Electrolytic tank health state detection method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to an electrolytic cell health state detection method and device, computer equipment, a readable storage medium and a program product. A node relation graph is constructed based on time-domain signals such as high-frequency impedance and oxygen flow of a to-be-detected electrolytic tank, frequency-domain signals such as impedance spectrum and the like and the real-time health state of the electrolytic tank, the health state of the electrolytic tank is detected based on the node relation graph through a detection model, and a health state detection result corresponding to the electrolytic tank is output. Compared with traditional health state detection depending on an equivalent circuit model, the method has the advantages that the node relation graph is constructed by combining the time-domain signal and the frequency-domain signal of the electrolytic tank and the health state node of the electrolytic tank, and the health state of the electrolytic tank is detected by combining the detection model with the node relation graph; and the detection accuracy of the health state of the electrolytic tank is improved.

Description

Technical Field

[0001] This application relates to the field of electrolytic cell technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for detecting the health status of an electrolytic cell. Background Technology

[0002] Proton exchange membrane electrolyzers (PEMECs) have become an important part of the future green energy architecture due to their high energy efficiency and cleanliness. However, the performance degradation of electrolyzers is caused by multiple heterogeneous sources. They are not only subject to the coupled degradation mechanisms of platinum, carbon, and membrane, but also constrained by cumulative operating variables such as operating parameters and membrane water content. Therefore, the health status monitoring of highly reliable electrolyzers remains a major challenge.

[0003] During long-term operation, electrolytic cells inevitably experience performance degradation due to multiple factors, including electrochemical reactions, uneven mass and heat transfer, material aging, and fluctuations in operating conditions. This degradation typically manifests as increased voltage, decreased efficiency, and increased energy consumption. Current methods for assessing the health status of electrolytic cells usually rely on equivalent circuit models, which have limited adaptability to complex changes in operating conditions, leading to reduced detection accuracy.

[0004] Therefore, current methods for detecting the health status of electrolytic cells suffer from low accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for detecting the health status of electrolytic cells that can improve the accuracy of detection, in order to address the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for detecting the health status of an electrolytic cell, including:

[0007] Acquire the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be tested; the time-domain signal includes the high-frequency impedance and oxygen flow rate of the electrolytic cell during operation, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell;

[0008] Obtain the node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell.

[0009] The node relationship diagram is input into the detection model; the detection model is used to output the health status detection result of the electrolytic cell based on the node relationship diagram.

[0010] Secondly, this application also provides an electrolytic cell health status detection device, comprising:

[0011] The first acquisition module is used to acquire the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be detected; the time-domain signal includes the high-frequency impedance and oxygen flow rate of the electrolytic cell during operation, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell.

[0012] The second acquisition module is used to acquire a node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell.

[0013] The detection module is used to input the node relationship diagram into the detection model; the detection model is used to output the health status detection result of the electrolytic cell according to the node relationship diagram.

[0014] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0015] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0016] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0017] The aforementioned electrolytic cell health status detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product construct a node relationship diagram based on time-domain signals such as high-frequency impedance and oxygen flow rate, frequency-domain signals such as impedance spectrum, and the real-time health status of the electrolytic cell. A detection model then detects the electrolytic cell's health status based on this node relationship diagram and outputs the corresponding health status detection result. Compared to traditional methods relying on equivalent circuit models for health status detection, this application improves the accuracy of electrolytic cell health status detection by combining the electrolytic cell's time-domain signals, frequency-domain signals, and health status nodes to construct a node relationship diagram. The detection model then uses this node relationship diagram to detect the electrolytic cell's health status, thus enhancing the accuracy of the detection. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a method for detecting the health status of an electrolytic cell in one embodiment;

[0020] Figure 2 This is a schematic diagram of the node relationship diagram in one embodiment;

[0021] Figure 3 This is a schematic diagram of the model detection steps in one embodiment;

[0022] Figure 4 This is a flowchart illustrating the electrolytic cell health status detection method in another embodiment;

[0023] Figure 5 This is a structural block diagram of an electrolytic cell health status detection device in one embodiment;

[0024] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0027] Among related technologies, proton exchange membrane electrolyzers (PEMECs) have become an important part of the future green energy architecture due to their high energy efficiency and cleanliness. However, the performance degradation of electrolyzers is caused by "multi-source heterogeneity," not only suffering from the coupled degradation mechanisms of platinum, carbon, and membrane, but also constrained by cumulative operating variables such as operating parameters and membrane water content. Assessing the health status of highly reliable electrolyzers remains a major challenge.

[0028] During long-term operation, electrolytic cells inevitably experience performance degradation due to multiple factors, including electrochemical reactions, uneven mass and heat transfer, material aging, and fluctuations in operating conditions. This degradation typically manifests as increased voltage, decreased efficiency, and increased energy consumption. However, the health status of an electrolytic cell cannot be characterized by a single physical quantity but is determined by multiple degradation mechanisms, such as the evolution of polarization loss, changes in mass transfer resistance, electrode activity decay, and structural aging. Existing methods often rely on equivalent circuit models, making it difficult to accurately model aging phenomena under multiple operating conditions and over long periods. Alternatively, they may rely on empirical indicators based on external phenomena such as voltage and current, which struggle to decouple the coupling effects between different degradation mechanisms and have limited adaptability to complex operating conditions, failing to meet the requirements for high-precision, real-time health status assessment. Data-driven methods, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs), only statistically analyze long-term aging trends, resulting in limited interpretability and generalization ability.

[0029] To address the aforementioned issues, this application proposes a method for assessing the health status of electrolytic cells based on Graph Convolutional Network-Long Short-Term Memory (GCN-LSTM). By constructing a graph model of degradation mechanisms to characterize the influence relationships between different degradation mechanisms, using the graph convolutional network to extract spatially coupled features of multiple mechanisms, and combining it with the long short-term memory network to perform temporal modeling of the degradation evolution process, a high-precision, real-time assessment and prediction of the health status of electrolytic cells can be achieved.

[0030] In one embodiment, such as Figure 1 As shown, a method for detecting the health status of an electrolytic cell is provided. This embodiment illustrates the application of this method to a server. It is understood that this method can also be applied to a terminal, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server, including the following steps S202 to S206. Wherein:

[0031] Step S202: Obtain the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be tested; the time-domain signal includes the high-frequency impedance and oxygen flow rate during the operation of the electrolytic cell, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell.

[0032] The server can be a device used to monitor the health status of an electrolytic cell, and the electrolytic cell to be monitored can be one that requires health status assessment and monitoring. The electrolytic cell can be a proton exchange membrane electrolytic cell. To monitor the health status of the electrolytic cell, the server can acquire the corresponding time-domain and frequency-domain signals. The time-domain signal can include various parameters, such as the high-frequency impedance and oxygen flow rate during electrolytic cell operation. The frequency-domain signal can include the impedance spectrum of the electrolytic cell.

[0033] Step S204: Obtain the node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell.

[0034] The server can predefine the health status of the electrolytic cell. For example, the server constructs the equation corresponding to the health status of the electrolytic cell through operating condition conversion and internal cumulative parameter normalization strategies. Based on the real-time operating parameters of the electrolytic cell and the above-mentioned equation corresponding to the health status, the server obtains the nodes corresponding to the real-time health status of the electrolytic cell. The server can also construct corresponding nodes based on the above-mentioned high-frequency impedance, oxygen flow rate, and impedance spectrum. Thus, the server can combine the above-mentioned constructed nodes and the relationships between the nodes to construct a node relationship graph corresponding to the health status of the electrolytic cell. Each node in the node relationship graph includes the nodes corresponding to the above-mentioned time-domain signals and frequency-domain signals, as well as the nodes corresponding to the health status. Each edge in the node relationship graph can be the influence relationship between the nodes, such as the influence relationship between the nodes corresponding to the time-domain signals or frequency-domain signals on the health status nodes, or the influence relationship between the nodes corresponding to the time-domain signals and frequency-domain signals.

[0035] Step S206: Input the above node relationship diagram into the detection model; the above detection model is used to output the health status detection result corresponding to the above electrolytic cell based on the above node relationship diagram.

[0036] The server can pre-train the detection model to be trained, resulting in a trained detection model. This model can include multiple networks, such as graph convolutional networks and long short-term memory networks. The server can input the node relationship graph into the detection model, which processes the graph using both the graph convolutional network and the long short-term memory network to detect the health status of the electrolytic cell and output the corresponding health status detection result. This result includes information about the electrolytic cell's health status for the current time period or a preset future time period. The health status detection result can be determined based on the parameters of the health status nodes in the node relationship graph. Different parameters represent different health states of the electrolytic cell.

[0037] In the aforementioned method for detecting the health status of an electrolytic cell, a node relationship graph is constructed based on time-domain signals such as high-frequency impedance and oxygen flow rate, frequency-domain signals such as impedance spectrum, and the real-time health status of the electrolytic cell. A detection model then detects the health status of the electrolytic cell based on this node relationship graph and outputs the corresponding health status detection result. Compared to traditional methods that rely on equivalent circuit models for health status detection, this application improves the accuracy of electrolytic cell health status detection by combining the time-domain signals, frequency-domain signals, and health status nodes of the electrolytic cell to construct a node relationship graph. The detection model then uses this node relationship graph to detect the health status of the electrolytic cell, thus enhancing the accuracy of the detection.

[0038] In one embodiment, obtaining the node relationship graph corresponding to the health status of the electrolytic cell includes: generating each first node based on the high-frequency impedance in the time-domain signal, the oxygen flow rate, and the impedance spectrum in the frequency-domain signal; generating a second node corresponding to the real-time health status of the electrolytic cell based on the initial electrical parameters and real-time electrical parameters of the electrolytic cell; generating each edge based on the relationship between each first node and the second node; and generating the node relationship graph based on each first node, the second node, and the edges.

[0039] In this embodiment, the server can construct a node relationship graph by combining various time-domain and frequency-domain signals. For example, the server generates first nodes based on the aforementioned high-frequency impedance and oxygen flow rate in the time-domain signal and the aforementioned impedance spectrum in the frequency-domain signal. There can be multiple first nodes, each representing a node generated based on the time-domain or frequency-domain signals of the electrolytic cell. The server can also generate second nodes corresponding to the real-time health status of the electrolytic cell based on its initial and real-time electrical parameters. The initial electrical parameters represent the electrical parameters of the electrolytic cell before its first use, and the real-time electrical parameters represent the electrical parameters of the electrolytic cell at the current time. These electrical parameters can include various types. The edges between the nodes can be generated based on the relationships between the first nodes and the relationships between the first and second nodes. Thus, the server can generate the node relationship graph based on the first nodes, the second nodes, and the edges.

[0040] Specifically, the server first defines the health state of the electrolyzer. Since apparent parameters such as voltage only reflect the battery's current operating conditions (e.g., current, pressure, oxygen permeate ratio, and temperature), internal cumulative parameters, such as the load-carrying capacity under membrane water content, are less relevant. To decouple the influence of irrelevant factors and define the true battery health state, the server can define the electrolyzer health state through operating condition calculations and internal cumulative parameter normalization strategies. Specifically, from the Nernst equation and polarization loss, the electrolyzer voltage equation satisfies:

[0041] V cell =E rev +RT / 2Fp H2 (p O2 ) 1 / 2 +η act +η ohm +η con Among them, E rev Represents the reversible voltage, R is the gas constant, T is the PEMEC temperature, F represents the Faraday constant, and p i This represents the gas pressure i. And η represents the gas pressure. act ,η ohm and η con Representing activation, ohmic, and concentration loss respectively, their specific expressions are as follows: {η act =RT / (α a a F)ln(I / I a,ref )+(RT / α c c F)ln(I / I c,ref );η ohm =R ohm I;η con =RT / 2F(ln(I lim / (I lim-I)))}。 Where, α a a and α c c I, representing the positive coefficients of the anodic and cathodic reactions. a,ref I c,ref and I lim These represent the battery current density, anode and cathode reference current densities, and limiting current densities, respectively. Furthermore, the conductivity of the Nafion proton exchange membrane is related to the modal water λ as follows: σ = exp[1268(1 / 303-1 / (273+T))](0.005139λ-0.00326); the ohmic resistance includes membrane resistance and other resistances (R). oth Examples include contact resistance and terminal resistance: R ohm =σ / S MEA +R oth Therefore, the equation for the electrolytic cell voltage can be rewritten as: V cell =f(I,p,T,λ). Wherein, the initial parameter set Ω of the above equation... bol (Initial electrical parameters) are affected by degradation effects and change with battery operation; their real-time parameter set Ω t (Real-time electrical parameters) can be obtained by inverse solution of the above equation: Ω t =f -1 (I t ,p t ,T t ,λ t ).

[0042] Based on operating condition conversion and modal water normalization, the server defines the standard condition health status index of the electrolyzer as SOH. ref The corresponding standard case is [I] ref ,p ref ,T ref ,λ ref So, what is the real-time health status (SOH) of the electrolytic cell? t The standard voltage ratio corresponding to the current parameter set and the initial parameter set can be defined: SOH t =(f(I ref ,p ref ,T ref ,λ ref )|Ω=Ω t ) / (f(I ref ,p ref ,T ref ,λ ref )|Ω=Ω bol The initial electrical parameters represent the internal characteristic parameters of the electrolyzer when it leaves the factory or is in a brand-new state, while the real-time electrical parameters characterize the current internal characteristic parameters of the battery after running for time t. Since Ω cannot be directly measured... t(For example, membrane water content), the server can measure the current voltage V. t and operating conditions (I) t ,p t ,T t ,λ t By performing inverse computation (or parameter identification) on model f, the current internal parameter set Ω is calculated. t The aforementioned real-time health status can be a dimensionless ratio between 0 and 1, where a value greater than 1 indicates an abnormal performance improvement, such as activation, and a value less than 1 indicates degradation.

[0043] In this embodiment, the server can combine the time-domain signal, frequency-domain signal, and health status nodes of the electrolytic cell to construct a node relationship graph for detecting the health status of the electrolytic cell, thereby improving the accuracy of the detection of the health status of the electrolytic cell.

[0044] In one embodiment, generating each first node based on the high-frequency impedance in the time-domain signal, the oxygen flow rate, and the impedance spectrum in the frequency-domain signal includes: generating corresponding mass transfer characteristic nodes, oxygen reduction reaction nodes, and cathode catalyst layer proton transfer effect nodes based on the impedance spectrum and relaxation time distribution algorithm; the mass transfer characteristic nodes characterize the porosity of the gas diffusion layer and the health status of the equivalent transfer coefficient of the electrolytic cell; the oxygen reduction reaction nodes characterize the electrochemical reaction health status of the catalyst, ionic conductor, and electronic conductor; the cathode catalyst layer proton transfer effect nodes characterize the proton transfer health status of the membrane components and the catalyst layer; and obtaining each of the first nodes based on the nodes corresponding to the oxygen flow rate, the high-frequency impedance, the mass transfer characteristic nodes, the oxygen reduction reaction nodes, and the cathode catalyst layer proton transfer effect nodes.

[0045] In this embodiment, each of the aforementioned first nodes can correspond to a time-domain signal and a frequency-domain signal. The time-domain signal may include data such as high-frequency impedance and oxygen flow rate, while the frequency-domain signal may include data such as impedance spectrum. The server can perform feature analysis and extraction on the frequency-domain signal to obtain the nodes corresponding to the frequency-domain signal. Specifically, the server generates corresponding mass transfer feature nodes, oxygen reduction reaction nodes, and cathode catalyst layer proton transfer effect nodes based on the aforementioned impedance spectrum and relaxation time distribution algorithm. The aforementioned mass transfer feature nodes characterize the porosity and equivalent transfer coefficient health status of the gas diffusion layer of the electrolytic cell; the aforementioned oxygen reduction reaction nodes characterize the electrochemical reaction health status of the catalyst, ionic conductor, and electronic conductor; and the aforementioned cathode catalyst layer proton transfer effect nodes characterize the proton transfer health status of the membrane components and the catalyst layer. The server can also generate corresponding nodes based on the aforementioned oxygen flow rate and the aforementioned high-frequency impedance, and then obtain each of the aforementioned first nodes based on the nodes corresponding to the aforementioned oxygen flow rate, the nodes corresponding to the high-frequency impedance, the aforementioned mass transfer feature nodes, the aforementioned oxygen reduction reaction nodes, and the aforementioned cathode catalyst layer proton transfer effect nodes.

[0046] Specifically, the server acquires time-domain signals X, such as high-frequency impedance and oxygen flow rate, under the operating conditions of the electrolytic cell training set through multi-source operational data acquisition and preprocessing. t =[X t1 ,X t2 [ ] and frequency domain electrochemical impedance spectroscopy R(f) and other operating data.

[0047] The server can identify polarization processes with different time constants in complex electrochemical systems through the Distribution of Relaxation Times (DRT). For impedance spectrum R(f) data, complete DRT analysis can yield j sets of characteristic DRT peak resistances and their corresponding frequencies [(R...]. DRT1 ,f DRT1 ),(R DRT2 ,f DRT2 ),…(R DRTj ,f DRTj Since the degradation process affects not only the magnitude of the characteristic impedance but also its response constant, the server introduces position coding to perform impedance-frequency fusion on the DRT features to obtain the final frequency domain signal X. f :

[0048] X f = =[X f1 ,X f2 ,…,X fj ]. Wherein, ψ and d model Both represent normalized parameters, log 10This is a logarithmic operation.

[0049] The server performs anomaly removal and max-min normalization on the above time-frequency signal to obtain x. norm The server's anomaly removal and normalization processing of the above signals can be expressed as: x norm =(xx min ) / (x max -x min ), where x norm x max and x min These represent the maximum and minimum values ​​of the variable group after normalization and before normalization, respectively.

[0050] The server can construct a degradation mechanism graph (node ​​relationship graph) by combining the above signals. Specifically, the server abstracts key variables characterizing different degradation mechanisms of the electrolytic cell as graph nodes, and the influence relationships between different degradation mechanisms as graph edges. The server can select six graph nodes, as shown below. Figure 2 As shown, Figure 2 This is a schematic diagram of the node relationship in one embodiment. Nodes 1, 2, and 3 represent the DRT and the mass transfer characteristics, oxygen reduction reaction, and proton transfer effect of the cathode catalyst layer obtained through impedance-frequency fusion processing, respectively. Nodes 4 and 5 represent the high-frequency impedance and oxygen flow time-domain signals acquired by the sensor, respectively. Node 6 represents the real-time health status (SOH). t Based on the coupling relationship of the above six nodes, the server predefines the graph and edges of the GCN and performs relationship analysis based on standard operating conditions.

[0051] Among them, Node 1: Mass transfer characteristics characterize the porosity and equivalent transfer coefficient of the gas diffusion layer, thus it is also coupled with Node 4 (oxygen flow rate). Node 2: Oxygen Reduction Reaction (ORR) characteristics characterize the health of the electrochemical reaction at the three-phase interface of the catalyst, ion conductor (ionomer), and electronic conductor (carbon support). ORR requires proton participation, and in PEMEC, proton transfer depends on sufficient modal water; therefore, it is also coupled with Node 3 (cathode proton transfer) and Node 5 (high-frequency impedance). Node 3: Cathode proton transfer characteristics characterize the health of proton transfer between the membrane components and the catalyst layer. In addition to the aforementioned coupling relationships, the hydrogen ion transfer rate affects the reaction rate, thus affecting the oxygen flow rate at Node 4. Node 4: Oxygen flow rate characterizes the gas diffusion layer transfer and the health of the electrochemical reaction. Besides Nodes 1, 2, and 3, there are no other nodes with strong coupling effects. Node 5: High-frequency impedance characterizes the health of the membrane components. Besides Nodes 2 and 3, there are no other nodes with strong coupling effects. Therefore, the server constructs a node relationship graph by combining the above nodes and the relationships between them.

[0052] Through this embodiment, the server can combine multiple nodes and the relationships between nodes to construct a node relationship graph for detecting the health status of the electrolytic cell, thereby improving the accuracy of the detection of the health status of the electrolytic cell.

[0053] In one embodiment, the detection model includes a graph convolutional network and a long short-term memory network; the graph convolutional network is used to obtain the corresponding feature extraction result based on the node relationship graph, and the long short-term memory network is used to output the corresponding health status detection result based on the feature extraction result.

[0054] In this embodiment, the detection model can include various network structures. For example, the detection model can include graph convolutional networks and long short-term memory networks. The graph convolutional network and the long short-term memory network can perform different processing on the node relationship graph to achieve the detection of the electrolytic cell's health status. The server uses the graph convolutional network in the detection model to obtain the corresponding feature extraction results based on the node relationship graph. The server uses the long short-term memory network in the detection model to output the corresponding health status detection results based on the feature extraction results.

[0055] Through this embodiment, the server can combine the various networks in the detection model to process the node relationship graph in different dimensions, thereby improving the accuracy of health status detection of the electrolytic cell.

[0056] In one embodiment, obtaining the corresponding feature extraction result based on the above node relationship graph includes: extracting the features of each node in the above node relationship graph; performing neighborhood aggregation and feature propagation on each of the above node features to obtain the spatial feature matrix corresponding to the above node relationship graph, which is used as the above feature extraction result.

[0057] In this embodiment, the server extracts features from the node relationship graph using a graph convolutional network in the detection model, obtaining the feature extraction result. Specifically, the server can extract the features of each node in the node relationship graph using the graph convolutional network, and perform neighborhood aggregation and feature propagation on each of the node features to obtain the spatial feature matrix corresponding to the node relationship graph, which serves as the feature extraction result.

[0058] Specifically, the server can perform graph convolutional feature extraction. Based on the constructed degradation mechanism graph (node ​​relationship graph), the server uses a graph convolutional network to perform neighborhood aggregation and feature propagation on node features, extracting spatially coupled features of multiple degradation mechanisms.

[0059] In the degradation mechanism graph G=(V,E,A), V is the set of nodes with a size of |V|=N, E is the set of edges representing the connections between nodes, and A∈R. N×N The adjacency matrix is ​​given by graph G. Furthermore, the input to graph G is X.tf =[X f ,X t In the serial structure, the GCN operator acts independently at each time step in the sequence. For time t, the graph convolutional network outputs the spatial feature matrix Z. t ∈R N×d' Z t k =σ(A'H t k-1 W gcn )=GCN(X t W gcn (Where, A'=D') -1 / 2 (A+I N ), D' -1 / 2 Let D be the normalized Laplace matrix, where the corresponding degree matrix is ​​D. ii =∑ j A' ij W gcn ∈R d×d' Let H be the weight matrix of the learnable GCN, where d and d' are the input and output feature dimensions, and σ(*) is the sigmoid activation function. t k-1 The node features after the (k-1)th iteration: H t k-1 =σ(A'H t k-2 W gcn ), where H t 0 =X t .

[0060] In this embodiment, the server extracts features from the node relationship graph based on a graph convolutional network, thereby using the extracted features to detect the health status of the electrolytic cell, thus improving the accuracy of the electrolytic cell health status detection.

[0061] In one embodiment, outputting the corresponding health status detection result based on the above feature extraction result includes: predicting the long-term degradation trend of the above feature extraction result and outputting the health status detection result corresponding to the above electrolytic cell.

[0062] In this embodiment, the server can analyze the features extracted by the graph convolutional network through the long short-term memory network in the detection model, and analyze the dynamic process of the electrolytic cell's health status degradation mechanism over time. For example, the server uses the long short-term memory network to predict the long-term degradation trend of the above feature extraction results and outputs the corresponding health status detection results of the electrolytic cell.

[0063] Specifically, the server utilizes a long short-term memory network for temporal degradation modeling. For example... Figure 3As shown, Figure 3 This is a schematic diagram of the model detection steps in one embodiment. The server outputs the temporal features Z=[Z1,Z2,…,Z…] from the GCN. N Input a long short-term memory network, establish a dynamic model of the degradation mechanism over time, and obtain hidden states and memory states that reflect long-term degradation trends.

[0064] To preserve the independence of each node, the Long Short-Term Memory (LSTM) network processes N nodes in parallel. Let h... t Let c be the hidden state at time t. t This represents the cell state. For input Z... t Its input gate i t Forgotten Gate f t Output gate o t and candidate cell state c' t The calculation satisfies the following formula: {i t =σ(W zi Z t +W hi h t-1 +b i );f t =σ(W zf Z t +W hf h t-1 +b f );o t =σ(W zo Z t +W ho h t-1 +b o );c' t =tanh(W zc Z t +W hc h t-1 +b c )}. Among them, W zα W zα and b α (α=i,f,o,t) represent the cyclic weight matrix of the input gate, forget gate, output gate, and candidate cell state, respectively, along with the hidden state weight matrix and bias term. The cell state c at time t is... t and the hidden state h t The state is updated by the following formula: {c t =f t ⊙c t-1 +i t ⊙c' t h t =o t ⊙tanh(c t The above formula can be simplified to: (c t ,ht =LSTM(Z) t ,c t-1 ,h t-1 ,θ LSTM ), where θ LSTM This includes gating and fully connected layer matrix weights and biases.

[0065] For the training process of the above detection model, since only historical data is available, the training of GCN-LSTM is essentially a combination of single-step prediction training and multi-step prediction inference. The loss function L for single-step prediction training is:

[0066] L= SOH' k+1 -SOH k+1 2 Among them, SOH' k+1 This represents the model's predicted values, such as predicted health status; SOH k+1 This represents the actual label value, such as the actual health status.

[0067] Based on the loss function above, the server uses the following gradient descent backdiffusion method to sequentially update the weights and biases of the prediction mapping layer, gated units, and graph convolutional network: W n+1 (l) =W n (l) -η∂L / (∂W n (l) The weight matrix for the (n+1)th iteration of the l-th layer is W. n+1 (l) .

[0068] The server combines the output of the Long Short-Term Memory (LSTM) network to predict the health status of the electrolytic cell. Specifically, based on the final memory state of the LTM network, the server outputs electrolytic cell health status indicators for the current and future prediction time domains through a fully connected prediction layer. The model parameters are updated online or periodically using newly acquired data to achieve adaptive optimization of the health status assessment model.

[0069] Specifically, assuming the prediction step size is H, the predicted electrolytic cell health status SOH' at any time in the time interval [t, t+H] is recursively calculated using an autoregressive method. t+i This can be expressed as: {h t+i =LSTM(h t+i-1 Z t+i );SOH' t+i =W o h t+i-1 +b0;i=1~H}. Where W o ∈ Rd'×dh and b o ∈R d'It predicts the weights and biases of the mapping layer.

[0070] Through this embodiment, the server can perform degradation and evolution analysis on the features output by the graph convolutional network based on the long short-term memory network to obtain the health status of the electrolytic cell, thereby improving the accuracy of the health status detection of the electrolytic cell.

[0071] In one exemplary embodiment, such as Figure 4 As shown, Figure 4 This is a flowchart illustrating a method for detecting the health status of an electrolytic cell in another embodiment. This embodiment includes the following steps:

[0072] (1) Decouple the cumulative effects of operating conditions and membrane water content on voltage and define them as the health status of the electrolyzer; (2) Extract the time-frequency features of key aging characteristics of the electrolyzer, where key lumped parameters such as oxygen content and high-frequency impedance are used as time-domain features, while frequency-domain features are derived from the characteristic impedance and frequency in the battery relaxation time distribution; (3) Abstract the aforementioned key aging features into graph nodes and construct a graph convolution-long short-term memory network health status assessment framework based on the endogenous correlation of aging features; (4) Design a fully connected layer to match the memory state output of the long short-term memory network and predict the subsequent health status of the battery. Among them, the graph convolutional network is used to characterize the degradation coupling mechanism, while the long short-term memory network can quantify the long-term degradation trend, thus enabling high-precision and highly interpretable online health status assessment of the electrolyzer.

[0073] Specifically, the server first defines the health state of the electrolyzer. Since apparent parameters such as voltage only reflect the battery's current operating conditions (e.g., current, pressure, oxygen permeate ratio, and temperature), internal cumulative parameters, such as the load-carrying capacity under membrane water content, are less relevant. To decouple the influence of irrelevant factors and define the true battery health state, the server can define the electrolyzer health state through operating condition calculations and internal cumulative parameter normalization strategies. Specifically, from the Nernst equation and polarization loss, the electrolyzer voltage equation satisfies:

[0074] V cell =E rev +RT / 2Fp H2 (p O2 ) 1 / 2 +η act +η ohm +η con Among them, E rev Represents the reversible voltage, R is the gas constant, T is the PEMEC temperature, F represents the Faraday constant, and p i This represents the gas pressure i. And η represents the gas pressure. act ,η ohm and η con Representing activation, ohmic, and concentration loss respectively, their specific expressions are as follows: {η act =RT / (αa a F)ln(I / I a,ref )+(RT / α c c F)ln(I / I c,ref );η ohm =R ohm I;η con =RT / 2F(ln(I lim / (I lim -I)))}。 Where, α a a and α c c I, representing the positive coefficients of the anodic and cathodic reactions. a,ref I c,ref and I lim These represent the battery current density, anode and cathode reference current densities, and limiting current densities, respectively. Furthermore, the conductivity of the Nafion proton exchange membrane is related to the modal water λ as follows: σ = exp[1268(1 / 303-1 / (273+T))](0.005139λ-0.00326); the ohmic resistance includes membrane resistance and other resistances (R). oth Examples include contact resistance and terminal resistance: R ohm =σ / S MEA +R oth Therefore, the equation for the electrolytic cell voltage can be rewritten as: V cell =f(I,p,T,λ). Wherein, the initial parameter set Ω of the above equation... bol (Initial electrical parameters) are affected by degradation effects and change with battery operation; their real-time parameter set Ω t (Real-time electrical parameters) can be obtained by inverse solution of the above equation: Ω t =f -1 (I t ,p t ,T t ,λ t ).

[0075] Based on operating condition conversion and modal water normalization, the server defines the standard condition health status index of the electrolyzer as SOH. ref The corresponding standard case is [I] ref ,p ref ,T ref ,λ ref So, what is the real-time health status (SOH) of the electrolytic cell? t The standard voltage ratio corresponding to the current parameter set and the initial parameter set can be defined: SOH t =(f(I ref ,p ref ,T ref ,λref )|Ω=Ω t ) / (f(I ref ,p ref ,T ref ,λ ref )|Ω=Ω bol The initial electrical parameters represent the internal characteristic parameters of the electrolyzer when it leaves the factory or is in a brand-new state, while the real-time electrical parameters characterize the current internal characteristic parameters of the battery after running for time t. Since Ω cannot be directly measured... t (For example, membrane water content), the server can measure the current voltage V. t and operating conditions (I) t ,p t ,T t ,λ t By performing inverse computation (or parameter identification) on model f, the current internal parameter set Ω is calculated. t The aforementioned real-time health status can be a dimensionless ratio between 0 and 1, where a value greater than 1 indicates an abnormal performance improvement, such as activation, and a value less than 1 indicates degradation.

[0076] The server acquires time-domain signals such as high-frequency impedance and oxygen flow rate under the operating conditions of the electrolytic cell training set through multi-source operational data acquisition and preprocessing. t =[X t1 ,X t2 [ ] and frequency domain electrochemical impedance spectroscopy R(f) and other operating data.

[0077] The server can identify polarization processes with different time constants in complex electrochemical systems by analyzing relaxation time distributions. For impedance spectrum R(f) data, complete DRT analysis can yield j sets of characteristic DRT peak resistances and corresponding frequencies [(R...]. DRT1 ,f DRT1 ),(R DRT2 ,f DRT2 ),…(R DRTj ,f DRTj Since the degradation process affects not only the magnitude of the characteristic impedance but also its response constant, the server introduces position coding to perform impedance-frequency fusion on the DRT features to obtain the final frequency domain signal X. f :

[0078] X f = =[X f1 ,X f2 ,…,X fj ]. Wherein, ψ and d model Both represent normalized parameters, log 10 This is a logarithmic operation.

[0079] The server performs anomaly removal and max-min normalization on the above time-frequency signal to obtain x. norm The server's anomaly removal and normalization processing of the above signals can be expressed as: x norm =(xx min ) / (x max -x min ), where x norm x max and x min These represent the maximum and minimum values ​​of the variable group after normalization and before normalization, respectively.

[0080] The server can construct a degradation mechanism graph (node ​​relationship graph) by combining the above signals. Specifically, the server abstracts key variables characterizing different degradation mechanisms of the electrolyzer as graph nodes, and the influence relationships between different degradation mechanisms as graph edges. The server can select six graph nodes: nodes 1, 2, and 3 represent the DRT and the mass transfer characteristics, oxygen reduction reaction, and proton transfer effect of the cathode catalyst layer obtained through impedance-frequency fusion processing, respectively. Nodes 4 and 5 represent the high-frequency impedance and oxygen flow time-domain signals acquired by the sensors, respectively. Node 6 represents the real-time health status (SOH). t Based on the coupling relationship of the above six nodes, the server predefines the graph and edges of the GCN and performs relationship analysis based on standard operating conditions.

[0081] Among them, Node 1: Mass transfer characteristics characterize the porosity and equivalent transfer coefficient of the gas diffusion layer, thus it is also coupled with Node 4 (oxygen flow rate). Node 2: Oxygen Reduction Reaction (ORR) characteristics characterize the health of the electrochemical reaction at the three-phase interface of the catalyst, ion conductor (ionomer), and electronic conductor (carbon support). ORR requires proton participation, and in PEMEC, proton transfer depends on sufficient modal water; therefore, it is also coupled with Node 3 (cathode proton transfer) and Node 5 (high-frequency impedance). Node 3: Cathode proton transfer characteristics characterize the health of proton transfer between the membrane components and the catalyst layer. In addition to the aforementioned coupling relationships, the hydrogen ion transfer rate affects the reaction rate, thus affecting the oxygen flow rate at Node 4. Node 4: Oxygen flow rate characterizes the gas diffusion layer transfer and the health of the electrochemical reaction. Besides Nodes 1, 2, and 3, there are no other nodes with strong coupling effects. Node 5: High-frequency impedance characterizes the health of the membrane components. Besides Nodes 2 and 3, there are no other nodes with strong coupling effects. Therefore, the server constructs a node relationship graph by combining the above nodes and the relationships between them.

[0082] The server can perform graph convolutional feature extraction. Specifically, based on the constructed degradation mechanism graph (node ​​relationship graph), the server uses a graph convolutional network to perform neighborhood aggregation and feature propagation on node features, extracting spatially coupled features of multiple degradation mechanisms.

[0083] In the degradation mechanism graph G=(V,E,A), V is the set of nodes with a size of |V|=N, E is the set of edges representing the connections between nodes, and A∈R. N×N The adjacency matrix is ​​given by graph G. Furthermore, the input to graph G is X. tf =[X f ,X t In the serial structure, the GCN operator acts independently at each time step in the sequence. For time t, the graph convolutional network outputs the spatial feature matrix Z. t ∈R N×d' Z t k =σ(A'H t k-1 W gcn )=GCN(X t W gcn (Where, A'=D') -1 / 2 (A+I N ), D' -1 / 2 Let D be the normalized Laplace matrix, where the corresponding degree matrix is ​​D. ii =∑ j A' ij W gcn ∈R d×d' Let H be the weight matrix of the learnable GCN, where d and d' are the input and output feature dimensions, and σ(*) is the sigmoid activation function. t k-1 The node features after the (k-1)th iteration: H t k-1 =σ(A'H t k-2 W gcn ), where H t 0 =X t .

[0084] The server utilizes a Long Short-Term Memory (LSM) network for temporal degradation modeling. The server outputs the temporal features Z=[Z1,Z2,…,Z…] from the GCN. N Input a long short-term memory network, establish a dynamic model of the degradation mechanism over time, and obtain hidden states and memory states that reflect long-term degradation trends.

[0085] To preserve the independence of each node, the Long Short-Term Memory (LSTM) network processes N nodes in parallel. Let h... t Let c be the hidden state at time t.t This represents the cell state. For input Z... t Its input gate i t Forgotten Gate f t Output gate o t and candidate cell state c' t The calculation satisfies the following formula: {i t =σ(W zi Z t +W hi h t-1 +b i );f t =σ(W zf Z t +W hf h t-1 +b f );o t =σ(W zo Z t +W ho h t-1 +b o );c' t =tanh(W zc Z t +W hc h t-1 +b c )}. Among them, W zα W zα and b α (α=i,f,o,t) represent the cyclic weight matrix of the input gate, forget gate, output gate, and candidate cell state, respectively, along with the hidden state weight matrix and bias term. The cell state c at time t is... t and the hidden state h t The state is updated by the following formula: {c t =f t ⊙c t-1 +i t ⊙c' t h t =o t ⊙tanh(c t The above formula can be simplified to: (c t ,h t =LSTM(Z) t ,c t-1 ,h t-1 ,θ LSTM ), where θ LSTM This includes gating and fully connected layer matrix weights and biases.

[0086] For the training process of the above detection model, since only historical data is available, the training of GCN-LSTM is essentially a combination of single-step prediction training and multi-step prediction inference. The loss function L for single-step prediction training is:

[0087] L= SOH' k+1 -SOH k+1 2 Among them, SOH' k+1 This represents the model's predicted values, such as predicted health status; SOH k+1 This represents the actual label value, such as the actual health status.

[0088] Based on the loss function above, the server uses the following gradient descent backdiffusion method to sequentially update the weights and biases of the prediction mapping layer, gated units, and graph convolutional network: W n+1 (l) =W n (l) -η∂L / (∂W n (l) The weight matrix for the (n+1)th iteration of the l-th layer is W. n+1 (l) .

[0089] The server combines the output of the Long Short-Term Memory (LSTM) network to predict the health status of the electrolytic cell. Specifically, based on the final memory state of the LTM network, the server outputs electrolytic cell health status indicators for the current and future prediction time domains through a fully connected prediction layer. The model parameters are updated online or periodically using newly acquired data to achieve adaptive optimization of the health status assessment model.

[0090] Specifically, assuming the prediction step size is H, the predicted electrolytic cell health status SOH' at any time in the time interval [t, t+H] is recursively calculated using an autoregressive method. t+i This can be expressed as: {h t+i =LSTM(h t+i-1 Z t+i );SOH' t+i =W o h t+i-1 +b0;i=1~H}. Where W o ∈ Rd'×dh and b o ∈R d' It predicts the weights and biases of the mapping layer.

[0091] Through the above embodiments, a node relationship graph is constructed based on time-domain signals such as high-frequency impedance and oxygen flow rate, frequency-domain signals such as impedance spectrum, and the real-time health status of the electrolytic cell under test. A detection model then detects the health status of the electrolytic cell based on this node relationship graph and outputs the corresponding health status detection result. Compared to traditional methods that rely on equivalent circuit models for health status detection, this application improves the accuracy of electrolytic cell health status detection by combining the time-domain signals, frequency-domain signals, and health status nodes of the electrolytic cell to construct a node relationship graph. The detection model then uses this node relationship graph to detect the health status of the electrolytic cell, thus enhancing the accuracy of health status detection.

[0092] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0093] Based on the same inventive concept, this application also provides an electrolytic cell health status detection device for implementing the electrolytic cell health status detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the electrolytic cell health status detection device provided below can be found in the limitations of the electrolytic cell health status detection method described above, and will not be repeated here.

[0094] In one exemplary embodiment, such as Figure 5 As shown, an electrolytic cell health status detection device is provided, comprising: a first acquisition module 500, a second acquisition module 502, and a detection module 504, wherein:

[0095] The first acquisition module 500 is used to acquire the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be detected; the time-domain signal includes the high-frequency impedance and oxygen flow rate during the operation of the electrolytic cell, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell.

[0096] The second acquisition module 502 is used to acquire the node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell.

[0097] The detection module 504 is used to input the above node relationship diagram into the detection model; the above detection model is used to output the health status detection result corresponding to the above electrolytic cell based on the above node relationship diagram.

[0098] In one embodiment, the second acquisition module 502 is configured to generate each first node based on the high-frequency impedance in the time-domain signal, the oxygen flow rate, and the impedance spectrum in the frequency-domain signal; generate a second node corresponding to the real-time health status of the electrolytic cell based on the initial electrical parameters and real-time electrical parameters corresponding to the electrolytic cell; generate each edge based on the relationship between each first node and the second node; and generate the node relationship graph based on each first node, the second node, and each edge.

[0099] In one embodiment, the second acquisition module 502 is used to generate corresponding mass transfer characteristic nodes, oxygen reduction reaction nodes, and cathode catalyst layer proton transfer effect nodes according to the impedance spectrum and relaxation time distribution algorithm. The mass transfer characteristic nodes characterize the health status of the gas diffusion layer porosity and equivalent transfer coefficient of the electrolytic cell. The oxygen reduction reaction nodes characterize the electrochemical reaction health status of the catalyst, ionic conductor, and electronic conductor. The cathode catalyst layer proton transfer effect nodes characterize the proton transfer health status of the membrane components and the catalyst layer. Each of the first nodes is obtained based on the nodes corresponding to the oxygen flow rate, the nodes corresponding to the high-frequency impedance, the mass transfer characteristic nodes, the oxygen reduction reaction nodes, and the cathode catalyst layer proton transfer effect nodes.

[0100] In one embodiment, the detection module 504 is used to extract the features of each node in the node relationship graph; perform neighborhood aggregation and feature propagation on each node feature to obtain the spatial feature matrix corresponding to the node relationship graph, which is used as the feature extraction result.

[0101] In one embodiment, the detection module 504 is used to predict the long-term degradation trend of the feature extraction results and output the health status detection results corresponding to the electrolytic cell.

[0102] Each module in the aforementioned electrolytic cell health status detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0103] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores electrolytic cell operating data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for detecting the health status of an electrolytic cell.

[0104] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0105] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described electrolytic cell health status detection method.

[0106] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the electrolytic cell health status detection method described above.

[0107] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described electrolytic cell health status detection method.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0111] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting the health status of an electrolytic cell, characterized in that, The method includes: Acquire the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be tested; the time-domain signal includes the high-frequency impedance and oxygen flow rate of the electrolytic cell during operation, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell; Obtain the node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell. The node relationship diagram is input into the detection model; the detection model is used to output the health status detection result of the electrolytic cell based on the node relationship diagram.

2. The method according to claim 1, characterized in that, The step of obtaining the node relationship diagram corresponding to the health status of the electrolytic cell includes: Each first node is generated based on the high-frequency impedance in the time-domain signal, the oxygen flow rate, and the impedance spectrum in the frequency-domain signal; Based on the initial and real-time electrical parameters of the electrolytic cell, a second node corresponding to the real-time health status of the electrolytic cell is generated. Based on the relationship between each of the first nodes and the second nodes, generate each edge; The node relationship graph is generated based on each of the first node, the second node, and each edge.

3. The method according to claim 2, characterized in that, The step of generating each first node based on the high-frequency impedance in the time-domain signal, the oxygen flow rate, and the impedance spectrum in the frequency-domain signal includes: Based on the impedance spectrum and relaxation time distribution algorithm, corresponding mass transfer characteristic nodes, oxygen reduction reaction nodes, and cathode catalyst layer proton transfer effect nodes are generated. The mass transfer characteristic nodes characterize the porosity of the gas diffusion layer and the health status of the equivalent transfer coefficient of the electrolytic cell. The oxygen reduction reaction nodes characterize the electrochemical reaction health status of the catalyst, ion conductor, and electronic conductor. The cathode catalyst layer proton transfer effect nodes characterize the proton transfer health status of the membrane components and the catalyst layer. Each of the first nodes is obtained based on the node corresponding to the oxygen flow rate, the node corresponding to the high-frequency impedance, the mass transfer characteristic node, the oxygen reduction reaction node, and the cathode catalyst layer proton transfer effect node.

4. The method according to claim 1, characterized in that, The detection model includes a graph convolutional network and a long short-term memory network; the graph convolutional network is used to obtain the corresponding feature extraction results based on the node relationship graph, and the long short-term memory network is used to output the corresponding health status detection results based on the feature extraction results.

5. The method according to claim 4, characterized in that, The step of obtaining the corresponding feature extraction results based on the node relationship graph includes: Extract the features of each node in the node relationship graph; Neighborhood aggregation and feature propagation are performed on the features of each node to obtain the spatial feature matrix corresponding to the node relationship graph, which is used as the feature extraction result.

6. The method according to claim 4, characterized in that, The step of outputting the corresponding health status detection result based on the feature extraction result includes: The long-term degradation trend of the extracted features is predicted, and the health status detection result of the electrolytic cell is output.

7. A device for detecting the health status of an electrolytic cell, characterized in that, The device includes: The first acquisition module is used to acquire the time-domain signal and frequency-domain signal corresponding to the electrolytic cell to be detected; the time-domain signal includes the high-frequency impedance and oxygen flow rate of the electrolytic cell during operation, and the frequency-domain signal includes the impedance spectrum of the electrolytic cell. The second acquisition module is used to acquire a node relationship diagram corresponding to the health status of the electrolytic cell; each node in the node relationship diagram is determined based on the high-frequency impedance and oxygen flow rate in the time domain signal, the impedance spectrum in the frequency domain signal, and the real-time health status node of the electrolytic cell. The detection module is used to input the node relationship diagram into the detection model; the detection model is used to output the health status detection result of the electrolytic cell according to the node relationship diagram.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images