generated second data 33 is transmitted via a second interface 75 to a second cloud-based data processing unit 80.

[0050] In step 110, the second data 33 is processed by applying a data-based algorithm 84 stored in the second cloud-based data processing unit 80 to generate third data 34.

[0051] Optionally, the processing of the second data 33 into third data 34 may consist of providing adapted or optimized parameters for the mathematical model 31 of the technical component 30 from the second data 33 in the form of the third data 34.

[0052] Furthermore, the processing of the second data 33 into third data 34 can be performed using a virtual data processing unit 82 installed in the second cloud-based data processing unit 80. The virtual data processing unit 82, for example, a cloud application that can be configured with appropriate computing or data processing capacity depending on the application, receives the second data 33 from the first cloud-based data processing unit 80 and accesses the data-based algorithm 84 to process the second data 33.

[0053] The virtual data processing unit 82 can further be connected to an output unit 86 via a third interface 85 in order to output the generated third data 33.

[0054] In step 112, the generated third data 34 is transmitted via a fourth interface 87 to the data control unit 50.

[0055] In step 114, the preset first data 32 of the technical component 30 is at least partially replaced by the generated third data 34 in order to improve the control of the technical system 40 with the technical component 30.

[0056] Figure 2 (in combination with Figure 1) shows a schematic architecture for implementing the method 100 according to an embodiment of the present invention.

[0057] The method 100 according to the invention can be used for improved control of a technical system 40, for example a fuel cell or a fuel cell stack, by means of a technical component 30, for example an HGI valve.

[0058] For this purpose, the technical system 40 is connected to a data control unit 50, which can be configured as a control device. The data control unit 50 acquires initial data 32 that determines a characteristic operating behavior of the technical component 30. This initial data 32 could, for example, be:

[0059] Print before the HGI,

[0060] Current at the stack, which is converted into the flow at the HGI and

[0061] Power to control the HGI. This first data 32 is transmitted—optionally by means of a data logger 55, which acts as a data provision unit—via a first interface 57 to a first cloud-based data processing unit 70, which can be configured as a computer network.

[0062] This first cloud-based data processing unit 70 is designed as a first area, the so-called data landing zone, with lower security requirements than the second cloud-based data processing unit 80, which can be regarded as a second area, the so-called data processing zone.

[0063] In this example, this first cloud-based data processing unit 70 consists of a data storage unit 72 and a data processing unit 74.

[0064] The first data 32 is stored in the data memory 72.

[0065] In the data processing unit 74, the first data 32 is processed to form second data 33: The first data 32 can be checked for data quality, the first data 32 can be cleaned, and the first data 32 can be converted into a data format suitable for further processing.

[0066] The generated second data 33 of the first area are then transmitted to the second cloud-based data processing unit 80 via a second interface 75.

[0067] In the example of Figure 2, the second cloud-based data processing unit 80 includes a virtual data processing unit 82 and a data-based algorithm 84.

[0068] To prevent unauthorized access to this algorithm 84, the two areas 70, 80 are separated from each other, for example, by a technical security structure such as a firewall. Direct external access to the algorithm is therefore generally not possible, which increases the security of the system. The algorithm 84 in the second cloud-based data processing unit 80 serves to process and interpret the received second data 33 in order to generate third data 34 therefrom. The third data 34 represents updated parameters of the mathematical model 31 of the technical component 30, which describes the characteristic operating behavior of the technical component 30.

[0069] The generated third data 34 are then transmitted via a fourth interface 87 to the data control unit 50, optionally to the data logger 55.

[0070] Subsequently, the preset first data 32 of the technical component 30 are at least partially replaced by the generated third data 34 in order to improve the control of the technical system 40, the fuel cell, with the technical component 30, the HGI valve.

[0071] As further shown in Figure 2, the virtual data processing unit 82 can optionally be connected via a third interface 85 to an output unit 86, for example, a monitor, a tablet, or a smart device, in order to output the generated third data 33 for review. This can be of safety-relevant importance because, in the present invention, the generated third data 34 are fed directly into the control unit 50 and thus directly intervene in or influence the control of the technical system 40 via the control of the component 30.

[0072] Optionally, the second cloud-based data processing unit 80 can include a so-called watchdog, a type of data arrival recorder. The watchdog detects when new data arrives in the second area 80, the so-called data processing zone, and triggers processing. For this purpose, for example, a container can be started in an environment containing an Octave kernel.

[0073] Suitable environments include Databricks clusters, Batch pools, Kubernetes / Docker servers, or simply Linux VMs. This container loads both the newly received data and the algorithm, for example, in the form of an Octave file. The algorithm describes how the parameters of the HGI valve model can be determined from the measured values. This algorithm massively compresses the data. For example, in the HGI valve example, only four parameters are determined from measurement files containing several kilobytes of data.

[0074] These parameters, as well as some characteristic variables of the measurements, can be visualized in the cloud via a dashboard (based on Tableau, Plotly, Dash, or Seaborn). This allows for the rapid detection of outliers, drift, and malfunctions in the algorithm.

[0075] After checking the result, which can be done manually in an early phase and then implemented using a suitable algorithm in a later phase, the data is transferred back to the control unit (= FCCU) 50 via a suitable data connection and a data logger.

[0076] In the control unit 50, the data is either stored in RAM (working page) during an early development phase or transferred to flash memory (reference page) at a later stage (e.g., after SOP). The changed parameters influence the model-based controller in the control unit, since both the pilot control behavior and the behavior of the closed control loop depend on the mathematical model of the HGI valve 30. This adapts the control behavior to the changed behavior of the HGI valve 30, allowing a consistent quality of the system behavior 40 to be achieved.

[0077] In the control unit 50, an equation is evaluated that calculates the manipulated variable for controlling the HGI valve (current = x) as a function of the desired flow rate y and under the given pressure conditions p (e.g., pressure upstream of the HGI valve), i.e., x = f(y, p). The function f can combine the variables y and p in a linear or nonlinear form to produce the result x.

[0078] To estimate the function f (y, p), data is now transmitted from the control unit 50 to the cloud, for example, via mobile communications (4G or 5G). In this case, the current, flow, and pressure are transmitted for many individual measurements / samples.

[0079] It is important that triplets of the three variables are transmitted, meaning that the cloud can identify which values ​​were recorded at the same time / sampling step. Furthermore, status information can be transmitted to indicate that these values ​​were recorded in a relevant operating state, for example, that no specific manual application interventions were performed or that a certain operating state of the fuel cell or fuel cell stack was present (e.g., purge and / or drain valve closed).

[0080] Thus, valid values ​​for x, y, and p are available in the cloud for many measurements / samples. For example, the quantities can be linked in the following way (y=k(p,x)):

[0081] ■y( ') = (m m p -T m c ) x + c m p + c c ) = rn m px + m ( . x + c m p + c c

[0082] From this, the four quantities mm, mc, cm, and cc can now be estimated, for example, using a least-squares estimator in the cloud and numerous measurements / samples. The advantage of this approach / form is that it presents a linear problem, which is then easier to solve. This form is then transformed into the form x=f(y,p) relevant to the control unit, where the estimated quantities mm, mc, cm, and cc are then linked to the inputs p and y in the function f.

[0083] In this specific case, this function is nonlinear. The estimated values ​​mm, mc, cm, and cc are then transmitted back to the FCCU via mobile radio, for example, using the MQTT protocol, which allows x=f(y,p) to be evaluated again based on the current behavior of the valve.

[0084] Alternatively, the function f can be represented using various methods, for example, with a Gaussian process or a neural network, which takes the inputs y and p and delivers the output x. Furthermore, a characteristic map could be used in which the current x is plotted as a function of p and y. Interpolation between the sampling points is then performed in a known manner.

[0085] The implementation using four parameters is only an example. Depending on the chosen model, more or fewer parameters may be required.

[0086] In a further development, the cloud can determine in advance whether, for example, sufficient excitation is present in the signals, i.e., whether the entire operating range can be estimated from the available data. If this were not the case, for example, if the pressure p were constant, then a separate procedure would be required.

[0087] For example, if there is insufficient excitation in the signals and one or more of the four quantities mm, mc, cm, and cc cannot be estimated, the quantities that can be determined using the excitation are estimated first. The remaining quantities not yet estimated are then estimated, if the excitation in the signals allows.

[0088] As long as no estimates are available for the respective parameters, initial values ​​are assumed, which may, for example, be derived from the measurement of the HGI in its new state. Therefore, values ​​for the four variables mm, mc, cm, and cc are always available, so that the control of the HGI valve or the calculation of the control current (current = x) is always possible.

[0089] In summary, the method described above, or rather the corresponding architecture for implementing the method, aims to determine the parameters of the HGI valve from the available measured variables during operation of a fuel cell system, thus continuously adjusting the mathematical model to the actual behavior. This allows the controller to be automatically adapted to the valve used in a new system, and any drift in the valve behavior during operation is detected and compensated for. Processing the measured variables present in the control unit in the cloud allows the use of complex and computationally intensive algorithms to determine the mathematical model of the HGI valve. In addition to the HGI valve, the method described here can also be used to identify the model parameters of other components of the control unit.Examples can be: chokes in cathode, anode, Thermalsys pumps / compressors in cathode, anode, Thermalsys volumes in anode, cathode stack characteristics such as.

[0090] Polarization characteristic or water diffusion.

Claims

Claims 1 . Method (100) for improved control of a technical system (40) by means of a technical component (30), comprising the following steps: Acquiring (102) first data (32) of the technical component (30) by a data control unit (50), wherein the first data (32) are used to determine a characteristic operating behavior of the technical component (30); Transmitting (104) the first data (32) via a first interface (57) to a first cloud-based data processing unit (70); Generating (106) second data (33) by processing the first data (32) in the first cloud-based data processing unit (70); transmitting (108) the generated second data (33) via a second interface (75) to a second cloud-based data processing unit (80); Processing (110) the second data (33) by applying a data-based algorithm (84) stored in the second cloud-based data processing unit (80) to generate third data (34); Transmitting (112) the generated third data (34) to the data control unit (50) via a fourth interface (87); at least partially replacing (114) the preset first data (32) of the technical component (30) with the generated third data (34) in order to thereby improve the control of the technical system (40) with the technical component (30).

2. The method (100) according to claim 1, wherein the first data (32) are measurable quantities that are influenced by the technical component (30) and / or influence the technical component (30), so that the first data (32) describe the characteristic operating behavior of the technical component (30).

3. The method (100) according to claim 2, wherein the processing (110) of the second data (33) to third data (34) consists in providing adapted or optimized parameters for the mathematical model (31) of the technical component (30) in the form of the third data (34) from the second data (33).

4. Method (100) according to one of the preceding claims, wherein the processing of the first data (32) comprises at least one of the following measures: error and completeness checking, cleaning of interference signals, data format conversion.

5. The method (100) according to any one of the preceding claims, wherein the processing (110) of the second data (33) to form third data (33) is carried out using a virtual data processing unit (82) installed in the second cloud-based data processing unit (80), which receives the second data (33) from the first cloud-based data processing unit (80) and which accesses the data-based algorithm (84) to process the second data (33).

6. The method (100) according to claim 5, wherein the virtual data processing unit (82) is connectable to an output unit (86) via a third interface (85) in order to output the generated third data (33).

7. Method (100) according to one of the preceding claims, wherein the technical system (40) is designed as a fuel cell or as a fuel cell stack, which is controlled via a hydrogen injection valve (HGI) designed as a technical component (30).

8. A computer program containing machine-readable instructions which, when executed on one or more computers and / or compute instances, cause the computer or computers or compute instances to carry out the method according to any one of claims 1 to 7.

9. Machine-readable data carrier and / or download product with the computer program according to claim 8.

10. One or more computers and / or compute instances with the computer program according to claim 8, and / or with the machine-readable Data carrier and / or download product according to claim 9.

Images