Method for controlling the operation of an electrolytic cell plant - Patents.com

Abstract

The present invention provides a computer-implemented method for controlling the operation of an electrolyzer plant, comprising: determining a first operating point of the electrolyzer plant for a first predetermined time period using a first model; simulating the operation of the electrolyzer plant for the first operating point for a second predetermined time period shorter than and included in the first time period using a second model, where the second model has a higher prediction accuracy for the operation of the electrolyzer plant than the first model, and determining whether the simulated operation satisfies a predetermined requirement. The method comprises adjusting one or more parameters and / or one or more boundary conditions of the first model when it is determined that the simulated operation does not satisfy the predetermined requirement, and setting the first operating point as a target operating point for the second predetermined time period when it is determined that the simulated operation satisfies the predetermined requirement.

Description

[Technical field]

[0001] The present invention relates to a method, system, computer program product, and computer readable medium for controlling the operation of an electrolyzer plant. [Background technology]

[0002] Determining the optimal operating point of an electrolyser plant is usually a large optimization problem.

[0003] An electrolyser plant can be considered as a linking component in a complex energy system, linking different networks, namely electricity, heat, hydrogen and oxygen networks. Thus, there are different aspects that can be considered in the optimization, both on the electricity supply side of the electrolyser plant and on the hydrogen, oxygen and heat output side. For example, there may be limited available grid energy, or there may be an upper limit on the amount of hydrogen, heat or oxygen that can be pumped into the respective networks.

[0004] Furthermore, further aspects may be taken into account in the optimization, such as the available reserves for energy, hydrogen, oxygen and heat, and their respective efficiencies.

[0005] Further aspects that may be taken into account in the optimization are, inter alia, the properties of the electrolyser modules, the operating conditions or even the wear of the equipment.

[0006] Optionally, in addition to the technical aspects mentioned above, there may also be aspects relating to pricing or contractual obligations etc. that can be taken into account in the optimization.

[0007] Currently, electrolyser plants are often not operated at ideal operating points. Modelling is used to determine the operating points of the electrolyser plant and / or its components, but the results of the modelling are often inaccurate. Inaccurate modelling may result in non-optimal operation of the electrolyser modules, e.g. in terms of efficiency and / or wear, and may result in the actual power demand and / or hydrogen, oxygen and heat output not being as predicted, which may cause technical problems and in some cases also contractual and / or cost problems. For example, it may be impossible to output the required amounts of hydrogen, oxygen and / or heat at all, or only by operating in non-ideal operating conditions (e.g. in terms of wear and / or efficiency) or by resorting to purchasing electricity at a higher cost than expected.

[0008] Using known methods, high accuracy requires the use of more complex models and, at the same time, frequent recalculation of the operating point to take into account changing conditions, and therefore often has to be sacrificed in the interest of robustness and resolvability within the limits of available time and computational resources.

[0009] It is an object of the present invention to provide a method for controlling the operation of an electrolyser plant, which alleviates at least some of the above mentioned drawbacks. Summary of the Invention

[0010] This object is achieved by the present invention, which provides a method, a system, a computer program product and a computer readable medium as set forth in the independent claims. Preferred embodiments are defined in the dependent claims.

[0011] The present invention provides a computer-implemented method for controlling the operation of an electrolyzer plant, comprising: (a) determining a first operating point of the electrolyzer plant for a first predetermined time period using a first model; (b) simulating the operation of the electrolyzer plant for the first operating point for a second predetermined time period shorter than and included in the first time period using a second model, where the second model has a higher predictive accuracy for the operation of the electrolyzer plant than the first model; and (c) determining whether the simulated operation satisfies a predetermined requirement. The method comprises: (d) adjusting one or more parameters and / or one or more boundary conditions of the first model when determining that the simulated operation does not satisfy the predetermined requirement; and (e) setting the first operating point as a target operating point for the second predetermined time period when determining that the simulated operation satisfies the predetermined requirement.

[0012] In the present disclosure, an operating point is the value of power at which the electrolyzer plant is operated at a given time. The operating point is also referred to in the present disclosure as a set point or an operating set point.

[0013] The first operating point may comprise all operating points during the second time period. The operating points may be different at different times during the first time period. This may be described as a function of the operating point as a function of time.

[0014] As mentioned above, the second time period is included in the first time period. The first time period and the second time period may, for example, start at the same time and extend to different times. As an example, the first time period may be a number of days, in particular one week or more, in particular one month or more, and the second time period may be a number of hours, in particular one day. Thus, the first period may, for example, be divided into consecutive second periods of equal or different length.

[0015] The first and second models may each include, among other things, a representation of the interdependence between the input power and the output of hydrogen and / or oxygen and / or heat.

[0016] The first model can be considered a long-term planning model.

[0017] The first model may be, for example, a model that determines an operating point based on first parameters, which may include external parameters, for example, a projected availability of grid energy over a first time period, or a projected hydrogen demand, etc. The first parameters may also include parameters representing general electrolyzer plant characteristics, for example, the number of electrolyzer modules and / or a nominal overall hydrogen production efficiency of the plant.

[0018] The first model may not take into account, for example, parameters that characterize the electrolyzer plant on an individual module basis. Thus, the number of parameters may be kept small and the modeling may not be computationally excessively expensive. For example, the model may be a linear model. Thus, the modeling may be performed with reasonable computational resource consumption even for a longer first period. However, the accuracy of this model may be limited, for example due to not taking into account the efficiency characteristics of the individual modules.

[0019] Determining the first operating point using the first model may comprise optimizing the operating point based on the (these) first parameters and optionally one or more first boundary conditions.

[0020] The first model may predict future power consumption and / or hydrogen production curves. The first model may enable solving an optimization problem to determine an optimal operating point.

[0021] Optionally, the first model may be configured as a conservative model with respect to selected parameters. For example, the first model may be configured to underestimate the hydrogen production amount, e.g., to ensure that the minimum amount of hydrogen required is always produced even if the model is inaccurate. Alternatively or additionally, the first model may be configured to underestimate the available storage volume of hydrogen, e.g., to ensure that some storage volume remains available even if the model is inaccurate.

[0022] The second model may for example be a model that determines an operating point based on second parameters, which may include plant internal parameters, i.e. parameters characterizing the electrolyzer plant, in particular parameters characterizing the electrolyzer plant on an individual module basis. Optionally, it may not take into account plant external parameters, e.g. all of the parameters considered by the first model.

[0023] The second model may, for example, be a plant model that has already been constructed previously, for example when setting up the plant for operation, for example for optimization of the plant setup and / or for control and / or optimization of the plant operation.

[0024] The second model may be more accurate in predicting the effects of the operating point determined by the first model, e.g., it may use more parameters and / or boundary conditions and / or interdependencies and / or more appropriate equations to represent the operation of the plant.

[0025] The electrolyzer plant may comprise a number of electrolyzer modules, each of which may be characterized by a number of properties that may be represented by a second parameter. Furthermore, the second parameter may also include a parameter that represents the current state of one or more modules, such as temperature. The second model may therefore take into account a larger number of parameters than the first model. It may therefore provide more accurate results. However, the determination using the second model may also be computationally more expensive, for example due to its complexity, and therefore calculations for shorter time periods than the first model help to avoid an overall excessive computational resource consumption. Furthermore, the second parameters may include more dynamic parameters than the first parameters. A shorter time period also allows for a higher accuracy.

[0026] The second model can be considered as a short-term model for simulating the operating point of the first model.

[0027] Simulating the operation of the electrolyzer plant with the second model may comprise optimization. As an example, optimization may not be necessary if the second model is completely determined by the output from the first model. However, simulating may comprise optimization, especially if the second model is not completely determined by the output from the first model. In other words, simulating may comprise optimization, for example, if there are free parameters in the second model. In such cases, the second model has certain degrees of freedom. The present disclosure refers to simulation with optimization and simulation without optimization unless otherwise specified.

[0028] The second model may receive the first operating point as input and may simulate the operation of the electrolyzer plant at the first operating point for a second time period based on the (these) second parameters. The method comprises determining whether the simulated operation satisfies at least one predetermined requirement. The requirement may be that one or more conditions are satisfied. For example, the satisfaction of the condition may comprise that one or more boundary conditions are not violated, e.g. that the value of the parameter remains above and / or below a predetermined threshold, in particular within a predetermined interval. The boundary condition may be, for example, a safety-related requirement, e.g. an allowed temperature range. Alternatively or additionally, the satisfaction of the condition may comprise that a difference between the modeled operation obtained by the first model and the modeled operation obtained by the second model of the electrolyzer plant remains below a predetermined threshold.

[0029] The method comprises adjusting one or more parameters and / or one or more boundary conditions of the first model when it is determined that the simulated operation does not meet the predetermined requirements, and setting the first operating point as a target operating point for a second predetermined time period when it is determined that the simulated operation meets the predetermined requirements. The method may comprise, for example, predicting expected results for the adjusted parameters and / or the one or more adjusted boundary conditions to determine the adjustments.

[0030] In this disclosure, "upon determining" is used synonymously with "in response to determining."

[0031] In other words, only if the operating points determined by the first model turn out to be acceptable when applying the more accurate second model, these operating points will be used for the operation of the electrolyzer plant during the second time period. Otherwise, for example if it is determined based on the second model that the operating points obtained by the first model are not acceptable, i.e. in other words that a corrected operating point is required, at least for the predefined second time period, a recalculation of the operating point using the second model with different parameters and / or boundary conditions is triggered, on the basis of which the first operating point is corrected using the results from the second model which more comprehensively describes the plant operation.

[0032] Thus, results from a less resource intensive second model can be used if the second model yields accurate results, with more computational resources only required when the results are suboptimal.

[0033] Thus, the method of the present disclosure can improve the accuracy of determining the operating point without significantly increasing the overall computational resource consumption. Thus, the method of the present disclosure solves at least the problems identified above.

[0034] Determining using the first model may be computationally less expensive than simulating using the second model. In particular, the first model may be a linear model and the second model may be a non-linear model and / or may take into account a larger number of parameters than the first model. As an example, the first model may assume a constant efficiency for different operating points. This does not generally correspond to the reality that efficiency depends on the operating point.

[0035] A model that is less computationally expensive than another model may involve determining an operating point or simulating operation using that model requiring less computational resources, particularly CPU and / or data transfer, and / or memory usage, than determining an operating point or simulating operation using the other model.

[0036] For example, as described above, determining the operating point using the first model and / or the second model may comprise solving an optimization problem, which becomes computationally more expensive for non-linear models, a larger number of parameters, a longer time horizon being modeled, and / or a finer time discretization.

[0037] By using a computationally more expensive model only to refine a computationally less expensive model when necessary, improved performance can be ensured without unduly increasing computational resource usage.

[0038] In the present disclosure, a first model may determine an operating point based on a first data set, the first data set comprising one or more plant external parameters, and a second model may simulate the operation of the electrolyzer plant based on a second data set, the second data set comprising one or more plant specific parameters not included in the first data set.

[0039] This allows the two models to serve different primary objectives. For example, the first model may serve the primary objective of adjusting the operation to plant external parameters, often within the capacity of the electrolyser plant and not necessarily influenced by the plant operator. These parameters represent external limitations, including, for example, the availability of electricity and technical limitations on the possible amount of hydrogen that can be output by the plant, for example fed into an external network. Furthermore, optionally, they may also include other types of limitations, for example contractual limitations. In addition to technical limitations, the parameters may also optionally represent, for example, price forecasts for electricity and hydrogen. The second model may serve the primary objective of ensuring, for example, an optimal operation of the electrolyser plant with respect to the plant internal parameters. For example, module specific characteristics, for example efficiency curves and / or predicted module degradation, and / or overall plant characteristics, for example overall efficiency curves and / or predicted overall degradation, may be taken into account. Alternatively or additionally, operation specific parameters may be taken into account, in particular safety-related parameters.

[0040] Examples of aspects that may be taken into account by the second model are, for example, production rate, power consumption, temperature, pressure, ramping behavior, production distribution among several modules / stacks, degradation, mechanical stress, harmonics, reactive power, and / or hydrogen and oxygen mixing, on a module basis. These parameters may follow non-linear relationships.

[0041] High-precision optimization may require a high-precision understanding of the module efficiency and degradation. Models that accurately reflect this understanding are complex and optimization based on them may be computationally expensive. Thus, even if the first model considers parameters that allow the efficiency and expected degradation to be derived, in order to keep the model simple, it will not provide a thorough representation and therefore will lack precision. For example, information may be information obtained at installation using data from manufacturer data sheets or measurements. However, these may not be accurate for the actual electrolyzer module, dependencies may not be fully taken into account, effects from parameter interdependencies and from specific operating conditions may not be fully taken into account, and parameter changes over time may not be fully reflected.

[0042] Please also refer to the above description regarding first parameters, e.g., plant external parameters, that may be part of the first data set, and second parameters, e.g., plant internal parameters, that may be part of the second data set.

[0043] The first and second data sets may partially overlap, i.e. they need not be mutually exclusive, for example both data sets may comprise overall electrolyser plant nominal characteristics, such as the number of electrolyser modules and / or the operating efficiency.

[0044] The overlap may allow for faster convergence of the results from the first and second models.

[0045] In a method according to the present disclosure, determining whether the simulated operation satisfies the predetermined requirement may comprise: determining, for each of one or more boundary conditions, whether the simulated operation causes a violation of the boundary condition, and determining that the simulated operation satisfies the predetermined requirement if it is determined that no violation of any boundary condition is present.

[0046] In other words, the predefined requirements are determined to be met only when there is no violation of any boundary conditions, which is advantageous in that it can ensure that operation of the electrolyser plant remains within a safe and / or efficient operating range.

[0047] In particular, if it is determined that the simulated operation causes a violation of one or more boundary conditions, adjusting may comprise adjusting one or more corresponding boundary conditions of the first model for at least a second time period and optionally issuing an alert.

[0048] The boundary conditions may comprise the operating parameters of the electrolyser plant, in particular of the electrolyser modules, such as the minimum and / or maximum allowable set points of each electrolyser module and / or the allowable operating temperature of each electrolyser module. The boundary conditions may be selected to ensure safe operation and / or to reduce wear and / or to best exploit the efficiency curves characterizing each electrolyser module. The latter expresses the efficiency as a function of the electrolyser set points, which may for example be rather low at low set points.

[0049] The "corresponding" boundary condition of the first model according to the present disclosure may be, for example, the same boundary condition as the boundary condition in the second model, if the boundary condition of the second model is directly applicable to the first model. This may be the case when the parameter of the second model is also considered in the first model. The corresponding boundary condition of the first model may also be the boundary condition of the parameter of the first model that is related to the parameter in the second model to which the boundary condition of the second model is applied. For example, if the parameter of the second model is not considered in the first model, the boundary condition on the parameter of the second model may be converted into the corresponding boundary condition of a different parameter of the first model. This corresponds to an indirect adjustment of the boundary condition of the first model.

[0050] For example, the first model may not take into account the operating temperature of the electrolyser module, but boundary conditions in the second model related to the operating temperature may be transformed into corresponding boundary conditions, e.g. boundary conditions on the overall rate of hydrogen production.

[0051] The use of boundary conditions can ensure safe and reliable operation by imposing limits on the allowable parameter values ​​for operation. They can also serve to improve accuracy depending on the parameters to which the boundary conditions are applied.

[0052] In some cases, the disclosed method may not be limited to identical boundary conditions of the first and second models, but may instead apply corresponding boundary conditions, which allows for combining very different first and second models, in particular without restrictive requirements regarding overlap of parameters and parameter compatibility of the two models.

[0053] The disclosed method may further comprise determining a first operation of the electrolyzer plant, where the operation is characterized by a first operating point determined using the first model for a second predetermined time period, and calculating a difference between the first operation obtained using the first model and a simulated operation obtained using the second model. In particular, determining whether the simulated operation satisfies predetermined requirements may comprise determining whether the difference satisfies predetermined similarity requirements, for example whether it remains within predetermined limits and / or below a predetermined threshold.

[0054] The operation of the plant may be characterized, for example, by at least the power input and hydrogen output and / or the overall plant efficiency of hydrogen production. Additional parameters for characterizing the operation are contemplated.

[0055] If it is determined that the differences do not meet the predetermined similarity requirement, the adjusting may comprise adjusting one or more parameters of the first model.

[0056] As an example, the difference may be the difference in the overall operating point at a given time, or the difference between curves representing the overall operating point as a function of time. Further, adjusting the parameters as described above may enable the output of the first model to become more similar to the output of the second model.

[0057] This allows for an improvement in the overall accuracy. For example, even if all of the above boundary conditions are met, a further improvement in the accuracy of the first model can be achieved by reducing the difference, or in other words, by increasing the similarity.

[0058] The method of the present disclosure may comprise, in response to determining that the simulated operation causes a violation of a boundary condition, temporarily overriding the first operating point for a second time period with a temporary operating point obtained, for example, using the second model, such that operation for the temporary operating point does not cause a violation of the boundary conditions, in particular any boundary conditions. Determining that the simulation causes a violation may be part of determining whether a predetermined requirement is satisfied.

[0059] During operation, scenarios may arise where operating the plant at the first operating point is highly undesirable, for example in terms of safety, degradation and / or efficiency. This is due to the fact that the first model does not take into account all internal parameters of the electrolyzer plant, which may result in blind spots for certain scenarios. As an example, operation at the first operating point may cause the electrolyzer plant to operate in an unsafe manner if a violation of a boundary condition on a safety-critical parameter, for example the operating temperature of the electrolyzer module, is determined based on the second model.

[0060] In particular, the override may occur in response to determining that the expected time required to adjust the first model and recalculate the operating point using the first model exceeds a predetermined threshold and / or in response to determining that a safety-critical boundary condition is violated.

[0061] Thus, temporarily overriding the first operating point and operating the electrolyzer plant at the temporary operating point may not be ideal with respect to the optimization objectives underlying the first model, but may ensure safe, reliable and / or efficient operation of the electrolyzer plant.

[0062] Temporarily overriding may be performed at least until it is determined that the adjustments have resulted in simulated behavior that does not cause violations of boundary conditions, and in particular simulated behavior that meets predetermined requirements.

[0063] In particular, the temporary override may be terminated once it is determined that the adjustment has resulted in a simulated operation that does not cause a violation of a boundary condition that meets particular predetermined requirements, and, optionally, it is determined that no other violations of the boundary conditions will occur.

[0064] Optionally, a temporary operating point may be determined that overrides the first operating point while not violating boundary conditions of the first and / or second models, such that the temporary operating point remains as close as possible to the first operating point.

[0065] The method of the present disclosure may comprise repeating steps (a)-(d) at least until determining that the simulated operation meets predetermined requirements.

[0066] That is, the method may iteratively adjust one or more parameters and / or one or more boundary conditions of the first model and, after the adjustment, recalculate the operating point using the first model. This may be done at least until a predetermined requirement is met. Once the predetermined requirement is met, the iterations may end. This is advantageous in terms of computational resources. Alternatively, the steps may be repeated after the simulated operation meets the predetermined requirement, for example for monitoring and / or correction purposes. For example, the steps may be repeated at predetermined times or time intervals within the second time period. This may improve accuracy and reliability.

[0067] The method of the present disclosure may comprise carrying out steps (a)-(d), in particular carrying out any of the method steps outlined above, during one or more additional time periods as a second time period, in particular at predetermined times and / or at predetermined time intervals during a subsequent time period that is shorter than the first time period and included within the first time period.

[0068] As an example, the first time period may be multiple days and each second time period may be one day or less. Steps (a)-(d) may then be performed at least once each day, for example at a predetermined time and / or at predetermined intervals, for example 24 hour intervals.

[0069] This makes it possible to ensure that after an initial check and possibly adjustment, i.e. after steps (a) to (d) are carried out once, the changing accuracy of the first model over time is detected and corrected in time.

[0070] The methods of the present disclosure may comprise storing the training data in a computer-readable form and / or improving the first model based on the training data, in particular using machine learning methods, in particular an improvement that takes into account contextual information.

[0071] The adjustment data may comprise, for example, one or more adjusted parameters, and / or one or more adjusted boundary conditions, and / or an adjusted operating point, i.e., an operating point determined and obtained based on the first model using the adjusted parameters and / or one or more adjusted boundary conditions, and / or an adjustment step performed to obtain the adjusted parameters and / or adjusted boundary conditions, and optionally associated context data, for example reflecting operating conditions and / or environmental factors associated with the second time period.

[0072] Improving the first model may comprise modifying parameters and / or boundaries of the first model after the second time period, and in particular even during a time period after the first time period, to obtain more accurate results.

[0073] The associated context data may be context data associated with one or more adjusted parameters, and / or one or more adjusted boundary conditions, and / or an adjusted operating point, and / or an adjustment step performed to obtain the adjusted operating point.

[0074] The associated contextual data may comprise data enabling a classification of time periods, for example, day classes, season classes, weather classifications, classes related to shifts or production schedules. Specific adjustments associated with particular classes may be used to determine modifications to the first model that enable the model to make improved predictions by taking into account the classification of the time period in question.

[0075] The labeling may be used to store the context data and / or to define an association between the context data and the remaining adjustment data.

[0076] By providing and / or using the adjustment data outlined above, the first model may be tuned based on adjustments that better reflect the actual operation of the electrolyzer plant, particularly under the particular circumstances reflected by the context data.

[0077] Thus, the accuracy of the first model can be improved, thereby reducing the need for future adjustments using the second model.

[0078] In the method of the present disclosure, simulating using the second model may comprise performing independent simulations in parallel at electrolyser module level for a plurality of electrolyser modules of the electrolyser plant, and then aggregating results of the independent simulations taking into account interdependencies between the electrolyser modules, and optionally between the electrolyser modules and other equipment of the electrolyser plant.

[0079] In other words, at least a partial parallelization of the calculations involved in the simulation may take place. This improves the overall performance and also means that the delays in adjustments and subsequent recalculations are reduced. As a result, the overall operation may become more efficient and potentially safer. Furthermore, if the above-mentioned overrides are performed, the time during which the overrides are performed may be reduced.

[0080] The present invention also provides a system comprising a processing system, also referred to in the present disclosure as a computing system, configured to perform any of the methods of the present disclosure.

[0081] The system may further comprise a plurality of electrolyser modules of the electrolyser plant, the processing system being configured to control operation of the electrolyser modules to operate at a target operating point during the second period of time.

[0082] The system may be or may comprise an electrolyser plant.

[0083] The system may be configured to provide operational and / or specification data of the electrolyser module as input to the first model and / or the second model.

[0084] The operational data may comprise current and / or historical operating parameters. The specification data may comprise data describing characteristics of the electrolyzer module, for example characteristics provided by a manufacturer and / or characteristics derived therefrom and / or from the operational data.

[0085] The electrolyzer module may be configured to directly provide operational and / or specification data as input for the first model, and / or the operational and / or specification data may be retrieved from data storage, and / or the operational data may be obtained via monitoring equipment.

[0086] The present invention also provides a computer program product comprising instructions which, when executed by a computer, cause the computer to perform any of the methods of the present disclosure.

[0087] The present invention also provides a computer readable medium comprising instructions which, when executed by a computer, cause the computer to perform any of the disclosed methods.

[0088] The features and advantages outlined above in the context of the method apply equally to the systems, computer program products, and computer readable media described herein.

[0089] Further features, embodiments, and advantages will become apparent from the detailed description which refers to the accompanying drawings, in which: [Brief description of the drawings]

[0090] [Figure 1a] 1 illustrates a schematic diagram of an electrolyser plant. [Figure 1b] 1 illustrates a schematic diagram of an electrolyser plant. [Diagram 2] 1 is a flow chart illustrating a method according to the present disclosure. [Diagram 3] 1 is a flow chart illustrating a method according to the present disclosure. [Figure 4] 1 illustrates an exemplary efficiency curve for a module or plant. [Diagram 5] 1 illustrates an exemplary operating profile for a plant. [Figure 6] 1 illustrates an exemplary hydrogen production curve. [Figure 7] 1 illustrates an exemplary power demand curve. [Figure 8] 1 illustrates an exemplary power forecast for one day. [Figure 9] 1 illustrates exemplary efficiency curves for two stacks of a plant. [Figure 10] 1 illustrates an example power schedule for two stacks. [Figure 11] 13 illustrates another example power schedule for two stacks. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0091] Figures 1a and 1b illustrate a system 1 according to the present disclosure, which is an electrolyser plant. The electrolyser plant comprises a number of electrolyser modules 2, a computing system 3, an optional monitoring device 4, an optional electrical storage 5, an optional hydrogen storage 6, and an optional oxygen storage 7. Furthermore, arrow 8 indicates the electrical input to the electrolyser plant, arrow 9 indicates the hydrogen output from the electrolyser plant, arrow 10 indicates the oxygen output from the electrolyser plant, arrow 11 indicates the heat output from the plant, and arrow 12 indicates the water input to the electrolyser plant.

[0092] For illustrative purposes only, and although not part of the system of this embodiment, a power grid 13 and a network 14 into which hydrogen, oxygen and heat are fed are shown.

[0093] It should be noted that FIG. 1a illustrates a plant with less detail of the individual components of the plant than FIG. 1b, solely to reflect the level of detail or aspects with which the first and second models of the present disclosure may each model an electrolyzer plant.

[0094] The method of the present disclosure may be performed in the system shown in Figures 1a and 1b, the steps of the method being performed, for example, by a computing system 3 or any other suitable system, in particular the system according to the present disclosure.

[0095] FIG. 2 is a flow chart illustrating a method according to the present disclosure.

[0096] The method comprises a step S11 of determining a first operating point of the electrolyzer plant for a predetermined first time period using a first model, a step S12 of simulating operation of the electrolyzer plant for the first operating point for a predetermined second time period that is shorter than and included in the first time period using a second model, where the second model has a higher prediction accuracy for the operation of the electrolyzer plant than the first model, and a step S13 of determining whether the simulated operation satisfies predetermined requirements.

[0097] In response to determining that the simulated operation does not meet the predetermined requirements, in step S14, one or more parameters and / or one or more boundary conditions of the first model are adjusted.

[0098] The method may then return to step S11 and repeat the method steps with the adjusted parameters and / or boundary conditions, for example, until it is determined that the simulated operation meets predetermined requirements, in which case the method proceeds to step S15.

[0099] In response to determining that the simulated operation meets the predetermined requirements, in step S15, the first operating point is set as a target operating point for a predetermined second time period.

[0100] By setting the first operating point as the target operating point, the plant is controlled to operate at the target operating point.

[0101] FIG. 3 is a flow chart illustrating a method according to the present disclosure that may illustrate in more detail how the first and second models are used.

[0102] In step S21, the first model is used to determine operational set points for a first time period. In step S22, the set points are used to predict the power input and hydrogen output, and optionally the heat output, for a second, shorter time period using the first model. For example, the first model may be a linear model that takes into account external technical aspects such as electricity supply and / or hydrogen demand, and optionally external aspects related to pricing and / or contractual aspects.

[0103] The set points are also used in step S23 to simulate the plant operation during a second time period using a second model, also outputting at least the electrical power input and hydrogen output, and optionally a thermal output. Additionally, the simulation may yield results for other parameters, e.g., operationally related.

[0104] More specifically, the set points of step S22 are received as inputs for the simulation in step S23. If the second model is completely determined by the results from the first model, e.g., by power consumption, hydrogen production, and heat production, the simulation does not necessarily have to comprise optimization. Alternatively, the simulation may comprise optimization. This may be the case when the second model has free parameters, e.g., reactive power, temperature profile, or degradation.

[0105] In step S24, the difference between the results obtained by the first model and the results obtained by the second model is determined, e.g., with respect to the predicted or simulated power input and hydrogen output, and optionally the heat output. The comparison may be made using different distance metrics, such as, for example, individual values ​​at a given time, average values ​​over a predefined time interval, or average difference values.

[0106] In a subsequent step S25, it is determined whether the difference exceeds a predefined threshold. In response to determining that it does, in step S26, the parameters of the first model for the second time period are adjusted. The parameters may be adjusted such that a recalculation with the adjusted parameters is expected to reduce the difference between the results obtained by the first model and the results obtained by the second model in terms of the predicted or simulated power input and hydrogen output, and optionally the thermal output, for the second time period. In other words, improved parameters for the first model are calculated that result in a more accurate prediction by the first model and thus a reduction in the difference between the results of the first model and the results of the second model. Step S26 is performed only if the threshold is exceeded. That is, if the threshold is not exceeded, no adjustment is performed.

[0107] In step S27, it is determined whether the simulation in step S23 resulted in parameters violating the predetermined boundaries. In response to determining that this is the case, in step S28, the corresponding boundaries of the first model are adjusted. The corresponding boundaries are adjusted such that a recalculation with the adjusted corresponding boundaries is expected to result in a set point that, when simulated with the second model, does not result in a violation of the predetermined boundaries, in other words, ensures that the boundaries are not violated.

[0108] Following steps S26 and / or S28, in step S29, the parameters and / or boundary conditions of the first model are updated and a recalculation is triggered, and the method returns to step S21, i.e., the first model with the adjusted parameters and / or boundaries is used to determine the operational set points.

[0109] The steps may be repeated until it is found in step S25 that the difference between the models does not exceed a predefined threshold, i.e., no parameter adjustments are necessary and no violation of any bounds is detected, sometimes called convergence. The checks for whether both are true may be performed sequentially, for example in the order shown in the figure, or in reverse order, or in parallel. Alternatively or additionally, one or more abort criteria may be applied that interrupt the loop, for example if convergence is not reached after a predefined maximum number of iterations, or if the results appear to be fluctuating.

[0110] The method may then proceed to step S30, i.e., setting the operational set point as the target operational set point for the second time period.

[0111] Further examples of the method according to the present disclosure are outlined below with reference to Figures 4 to 11. It should be noted that the first model is called M1 and the second model is called M2. The electrolyser plant is also called a hydrogen production plant.

[0112] [Example 1] Example 1 focuses on the power set point and efficiency at the power set point, which in this example having a plant with only one module for illustrative purposes only is the overall power at which the plant is operated.

[0113] 4, which illustrates an exemplary simplified efficiency curve of a hydrogen production plant. This curve can be approximated in a first model M1 by a constant value, for example 74% in this example.

[0114] In this example, at a nominal operating point of 5 MW, 73% of the power can be converted into hydrogen, while the rest is a loss, e.g., a heat loss. Towards lower power set points, the efficiency rises to a peak of about 78% and then drops sharply. The assumption of M1, i.e., hydrogen production at a constant efficiency of 74%, can already take into account the fact that the plant is not always operated at the nominal operating point value (5 MW with an efficiency of 73%), but sometimes at a value below the nominal value with a higher efficiency. Furthermore, for this example, it is assumed that M1 determines a daily power set point as shown in FIG. 5, which illustrates an exemplary operating profile of a hydrogen production plant for one day. One day corresponds to the second time period. The first time period (not shown) can be, for example, one week.

[0115] The amount of hydrogen production determined by M2 is the power set point shown in FIG. 5 multiplied by the efficiency shown in FIG.

[0116] 6 illustrates that assuming an efficiency of M1 (74%) results in an unexpectedly large excess production ("Delta 74%"). Optimizing M1 efficiency parameters using M2 using the methods of the present disclosure results in a production delta that is close to zero ("Delta 75.1% (Optimized)").

[0117] Figure 6 shows that there appears to be good agreement between the actual production volume ("Actual Production") and the simulated production volume at 74% ("Simulated at 74%). However, integrating the deviation over a day ("Delta 74%) results in an additional 900 Nm2 of production, corresponding to an additional hour of production, compared to the M1 prediction. 3 There is a deviation of hydrogen.

[0118] Overproduction is problematic on multiple levels, as it can result in entire production hours being lost if the hydrogen cannot be pumped into the network due to a lack of storage capacity. Additionally, the price at which hydrogen can be sold overhead can be low.

[0119] As can be seen, with a fixed M1 set point of 76% ("Simulated at 76%), the M1 model produces 700Nm less torque than the actual production output. 3 Forecasting high hydrogen production volumes ("Delta 76%), which could result in underproduction of hydrogen, which could result in contracts not being fulfilled and potentially having to pay penalties.

[0120] By employing a more accurate second model M2 according to the present disclosure, this discrepancy can be found for this particular day and an adjusted value, e.g., optimal value, for the constant M1 set point of 75.1% can be calculated. Re-running M1 based on the updated values ​​for that day may still predict slightly incorrect values ​​for each hour, but the total production over the day will be accurate.

[0121] The optimum value (e.g., 75.1%) depends on the M1 profile on that day (e.g., as shown in FIG. 5) and therefore may not be accurate for other days. This means that tuning with M2 may need to be repeated for other days.

[0122] However, subsets of profiles for different days can be grouped into "typical profiles" and day classes may be defined, and M2 can be used to determine the optimal value for each of the day classes. For example, a value for weekdays and another value for weekends. Similarly, a value for summer and a value for winter, or a value depending on the weather forecast (calm or windy days) are also conceivable.

[0123] Just for the sake of completeness, as can be seen above, an example for keeping the model of a hydrogen production plant linear is a constant hydrogen (and heat) production efficiency of the production plant, which is not the case in reality. Thus, a constant model is clearly out of touch with reality and gives rise to the following problems or drawbacks:

[0124] As can be seen above, the predicted production rate is misestimated, causing more or less hydrogen to be produced than expected depending on the set point. The estimated power will also be different. This can be compensated for via electricity purchased from / sold to the grid or via electricity storage, but the SOC (State of Charge) will be different from the estimated SOC of M1. This may also involve higher costs. The actual operation may enter an operating area that is not allowed due to technical reasons (safety, grid limits) or contractual reasons. For example, the power grid limits of the plant may be exceeded, the set point may cause an extreme degradation situation of the plant, and / or the set point will cause the plant operation in a non-safe area, for example due to high mixing of hydrogen into the oxygen exhaust causing explosive gases. This may be due to the model M1 not including the respective physical parameter(s) such as electrolyzer temperature and pressure, grid side reactive power or harmonics, or plant degradation. Potential further problems may include not being able to produce enough hydrogen, causing a contract violation, or violating grid code compliance.

[0125] [Example 2] Example 2 focuses on the manufacturing set point.

[0126] Figure 7 shows the relationship between power demand [kW] and desired hydrogen production [Nm m 3 ] curve is shown as an example.

[0127] The M2 model is calculated with realistic non-linear values ​​("M2 power demand"). The simpler M1 model is calculated with a linearized representation ("M1 power demand"). Based on this, the power demand for a given desired hydrogen production curve looks different, as shown in Figure 8.

[0128] More specifically, Figure 8 visualizes the daily production of the hydrogen demand profile ("H2 Demand"). The power difference ("Delta P") between the M1 schedule ("M1 Power") and the M2 schedule ("M2 Power") is integrated to 900 kWh ("Delta E") over the course of a day. The day corresponds to a second time period. The first time period (not shown) can be, for example, a week.

[0129] As in the first example, the integrated energy demand over the day shows a deficit value of 850 kWh. However, in this example, the discussion focuses on the maximum power. M1 predicts 4877 kW, while M2 yields a more accurate higher value of 5000 kW.

[0130] In practical operation there is a risk that the difference will cause the grid or transformer limits to be exceeded, either alone or together with other consumers. According to the method of the present disclosure, limit violations are detected and corrected, for example by steps S27, S28, S29 and (repeated) step S21 described above.

[0131] For example, 978 Nm in a given time frame 3 A value for maximum hydrogen production / h is calculated and the M1 model is updated with this adjusted value, triggering a recalculation for that particular day, resulting in, for example, not exceeding grid limits.

[0132] Potentially technically dangerous situations such as overloading equipment can be avoided, and, optionally, contractually prohibited situations such as receiving more power than permitted can be avoided.

[0133] [Example 3] Example 3 considers the method of the present disclosure in the context of stack degradation. The degradation of electrolyzer stacks depends on many factors. The operating time is typically provided on the product sheet as a value of efficiency loss over a certain operating time, e.g., 0.1% efficiency loss per 1000 hours of operation. Other factors may include cycle count (comparable to a battery), operating temperature, ramp geometry, and water quality.

[0134] In this example, for illustrative purposes only, a plant with two stacks is considered, stack 1 is a new or refurbished stack (which may have an efficiency as illustrated in Figure 4) and stack 2, which has not yet been replaced, is approaching the end of its life and therefore has a reduced efficiency. Figure 9 illustrates the different efficiency curves of the two stacks.

[0135] The end of life determination may be based on several operating factors (among them operating time and operating temperature) and may ensure that the stack is removed from operation before a safety risk arises due to hydrogen permeation on the oxygen side.

[0136] M1 may result in a production schedule and distribution between two stacks as shown in Figure 10, which illustrates a daily schedule for an exemplary plant consisting of two stacks, one ("P Setting Stack 2") near the end of its life. The day corresponds to the second time period. The first time period (not shown) may be, for example, one week.

[0137] In this example, M1 contains an impairment model based on the total current through the stack, which gives rise to a current limit for stack 2, resulting in P-max=1000kW.

[0138] However, M2 in this example calculates that due to effects not modelled in M1, stack 2 will reach the end of its life at around 15:00. Switching stack 2 off at that point will result in a loss of 1000Nm by the end of the day. 3This causes a shortage of hydrogen.

[0139] To still ensure sufficient hydrogen production while also ensuring plant safety, the method of the present disclosure causes the M2 model to adjust parameters and / or boundary conditions to specifically reduce the M1 current limit of stack 2 for this day.

[0140] The updated M1 calculation after the adjustment results in a revised distribution of the desired production between the two stacks, for example as shown in Fig. 11. More specifically, Fig. 11 shows a daily schedule of an exemplary plant consisting of two stacks. Stack 2, after the adjustment and recalculation, is further reduced as it can be compensated by stack 1 to increase its life.

[0141] Note that the overall hydrogen production schedule is the same as in Figure 11, but the distribution between the two stacks is using stack 2 less, so it can still operate until the end of the day before reaching the end of its life. This effect is even more pronounced if no degradation is modeled at all in M1.

[0142] The above examples 1-3 provide several advantages in addition to those already outlined above, such as: The error of the first model M1 can be limited to a predefined value (particularly embodiment 1). · Due to the increased accuracy of the M1 results after tuning, better decisions can be made, potentially resulting in higher profits (especially in Example 1). Violations of safety and / or contractual obligations can be avoided, again resulting in potential cost savings (especially in examples 2 and 3). All this is achieved without increasing the complexity of the M1 model, which ensures that the robustness and response time of M1 remain constant (especially in Examples 1-3). Effects on parameters that change over time (seasonal effects or degradation effects) can be taken into account (especially in Examples 1 to 3). The plant is prevented from operating at potentially unsafe operating points (especially embodiment 2). Plant life can be increased without sacrificing financial performance (especially example 3). Additional measurements are avoided (especially examples 1 to 3). M1 can be tuned by conservative inputs by M2 to stay on the safe side. An example would be to reach as close as possible to the hydrogen production target, but any deviation would be associated with a higher production, thus avoiding, for example, penalties. The method can tune M1 over time, while still allowing M1 to remain robust to changes in external conditions. The method offers the possibility of parallelization. It can be performed at specific time intervals (e.g. once a day). Any recalculation of M1 that may be necessary therefore only needs to be done for the time interval where a problem has been identified. Technically, the recalculation can be parallelized with other calculations, avoiding recalculating the complete time horizon of M1.

[0143] While the present invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered as illustrative and not restrictive. The invention is not limited to the disclosed embodiments. In view of the foregoing description and drawings, it will be apparent to those skilled in the art that various modifications may be made within the scope of the present invention as defined by the claims.

Claims

1. 1. A computer-implemented method for controlling the operation of an electrolyser plant (1), comprising the steps of: (a) determining a first operating point of the electrolyzer plant (1) for a first predetermined time period using a first model; (b) simulating operation of the electrolyzer plant (1) for the first operating point during a second predetermined time period that is shorter than and included in the first time period using a second model, where the second model has a higher predictive accuracy for the operation of the electrolyzer plant (1) than the first model; (c) determining whether the simulated operation meets predetermined requirements; and (d) adjusting one or more parameters and / or one or more boundary conditions of the first model upon determining that the simulated operation does not satisfy the predetermined requirements; and (e) setting the first operating point as a target operating point for the second predetermined time period upon determining that the simulated operation satisfies the predetermined requirements; and A method comprising:

2. 2. The method of claim 1, wherein determining using the first model is computationally less expensive than simulating using the second model, in particular the first model being a linear model and the second model being a non-linear model and / or may take into account a larger number of parameters than the first model.

3. the first model determines an operating point based on a first data set, the first data set comprising one or more plant external parameters; the second model simulates operation of the electrolyzer plant based on a second data set, the second data set comprising one or more plant specific parameters not included in the first data set. The method according to claim 1 or 2.

4. determining whether the simulated motion satisfies the predetermined requirement comprises: determining, for each of one or more boundary conditions, whether the simulated motion causes a violation of the boundary condition; and determining that the simulated motion satisfies the predetermined requirement if it is determined that there is no violation of any of the boundary conditions; In particular, if it is determined that the simulated operation causes a violation of one or more of the boundary conditions, said adjusting comprises adjusting one or more corresponding boundary conditions of the first model for at least the second time period and optionally issuing an alert. The method according to any one of claims 1 to 3.

5. determining a first operation of the electrolyzer plant, the operation being characterized by the first operating point determined using the first model during the second predetermined time period, and calculating a difference between the first operation obtained using the first model and a simulated operation obtained using the second model; Further comprising: A method according to any one of claims 1 to 4, in particular wherein determining whether the simulated behaviour satisfies the predetermined requirements comprises determining whether the differences satisfy predetermined similarity requirements, e.g. whether they remain within predetermined limits and / or below a predetermined threshold.

6. The method of claim 5 , wherein if it is determined that the difference does not meet the predetermined similarity requirement, the adjusting comprises adjusting one or more parameters of the first model.

7. In particular, as part of determining whether the predetermined requirement is satisfied, in response to determining that the simulated operation causes a violation of the boundary condition, temporarily overriding the first operating point during the second time period with a temporary operating point, e.g. obtained using the second model, so that operation on the temporary operating point does not cause a violation of the boundary conditions, in particular any boundary conditions; In particular, the temporary overriding is performed at least until it is determined that the adjusting has resulted in a simulated behavior that does not cause a violation of the boundary conditions, in particular a simulated behavior that meets the predetermined requirements. The method according to any one of claims 1 to 6.

8. A method according to any preceding claim, comprising repeating steps (a) to (d) at least until determining that the simulated behaviour satisfies the predetermined requirements.

9. The method according to any one of claims 1 to 8, comprising carrying out steps (a) to (d) during one or more additional time periods as the second time period, in particular at predetermined times and / or at predetermined time intervals during a subsequent time period which is shorter than the first time period and is included in the first time period, in particular carrying out any of the method steps according to any one of claims 1 to 8.

10. storing training data in a computer readable form and / or improving the first model based on the training data, in particular using machine learning methods, in particular an improvement taking into account contextual information, In particular, the adjustment data comprises the one or more adjusted parameters, and / or the one or more adjusted boundary conditions, and / or an adjusted operating point, and / or an adjustment step performed to obtain the adjusted operating point, and optionally associated context data reflecting operating conditions and / or environmental factors associated with the second time period. The method according to any one of claims 1 to 9.

11. 11. The method according to any one of claims 1 to 10, wherein simulating with the second model comprises performing independent simulations at electrolyser module level in parallel for a number of electrolyser modules (2) of the electrolyser plant (1) and subsequently aggregating results of the independent simulations taking into account interdependencies between the electrolyser modules (2) and optionally between the electrolyser modules (2) and other equipment of the electrolyser plant (1).

12. A system (1) comprising a processing system (3) configured to carry out the method according to any one of the claims 1 to 11.

13. 13. The system of claim 12, further comprising a plurality of electrolyzer modules (2) of an electrolyzer plant, the processing system being configured to control operation of the electrolyzer modules to operate at the target operating point during the determined second period of time.

14. A computer readable medium comprising instructions which, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 11.

15. A computer program product comprising instructions which, when executed by a computer, cause said computer to carry out the method according to any one of claims 1 to 11.

Images