A method and system for testing adaptability of electrolytic cell fluctuation based on working condition decomposition, and a safe operation control method

Abstract

The application relates to an electrolytic cell fluctuation adaptability test method and system and a safe operation control method of an electrolytic cell, wherein the electrolytic cell fluctuation adaptability test method comprises the following steps: obtaining a sample dynamic power sequence and performing cyclic characteristic decomposition to construct a load spectrum; extracting typical operation conditions by clustering the load spectrum and establishing corresponding single-condition degradation models; and then combining the typical condition distribution proportion to calculate the cumulative damage rate of the to-be-tested dynamic power sequence and determining the fluctuation adaptability index of the electrolytic cell. The application overcomes the defect that the damage of complex dynamic conditions cannot be quantified, can scientifically evaluate the damage contribution of different fluctuation modes to the electrolytic cell, and provides key technical support for the safe operation and boundary construction of a hydrogen production system in a new power system.

Description

Technical Field

[0001] This invention belongs to the field of electrolytic hydrogen control technology, specifically relating to an electrolyzer fluctuation adaptability test method, system, and safe operation control method based on operating condition decomposition. Background Technology

[0002] With the large-scale grid connection of high-proportion fluctuating power sources such as wind and solar power, electrolyzers, as the core electrochemical unit of electro-hydrogen production systems, will inevitably operate under frequent, random, and variable power disturbances. Compared with traditional steady-state operating conditions, such fluctuating conditions cause rapid changes in internal operating parameters such as current, stack temperature, gas yield, and pressure, subjecting the electrolyzer to repeated loading of electrochemical and mechanical coupled stresses, significantly accelerating the damage accumulation process of its materials and structure.

[0003] In existing technologies, research on the operating status of electrolyzers mainly focuses on energy efficiency testing, steady-state performance modeling, and lifetime degradation trend analysis, emphasizing long-term aging patterns or static performance changes. However, existing research is significantly insufficient in quantifying the short-term damage contribution under unstable operating conditions such as large fluctuations in wind and solar power, frequent start-ups and shutdowns, and dynamic ramp-ups, as well as in testing fluctuation adaptability and understanding the impact of different dynamic operating conditions on degradation mechanisms. This makes it difficult to accurately test the damage effects of different fluctuation modes on electrolyzers in engineering practice.

[0004] This is because power signals driven by actual renewable energy sources exhibit strong randomness and multi-scale characteristics. Existing technologies lack the ability to structurally describe complex fluctuations when dealing with such non-stationary fluctuation conditions, making it impossible to distinguish the differentiated impacts of different fluctuation modes on the electrochemical and mechanical coupling stresses within the electrolyzer. Furthermore, since different types of fluctuation conditions alternate in actual operation, their damage accumulation exhibits significant nonlinear characteristics. Existing single stress degradation models cannot reproduce the complex fluctuation mechanisms in real operation, making it difficult to establish a quantitative mapping relationship between fluctuation conditions and degradation behavior. This makes it impossible to independently quantify the damage contribution of different dynamic conditions, thus preventing the acquisition of the overall cumulative damage level based on a physically consistent damage superposition mechanism when facing combined conditions.

[0005] Therefore, there is an urgent need for a solution that can accurately quantify the cumulative damage under various dynamic fluctuation conditions and objectively evaluate the fluctuation adaptability of electrolyzers in complex power systems. Summary of the Invention

[0006] One of the objectives of this invention is to at least solve one or more of the aforementioned problems in the prior art. In other words, one of the objectives of this invention is to provide an electrolytic cell fluctuation adaptability testing method, system, and safe operation control method based on operating condition decomposition that meets one or more of the aforementioned requirements.

[0007] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for testing the adaptability of an electrolyzer to fluctuations based on operating condition decomposition, comprising: Obtain several sample dynamic power sequences of the electrolyzer during actual operation; Cyclic feature decomposition is performed on the dynamic power sequence of the sample to obtain the load spectrum containing several load cyclic units; Several typical operating conditions are generated by clustering the load cyclic units in the load spectrum; Experiments or simulations were conducted for each typical operating condition to generate a single-condition degradation model corresponding to each typical operating condition. Obtain the dynamic power sequence of the electrolyzer during actual operation; Calculate the load spectrum corresponding to the dynamic power sequence under test, and classify the load cycle units into a typical operating condition; Based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test, the cumulative damage rate corresponding to the dynamic power sequence under test is calculated. The fluctuation adaptability index of the electrolyzer is determined based on the cumulative damage rate, and the corresponding operational adaptability evaluation results are output.

[0008] In a preferred embodiment, the elements in the load spectrum include mean power, cyclic amplitude, and number of cycles.

[0009] As a further preferred implementation, several typical operating conditions are generated by clustering the load cyclic units in the load spectrum, including: Using the mean power and cyclic amplitude of the load cyclic unit as the sample feature vector and the number of times it appears in the cycle as the weight, the K-means clustering algorithm is used to minimize the intra-cluster variance to obtain multiple cluster centers; Each cluster center corresponds to a set of operating conditions, which is defined as a typical operating condition.

[0010] As a preferred implementation, the dynamic power sequence to be measured is subjected to cyclic feature decomposition to obtain a load spectrum containing several load cyclic units. Specifically, the peak-valley pairs in the dynamic power sequence to be measured are extracted using the rainflow counting method to filter out non-critical fluctuations, and closed-loop cyclic extraction is performed to obtain several load cyclic units and construct a load spectrum.

[0011] As a preferred implementation method, when conducting experiments or simulations for each typical operating condition to generate a single-condition degradation model corresponding to each typical operating condition, an exponential function is used to construct the single-condition degradation model for the early rapid decay stage. For the medium- and long-term nonlinear degradation stage, a single-condition degradation model is constructed using a polynomial function.

[0012] As a preferred implementation, based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test, the cumulative damage rate corresponding to the dynamic power sequence under test is calculated, including: Using the single-condition degradation model corresponding to each typical operating condition, the single-condition damage curve corresponding to each typical operating condition is calculated respectively. The cumulative damage model is obtained by weighting and superimposing the single-condition damage curves corresponding to each typical operating condition with the corresponding distribution ratio. The cumulative damage rate is calculated using the cumulative damage model.

[0013] As a preferred embodiment, the method further includes: Obtain the actual operational damage values ​​of the electrolytic cell; The error between the actual damage value and the predicted damage value calculated based on the cumulative damage rate is calculated. If the error exceeds a preset threshold, the model parameters of the single-condition degradation model are updated in reverse using an optimization algorithm to perform closed-loop correction.

[0014] On the other hand, the present invention also provides a method for safe operation control of an electrolytic cell, comprising: The fluctuation adaptability test method of the electrolyzer based on operating condition decomposition, as described above, is used to obtain the fluctuation adaptability index of the electrolyzer. Construct the safe operating boundary of the electrolyzer based on its fluctuation adaptability index; Power regulation strategies are developed based on safe operating boundaries to control the safe operation of the electrolyzer.

[0015] On the other hand, the present invention also provides an electrolytic cell fluctuation adaptability testing system based on operating condition decomposition, comprising: The data acquisition module is used to acquire several sample dynamic power sequences and the dynamic power sequence to be measured during the actual operation of the electrolytic cell. The decomposition module is used to perform cyclic feature decomposition on the sample dynamic power sequence and the dynamic power sequence to be tested to obtain a load spectrum containing several load cyclic units. The clustering module is used to cluster the load cycle units in the load spectrum to generate several typical operating conditions; The classification module is used to classify the load cyclic units of the dynamic power sequence under test into a typical operating condition. The simulation modeling module is used to conduct experiments or simulations for each typical operating condition and generate a single-condition degradation model corresponding to each typical operating condition. The calculation and evaluation module is used to calculate the cumulative damage rate corresponding to the dynamic power sequence under test based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test. The output module is used to output the cumulative damage rate as an indicator of the electrolyzer's fluctuating adaptability.

[0016] As a preferred embodiment, the system further includes: The closed-loop correction module is used to obtain the actual operating damage value of the electrolytic cell, calculate the error between the actual operating damage value and the predicted damage value calculated based on the cumulative damage rate, and if the error exceeds the preset threshold, the model parameters in the single-condition degradation model are updated in reverse using the optimization algorithm to perform closed-loop correction.

[0017] Compared with existing technologies, the electrolytic cell fluctuation adaptability testing method, system, and safe operation control method based on operating condition decomposition provided by this invention have the following beneficial effects: This invention obtains a load spectrum containing load cyclic units by performing cyclic feature decomposition on the acquired dynamic power sequence to be measured. Furthermore, it clusters the load cyclic units in the load spectrum to generate several typical operating conditions. This effectively reduces the dimensionality of the complex, non-stationary fluctuating power, which originally had strong randomness and multi-scale characteristics, into a finite number of standard stress modes with well-defined features. Subsequently, experiments or simulations are conducted on each typical operating condition to generate a corresponding single-condition degradation model. Based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum, the cumulative damage rate corresponding to the dynamic power sequence to be measured is calculated. This effectively overcomes the deficiency of traditional single aging models in being unable to reproduce real combined fluctuations. A quantitative mapping mechanism with physical consistency is established between complex fluctuating conditions and the actual degradation behavior of the electrolyzer. Finally, the fluctuation adaptability index is determined based on the cumulative damage rate, and the evaluation results are output. This achieves independent quantification of the dynamic damage contribution of the electrolyzer under unstable operating conditions and accurate assessment of the overall fluctuation adaptability. Attached Figure Description

[0018] Figure 1 This is a flowchart of the electrolytic cell fluctuation adaptability test method based on working condition decomposition of the present invention; Figure 2 This is a flowchart of the cyclic feature decomposition process of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0020] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes may be made to the function and arrangement of the described elements without departing from the scope of the invention. Various processes or components may be appropriately omitted, substituted, or added to the various examples. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined into other examples.

[0021] An embodiment of the present invention provides a method for testing the adaptability of electrolyzer fluctuations based on operating condition decomposition, the flowchart of which is shown below. Figure 1 As shown, steps S100-S600 are included: S100. Obtain the sample dynamic power sequence of the electrolytic cell during actual operation.

[0022] Since raw power sequences typically exhibit significant non-stationarity, randomness, and multi-scale fluctuations, it is difficult to identify the differentiated impacts of various fluctuation modes on the electrolyzer if durability testing or damage assessment is conducted directly using the raw power sequences. Therefore, this invention first performs cyclic feature decomposition on the sample dynamic power sequences to convert the raw power waveforms into load spectra that can be used for subsequent clustering and damage modeling.

[0023] In one specific embodiment of the present invention, the sample dynamic power sequence can be derived from the dynamic power input of the electrolyzer in a real operating scenario. The dynamic power input can be formed by factors such as renewable energy fluctuation power, grid disturbances, and load changes.

[0024] S200. Perform cyclic feature decomposition on the dynamic power sequence of the sample to obtain the load spectrum containing several load cyclic units.

[0025] In one specific embodiment of the present invention, when performing cyclic feature decomposition on the sample dynamic power sequence, the rainflow counting method is used to extract peak-valley pairs in the sample dynamic power sequence to filter out non-critical fluctuations, and further closed-loop cyclic extraction is performed to obtain several load cyclic units. Finally, the sample dynamic power sequence is transformed into a load spectrum composed of multiple load cyclic units.

[0026] The flowchart of cyclic eigenvalue decomposition is as follows: Figure 2 As shown.

[0027] Each load cycle unit includes mean power, cycle amplitude, and number of cycles, fully characterizing the cycle distribution of different amplitudes and mean values ​​for subsequent clustering.

[0028] As a feasible example, the load spectrum can be represented as a load cyclic matrix. : .

[0029] in, For the first The average power of each cycle, For amplitude, This represents the number of times the loop occurs. This represents the total number of cyclic types.

[0030] The above load spectrum retains the main statistical characteristics of power fluctuations and provides a unified input format for the subsequent extraction of typical operating conditions.

[0031] S300: Cluster the load cycle units in the load spectrum to generate several typical operating conditions.

[0032] Considering that the load spectrum usually contains a large number of load cycle elements, if each load cycle element is tested or simulated separately, the testing cost will be high and it will not be conducive to forming a load condition category that can be used for engineering judgment.

[0033] Therefore, in order to reduce the scale of experiments or simulations, the load spectrum is used as a feature space sample to input into the unsupervised clustering model to obtain a limited number of typical operating conditions. Each typical operating condition represents a set of operating conditions that are similar in power fluctuation patterns and have a consistent impact on the damage mechanism.

[0034] The processing in step S300 solves the problem that it is not feasible to use a large number of load cycle units directly in experiments and simulations in actual engineering. It reduces the dimensionality of a large number of working conditions to a limited number of typical stress modes, greatly reduces the scale of subsequent experimental calibration, and retains the main statistical characteristics of power fluctuations, accurately extracting the data related to the core degradation mechanism.

[0035] In one specific embodiment of the present invention, the load cyclic unit is clustered using the mean power and cyclic amplitude of the load cyclic unit as the sample feature vector and the number of cycles as the weight, so as to obtain several typical operating conditions that can represent the main fluctuation modes of the load spectrum.

[0036] As a feasible example, the clustering process uses the K-means clustering algorithm, and the specific process includes: initialization Given a random center, iteratively assign samples and update the center until convergence.

[0037] The number of clusters K can be optimized using the Elbow Method or the Silhouette Score, and is typically set to 3-7 to balance accuracy and experimental scale.

[0038] In this example, the K-means clustering algorithm obtains multiple cluster centers by minimizing the within-cluster variance. Its optimization objective function is expressed as: .

[0039] in, For the number of clusters, This is the feature vector of the sample. For the first Cluster centers of classes As weight.

[0040] During the clustering process, through iterative optimization, we find... Cluster centers The goal is to optimize the sum of the weighted Euclidean distances from all sample points to their respective cluster centers (i.e., the objective function). (Minimum)

[0041] Based on the above optimization objective function, load cycle units that are close to each other within a class and have representative frequency of occurrence can be grouped into the same cluster. The set of working conditions corresponding to each cluster center is determined as a typical operating condition, thereby obtaining several typical operating conditions and ensuring that the working conditions within each typical operating condition are highly similar.

[0042] Number of clusters The elbow rule can be used to help determine this.

[0043] As a further example, take Four typical operating conditions can be obtained, which respectively characterize the four typical stress modes faced by the PEM electrolyzer: TOP1-stable low load mode, TOP2-fluctuating medium load mode, TOP3-intermittent high load mode and TOP4-violent oscillation mode.

[0044] Understandable These are merely example values ​​used in the embodiments. Without departing from the technical concept of the present invention, the number of clusters can be adjusted according to the specific sample distribution and testing requirements.

[0045] S400: Conduct experiments or simulations for each typical operating condition to generate a single-condition degradation model corresponding to each typical operating condition.

[0046] In one specific embodiment of the present invention, after completing the division of typical operating conditions, the performance degradation process data of the electrolytic cell under different typical operating conditions is obtained by conducting experiments or simulations to reproduce each typical operating condition, and then a single-condition degradation model corresponding to each typical operating condition is generated based on the performance degradation process data.

[0047] As a feasible example, a hardware-in-the-loop (HIL) test platform can be used to reproduce the various typical operating conditions of a 200kW PEM hydrogen production system, and collect performance parameters such as voltage, impedance, and temperature during the test to monitor key health indicators such as the voltage decay rate and ohmic internal resistance changes of the electrolyzer in real time.

[0048] A single-condition degradation model was constructed based on the degradation data collected experimentally. For the first... A typical operating condition, its performance parameters The pattern of change over time is expressed as follows: .

[0049] in, These are the initial performance values. The attenuation magnitude, For normalized degenerate functions, This represents the degradation parameter obtained through experimental fitting.

[0050] The above expressions can be used to quantify the impact of different typical operating conditions on the performance degradation of electrolyzers.

[0051] Furthermore, for the early rapid degradation stage mainly characterized by catalyst dissolution or ionomer degradation, this invention employs an exponential function to construct the degradation function of the single-condition degradation model. The exponential function can be expressed as: .

[0052] in, This represents the degradation rate, used to characterize the rate of change during the early, rapid decay phase.

[0053] For the medium- to long-term nonlinear degradation stage, which includes membrane thinning or gas diffusion layer aging, this embodiment employs a polynomial function to construct the degradation function of the single-condition degradation model. Taking a quadratic polynomial as an example, it can be expressed as: .

[0054] in, The fitting coefficients are denoted as .

[0055] The aforementioned exponential and polynomial functions can be fitted with parameters based on experimental or simulation data using the least squares method to correspond to the damage evolution of the electrolytic cell at different degradation stages, respectively.

[0056] After step S400, a typical operating condition library and a single operating condition model library are obtained. The typical operating condition data in the typical operating condition library correspond one-to-one with the single operating condition model in the single operating condition model library. Each typical operating condition data is used to characterize a class of operating conditions with similar power fluctuation characteristics and similar damage mechanisms.

[0057] S500: Obtain the dynamic power sequence of the electrolytic cell during actual operation.

[0058] In step S500, the dynamic power sequence to be measured of the electrolytic cell during actual operation is collected to evaluate the fluctuation adaptability of the electrolytic cell during actual operation. Specifically, the dynamic power sequence to be measured can be the actual input power sequence of the electrolytic cell over a period of time.

[0059] It should be noted that the dynamic power sequence to be tested obtained in step S500 is used for online evaluation and is not used to rebuild the typical operating condition library and the single operating condition degradation model library.

[0060] S600: Perform cyclic feature decomposition on the dynamic power sequence under test to obtain a load spectrum containing several load cyclic units.

[0061] In one specific embodiment of the present invention, step S600 first employs the same cyclic feature decomposition method as step S200 to perform cyclic feature decomposition on the dynamic power sequence under test to obtain the load spectrum under test. Specifically, the rainflow counting method can be used to extract peak-valley pairs in the dynamic power sequence under test, and closed-loop cyclic extraction can be performed to obtain several cyclic units of the load under test.

[0062] S700 classifies the load cycle units in the load spectrum into a typical operating condition.

[0063] In one specific embodiment of the present invention, the load cycling units are classified into a typical operating condition in the following manner: For any dynamic power sequence to be measured, after completing the cyclic eigenvalue decomposition, multiple load cyclic elements can be obtained, and each load cyclic element can be represented as a corresponding eigenvector. Then, the eigenvectors of each load cycle element are calculated separately. Cluster centers corresponding to each typical operating condition Distance between The distance can be expressed as:

[0064] in, This represents the distance between the i-th load cycle unit and the k-th typical operating condition. This represents the eigenvector of the i-th load cycle unit. This represents the cluster center corresponding to the k-th typical operating condition.

[0065] Furthermore, based on the principle of minimum distance, the i-th load cycle unit can be assigned to a typical operating condition that satisfies the following formula: ; in, This represents the typical operating condition that the i-th load cycle unit is ultimately classified into.

[0066] It is understood that, without departing from the technical concept of this invention, a similarity measurement method can be used instead of a distance measurement method, and the attribution can be completed based on the principle of maximum similarity.

[0067] After classifying all load cycle units, the load cycle units belonging to each typical operating condition can be statistically analyzed, and the distribution ratio of each typical operating condition in the dynamic power sequence under test can be determined based on the number of cycles corresponding to each load cycle unit.

[0068] As a feasible example, the distribution ratio of the k-th typical operating condition It can be represented as: ; in, This represents the set of load cycle elements that are classified as the kth type of typical operating condition. denoted by , which represents the number of cycles corresponding to the i-th load cycle unit, and m represents the total number of load cycle units obtained after the cyclic feature decomposition of the dynamic power sequence under test.

[0069] After obtaining the load spectrum, the typical operating conditions corresponding to each load cycle unit in the load spectrum to be tested are further determined based on the typical operating condition library formed in step S300.

[0070] Using the above method, each load cycle unit in the dynamic power sequence under test can be classified into a typical operating condition in the typical operating condition library, thereby obtaining the distribution results of typical operating conditions corresponding to the dynamic power sequence under test.

[0071] S800, based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum, calculates the cumulative damage rate corresponding to the dynamic power sequence under test.

[0072] Through steps S500-S700, any complex fluctuating working condition is decomposed into the superposition of various typical operating conditions in step S300. Step S800 corresponds to the single-condition degradation model generated in step S400, which combines the single-condition damage according to the distribution ratio of the load spectrum, thus solving the problem of the overall damage being unquantifiable under complex combined working conditions.

[0073] In one specific embodiment of the present invention, step S800 firstly calculates the single-condition damage curve corresponding to each typical operating condition using the single-condition degradation model, and then weights and superimposes the single-condition damage curve with the corresponding distribution ratio to obtain the cumulative damage model.

[0074] As a feasible example, based on Miner's linear damage accumulation theory, the cumulative damage model... The single-condition damage of the single-condition degradation model is superimposed using the following formula: .

[0075] in, For the first Similar to TOP in duration The cumulative damage value generated internally.

[0076] After generating the cumulative damage model, the cumulative damage rate can be calculated from the cumulative damage model over time. The growth rate is represented by [the value / rate].

[0077] For any dynamic power sequence to be tested, steps S500-S700 can be executed to determine the distribution ratio of each typical operating condition in the load spectrum corresponding to the dynamic power sequence to be tested.

[0078] S900: Determine the fluctuation adaptability index of the electrolyzer based on the cumulative damage rate, and output the operation adaptability evaluation results corresponding to the fluctuation adaptability index.

[0079] Specifically, the fluctuation adaptability index is characterized by the cumulative damage rate, i.e. the growth slope of the cumulative damage curve. The larger the cumulative damage rate, the faster the damage accumulates under the dynamic power sequence under test, and the worse the electrolyzer's ability to adapt to the fluctuation condition. The smaller the cumulative damage rate, the better the electrolyzer's ability to adapt to the fluctuation condition.

[0080] Furthermore, the operational adaptability evaluation results can be output by comparing the fluctuation adaptability index with the preset adaptability index threshold range, so as to output the corresponding operational adaptability evaluation results according to the different index ranges in which the fluctuation adaptability index is located.

[0081] Specifically, a first threshold and a second threshold can be preset. When the fluctuation adaptability index is less than the first threshold, a high adaptability evaluation result is output, indicating that the fluctuation condition corresponding to the dynamic power sequence under test has a weak cumulative damage effect on the electrolyzer, and normal operation or priority access is allowed. When the fluctuation adaptability index is greater than or equal to the first threshold and less than the second threshold, a medium adaptability evaluation result is output, indicating that the fluctuation condition corresponding to the dynamic power sequence under test has a certain cumulative damage effect on the electrolyzer, and the fluctuation amplitude, duration, or access ratio can be constrained. When the fluctuation adaptability index is greater than or equal to the second threshold, a low adaptability evaluation result is output, indicating that the fluctuation condition corresponding to the dynamic power sequence under test will lead to rapid damage accumulation, and access should be restricted, derated, and / or life management interventions should be implemented for such high-damage fluctuation conditions.

[0082] To improve adaptability to different operating stages or application scenarios, in some improved embodiments, the electrolytic cell fluctuation adaptability test method based on operating condition decomposition of the present invention further includes step S1000 after step S900: obtaining the measured performance data of the electrolytic cell and determining the true damage value; calculating the error between the true damage value and the predicted damage value calculated based on the cumulative damage rate; if the error exceeds a preset threshold, then using an optimization algorithm to back-update the model parameters of the single-operating-condition degradation model or weighting and superimposing the corresponding weights to perform closed-loop correction.

[0083] The above closed-loop correction process involves collecting measured performance data of the electrolytic cell and determining the actual damage value based on this data. Then, the actual damage value is compared with the predicted damage value calculated from the cumulative damage rate to determine whether the model needs to be updated.

[0084] In a preferred embodiment, the damage value can be characterized by the amount of voltage degradation in the electrolytic cell.

[0085] In one specific embodiment of the present invention, when using a cumulative damage model over time... When the growth rate represents the cumulative damage rate, the model prediction value of the cumulative damage model can be directly used as the predicted damage value.

[0086] Specifically, during the actual operation of the electrolytic cell system, measured performance data of the electrolytic cell are collected periodically. and compared with model predictions Compare the results and calculate the test error.

[0087] As a feasible example, test error can be expressed as Mean Absolute Percentage Error (MAPE), calculated as follows: .

[0088] When the error exceeds a preset threshold (e.g., 5%), a closed-loop correction process is triggered.

[0089] Specifically, the closed-loop correction process can use the gradient descent optimization algorithm to update the parameters of the single-condition degradation model in step S400 in reverse. Another embodiment of the present invention provides a safe operation control method for an electrolytic cell. This safe operation control method first calls the electrolytic cell fluctuation adaptability test method based on operating condition decomposition of the aforementioned embodiment to obtain the fluctuation adaptability index corresponding to the electrolytic cell.

[0090] The safe operating boundary of the electrolyzer is constructed based on the fluctuation adaptability index. The safe operating boundary is used to distinguish between operating conditions that meet the safe operating requirements and high-damage fluctuation conditions that require restricted access.

[0091] In one specific embodiment of the present invention, the following power regulation strategy can be formulated based on the safe operating boundary to control the safe operation of the electrolyzer.

[0092] The dynamic power sequence to be accessed is determined based on the safe operation boundary, and the corresponding operation and control strategy is output: When the dynamic power sequence to be connected is within the safe operating boundary, it can be allowed to connect normally or operate according to the predetermined power command; when the dynamic power sequence to be connected approaches the safe operating boundary, a restrictive operation control strategy can be adopted for it, including limiting the power fluctuation amplitude, limiting the power ramp-up rate and / or decline rate, limiting the duration of high damage fluctuation conditions, and limiting the connection ratio of high damage typical operating conditions; when the dynamic power sequence to be connected exceeds the safe operating boundary, the control result of limiting connection, derated operation, and / or switching to conservative operation mode is output.

[0093] Furthermore, life management of the electrolyzer can be carried out based on fluctuation adaptability indicators, cumulative damage rate and / or cumulative damage amount, including adjusting the operating time under high damage conditions, increasing the proportion of low damage conditions, optimizing maintenance cycles and outputting life warning information, so as to meet the requirements of safe operation while taking into account the long-term durability of the electrolyzer.

[0094] The safe operation control method of the present invention can formulate power grid dispatch control strategies based on the results of adaptability tests, which is beneficial to restrict the access of low adaptability and high damage combination conditions from the source, thereby realizing proactive health management of hydrogen production equipment.

[0095] Another embodiment of the present invention provides an electrolytic cell fluctuation adaptability testing system based on operating condition decomposition. Specifically, the testing system includes a data acquisition module, a decomposition module, a clustering module, a classification module, a simulation modeling module, a calculation and evaluation module, and an output module.

[0096] Among them, the data acquisition module is used to acquire several sample dynamic power sequences and the dynamic power sequence to be measured during the actual operation of the electrolytic cell; The decomposition module is used to perform cyclic feature decomposition on the sample dynamic power sequence and the dynamic power sequence to be measured to obtain a load spectrum containing several load cyclic units. The clustering module is used to cluster load cyclic units in the load spectrum to generate several typical operating conditions; The classification module is used to classify the load cyclic units of the dynamic power sequence under test into a typical operating condition. The simulation modeling module is used to conduct experiments or simulations for each typical operating condition and generate a single-condition degradation model corresponding to each typical operating condition. The calculation and evaluation module is used to calculate the cumulative damage rate corresponding to the dynamic power sequence under test based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test. The output module is used to output the cumulative damage rate as an indicator of the electrolyzer's fluctuating adaptability.

[0097] In some further embodiments, the system also includes a closed-loop correction module.

[0098] The closed-loop correction module is used to obtain the actual operating damage value of the electrolytic cell, calculate the error between the actual operating damage value and the predicted damage value calculated based on the cumulative damage rate, and if the error exceeds the preset threshold, the model parameters in the single-condition degradation model are updated in reverse using the optimization algorithm to perform closed-loop correction.

[0099] The process of implementing the steps in the above embodiments for each module will not be described again here.

[0100] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0101] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and practicing the disclosure herein. This invention is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for testing the adaptability of electrolytic cells to fluctuations based on operating condition decomposition, characterized in that, include: Obtain several sample dynamic power sequences of the electrolyzer during actual operation; Cyclic feature decomposition is performed on the sample dynamic power sequence to obtain a load spectrum containing several load cyclic units; When performing cyclic feature decomposition on the sample dynamic power sequence, the rainflow counting method is used to extract peak-valley pairs in the sample dynamic power sequence to filter out non-critical fluctuations, and further closed-loop cyclic extraction is performed to obtain several load cyclic units. Finally, the sample dynamic power sequence is transformed into a load spectrum composed of multiple load cyclic units. Several typical operating conditions are generated by clustering the load cycle units in the load spectrum; Experiments or simulations were conducted for each of the typical operating conditions to generate a single-condition degradation model corresponding to each of the typical operating conditions. Obtain the dynamic power sequence of the electrolyzer during actual operation; Calculate the load spectrum corresponding to the dynamic power sequence under test, and classify the load cycle units into one of the typical operating conditions; Based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test, the cumulative damage rate corresponding to the dynamic power sequence under test is calculated. The fluctuation adaptability index of the electrolytic cell is determined based on the cumulative damage rate, and the operational adaptability evaluation result corresponding to the fluctuation adaptability index is output.

2. The electrolytic cell fluctuation adaptability testing method based on operating condition decomposition as described in claim 1, characterized in that, The elements in the load spectrum include mean power, cyclic amplitude, and number of cycles.

3. The electrolytic cell fluctuation adaptability test method based on operating condition decomposition as described in claim 2, characterized in that, Several typical operating conditions are generated by clustering the load cycle units in the load spectrum, including: Using the mean power and cycle amplitude of the load cycle unit as the sample feature vector, and the number of cycles as the weight, the K-means clustering algorithm is used to minimize the intra-cluster variance to obtain multiple cluster centers. Each cluster center corresponds to a set of operating conditions as a typical operating condition.

4. The electrolytic cell fluctuation adaptability test method based on operating condition decomposition as described in claim 1, characterized in that, Cyclic feature decomposition is performed on the dynamic power sequence to be measured to obtain a load spectrum containing several load cyclic units. Specifically, the peak-valley pairs in the dynamic power sequence to be measured are extracted using the rainflow counting method to filter out non-critical fluctuations, and closed-loop cyclic extraction is performed to obtain several load cyclic units and construct the load spectrum.

5. The electrolytic cell fluctuation adaptability test method based on operating condition decomposition as described in claim 1, characterized in that, When conducting experiments or simulations for each of the typical operating conditions to generate a single-condition degradation model corresponding to each of the typical operating conditions, an exponential function is used to construct the single-condition degradation model for the early rapid decay stage. For the medium- to long-term nonlinear degradation stage, the single-condition degradation model is constructed using a polynomial function.

6. The electrolytic cell fluctuation adaptability test method based on operating condition decomposition as described in claim 1, characterized in that, Based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test, the cumulative damage rate corresponding to the dynamic power sequence under test is calculated, including: Using the single-condition degradation model corresponding to each of the typical operating conditions, the single-condition damage curves corresponding to each of the typical operating conditions are calculated respectively. The single-condition damage curves corresponding to each typical operating condition are weighted and superimposed with the corresponding distribution ratios to obtain a cumulative damage model. The cumulative damage rate is calculated using the cumulative damage model.

7. The electrolytic cell fluctuation adaptability test method based on operating condition decomposition as described in claim 1, characterized in that, The method also includes: Obtain the actual operational damage value of the electrolytic cell; The error between the actual damage value and the predicted damage value calculated based on the cumulative damage rate is calculated. If the error exceeds a preset threshold, the model parameters of the single-condition degradation model are updated in reverse using an optimization algorithm to perform closed-loop correction.

8. A method for safe operation control of an electrolytic cell, characterized in that, include: The fluctuation adaptability test method of the electrolytic cell based on operating condition decomposition as described in any one of claims 1-7 is used to obtain the fluctuation adaptability index of the electrolytic cell. The safe operating boundary of the electrolyzer is constructed based on the fluctuation adaptability index of the electrolyzer. A power regulation strategy is formulated based on the safe operating boundary to control the safe operation of the electrolytic cell.

9. A test system for the adaptability of electrolytic cells to fluctuations based on operating condition decomposition, characterized in that, include: The data acquisition module is used to acquire several sample dynamic power sequences and the dynamic power sequence to be measured during the actual operation of the electrolytic cell. The decomposition module is used to perform cyclic feature decomposition on the sample dynamic power sequence and the dynamic power sequence to be tested to obtain a load spectrum containing several load cyclic units. When performing cyclic feature decomposition on the sample dynamic power sequence, the rainflow counting method is used to extract peak-valley pairs in the sample dynamic power sequence to filter out non-critical fluctuations, and further closed-loop cyclic extraction is performed to obtain several load cyclic units. Finally, the sample dynamic power sequence is transformed into a load spectrum composed of multiple load cyclic units. The clustering module is used to cluster the load cycle units in the load spectrum to generate several typical operating conditions; The classification module is used to classify the load cyclic units of the dynamic power sequence under test into a typical operating condition. The simulation modeling module is used to conduct experiments or simulations on each of the typical operating conditions and generate a single-condition degradation model corresponding to each of the typical operating conditions. The calculation and evaluation module is used to calculate the cumulative damage rate corresponding to the dynamic power sequence under test based on the single-condition degradation model and the distribution ratio of each typical operating condition in the load spectrum of the dynamic power sequence under test. The output module is used to output the cumulative damage rate as a fluctuation adaptability index of the electrolytic cell.

10. The electrolytic cell fluctuation adaptability testing system based on operating condition decomposition as described in claim 9, characterized in that, The system also includes: The closed-loop correction module is used to obtain the actual operating damage value of the electrolytic cell, calculate the error between the actual operating damage value and the predicted damage value calculated based on the cumulative damage rate, and if the error exceeds a preset threshold, use an optimization algorithm to update the model parameters in the single-condition degradation model in reverse to perform closed-loop correction.

Images