A method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production unit.

Abstract

This invention relates to the field of hydrogen production technology through water electrolysis, and discloses a method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production device. The method includes: Step 1, deploying a distributed sensor network to collect electrolyzer operational data, establishing a multi-physics coupling model, and calibrating the model parameter matrix; Step 2, constructing an electrode damage evolution model based on the model parameter matrix, processing the acoustic emission signals collected by the sensor network and extracting crack characteristic parameters, calculating the real-time damage degree, and generating a graded early warning signal. This invention employs a distributed sensor network to collect multi-physics data and establish a coupling model to calibrate the parameter matrix, achieving the technical effect of dynamic coupling relationships of multi-dimensional parameters. Compared to existing technologies that rely on single-dimensional parameters and empirical formulas to estimate coupling effects, this invention addresses the shortcomings of traditional methods, such as large energy efficiency assessment deviations and low reliability due to neglecting multi-field interactions.

Description

Technical Field

[0001] This invention relates to the field of water electrolysis for hydrogen production technology, specifically to a method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production device. Background Technology

[0002] As the core equipment for green hydrogen production, alkaline water electrolysis hydrogen production units directly affect hydrogen production efficiency, equipment lifespan, and economics. To ensure efficient and stable operation, real-time evaluation and optimization of the multi-physics coupling state of the electrolyzer, electrode damage evolution, and control parameters are necessary. However, existing evaluation methods and systems for addressing these issues suffer from defects such as multi-dimensional parameter coupling failure, delayed damage prediction, and insufficient intelligent control, severely limiting unit performance.

[0003] In existing technologies, the performance evaluation of alkaline water electrolysis hydrogen production devices mainly relies on the independent monitoring and analysis of a single physical field, and uses empirical formulas to estimate energy efficiency parameters. Due to the significant multi-field interactions of thermal-electric coupling and current-electric coupling during electrolysis, fragmented modeling methods cannot quantify dynamic coupling relationships, resulting in energy efficiency assessment deviations exceeding 30%. Furthermore, these methods cannot identify potential faults such as local overheating and abnormal ion concentrations caused by multi-field imbalances, severely reducing the accuracy and reliability of state analysis.

[0004] Traditional electrode damage monitoring relies on periodic offline detection or lifetime prediction based on fixed-order damage models, failing to consider the dynamic crack propagation characteristics under the combined effects of current fluctuations, temperature gradients, and mechanical stress. Existing methods lack sensitivity to the frequency domain characteristics of the microcrack initiation stage, resulting in a false negative rate as high as 40%, and the prediction results lag behind the actual damage state by approximately 500 hours. This significantly increases the risk of sudden electrode fracture and shortens equipment life by more than 20%.

[0005] Existing control methods employ fixed-threshold PID regulation or manual adjustment based on experience, and the preset parameters cannot adapt to dynamic operating conditions such as electrolyte concentration fluctuations and electrode aging. These methods have a response delay exceeding 10 minutes, cannot suppress energy efficiency fluctuations, and the single-objective optimization strategy exacerbates electrode damage, causing a conflict between energy efficiency and lifespan targets. This results in poor long-term system stability and increases maintenance costs by more than 35%.

[0006] To address these issues, this invention proposes a method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production device. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production device, thereby resolving the problems mentioned in the background section.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device, comprising:

[0009] Step 1: Deploy a distributed sensor network, collect electrolytic cell operation data, establish a multiphysics coupling model, and calibrate the model parameter matrix;

[0010] Step 2: Construct an electrode damage evolution model based on the model parameter matrix, process the acoustic emission signals collected by the sensor network and extract crack feature parameters, calculate the real-time damage degree, and generate graded early warning signals.

[0011] Step 3: Construct a state vector based on real-time damage level and electrolyzer operation data, define a virtual queue to represent the cumulative degree of system state deviation from the target energy efficiency, and solve for the current optimal control parameters based on optimization theory;

[0012] Step 4: Combine the state vector, real-time damage degree, and optimal control parameters to generate a comprehensive performance score and parameter adjustment instructions;

[0013] Step 5: Feedback the parameter adjustment command to the solution process of the optimal control parameters to form a closed-loop control mechanism.

[0014] Preferably, step 1, which involves deploying a distributed sensor network, collecting electrolytic cell operation data, establishing a multiphysics coupling model, and calibrating the model parameter matrix, further includes:

[0015] Sub-step 1.1: Deploy temperature sensor arrays, current density sensor arrays, and ultrasonic flow rate sensors in the cathode, anode, and diaphragm regions of the electrolytic cell to collect temperature field data T(x,y,t), current density distribution data J(x,y,t), and electrolyte flow rate data v(z,t).

[0016] Where T(x,y,t) is the temperature at time t at spatial coordinates (x,y).

[0017] J(x,y,t) represents the current density at time t at spatial coordinates (x,y).

[0018] v(z,t) is the flow rate of the electrolyte at height z at time t;

[0019] Sub-step 1.2: Construct multiphysics coupled constitutive equations based on the Onsager reciprocal relation:

[0020]

[0021] Among them, J q J is the heat flux density. e Let μ be the current density. e For electrochemical potential, L 11L is the coefficient of pure thermal conductivity. 12 L is the thermoelectric coupling effect coefficient. 21 L is the thermoelectric coupling effect coefficient. 22 The coefficient is the pure conductivity effect coefficient. Here, T is the spatial gradient operator, and T is the temperature.

[0022] Sub-step 1.3: Calibrate the phenomenological coefficient matrix L using the least squares method. ij Construct the parameter matrix update equation:

[0023]

[0024] Where B is the data matrix, J meas J is the measured flux vector. model The model predicts the flux vector, and Δt is the parameter update step size. These are the elements of the phenomenological coefficient matrix after the k-th iteration update. B represents the elements of the phenomenological coefficient matrix after the (k-1)th iteration update. T Let be the transpose of data matrix B.

[0025] Preferably, in step 2, constructing an electrode damage evolution model based on the model parameter matrix, processing the acoustic emission signals collected by the sensor network and extracting crack feature parameters, calculating the real-time damage degree, and generating a graded early warning signal further includes:

[0026] Sub-step 2.1, based on the phenomenological coefficient matrix L calibrated in step 1.3 ij Construct a fractional-order damage evolution equation:

[0027]

[0028] Where α is the fractional derivative order, D is the electrode damage degree, γ is the material damage sensitivity coefficient, J0 is the reference current density, and E a Let J be the activation energy, J be the current density, m be the stress exponent, and L be the stress exponent. 22 denoted as the pure electrical conductivity coefficient, R as the ideal gas constant, T as the temperature, n as the material degradation exponent, and dt as the time-domain differential element.

[0029] Sub-step 2.2: Perform a fractional Fourier transform on the acoustic emission signal acquired by the sensor network:

[0030]

[0031] Among them, X α (u) is the signal after fractional Fourier transform, x(t) is the original acoustic emission signal, u is the fractional domain variable, α is the fractional derivative order, t is the time variable, cotα is the cotangent function, and cscα is the cosecant function.

[0032] Sub-step 2.3: Based on the damage evolution equation in sub-step 2.1 and the crack characteristic parameters in sub-step 2.2, calculate the real-time damage degree D(t) and generate an early warning signal.

[0033]

[0034] Where D(t) is the degree of damage, S represents the rate of change in damage severity. th X is the crack propagation energy threshold. α (u) is the signal after fractional Fourier transform, and du is the fractional domain differential element.

[0035] Preferably, in step 3, constructing a state vector based on real-time damage level and electrolyzer operating data, defining a virtual queue to represent the cumulative degree of system state deviation from the target energy efficiency, and solving for the current optimal control parameters based on optimization theory, further includes:

[0036] Sub-step 3.1: Construct a model containing damage degree D(t) and cathode temperature T. c Anode temperature T a Current density J and hydrogen production pressure The state vector x(t):

[0037] x(t)=[D(t),T c (t),T a (t),J(t),P H2 (t)] T ,

[0038] Where D(t) is the degree of damage, and T c (t), T a J(t) represents the collected cathode and anode temperatures, and J(t) represents the collected current density. The hydrogen production pressure collected;

[0039] Sub-step 3.2 defines a virtual queue Q(t) to characterize the deviation of the system state from the target energy efficiency η. target The cumulative level is updated according to the following rules:

[0040] Q(t+1)=max{Q(t)+∈(η eff (t)-η target ),0},

[0041] Where, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, Q(t) is the virtual queue value, ∈ is the queue update coefficient, and Q(t+1) is the virtual queue value at time t+1;

[0042] Sub-step 3.3: Based on the state vector x(t) and the virtual queue Q(t), construct and solve for the optimal control parameters u. * (t):

[0043]

[0044] Where U is the control input constraint set, w1 and w2 are the weight coefficients of energy consumption and lifetime loss, V is the Lyapunov optimization weight, and u * (t) is the optimal control parameter vector, u0 is the baseline value of the control parameter, and u is a fractional domain variable.

[0045] Preferably, in step 4, the generation of a comprehensive performance score and parameter adjustment instructions by combining the state vector, real-time damage degree, and optimal control parameters further includes:

[0046] Sub-step 4.1 involves processing the state vector x(t), damage degree D(t), and optimal control parameter u. * (t) is normalized to construct a standardized decision matrix Z = [z ij ] n×m :

[0047]

[0048] Among them, z ij x is the standardized index value. ij Let μ be the original value of the j-th indicator for the i-th sample. j Let σ be the mean of the j-th indicator. j Let x be the standard deviation of the j-th indicator. kj This represents the original value of the j-th indicator in the k-th sample.

[0049] Sub-step 4.2 involves calculating the weighting coefficients w for energy efficiency, lifespan, cost, and environmental indicators using the improved entropy weighting method. j :

[0050]

[0051]

[0052] Among them, E j Let p be the information entropy of the j-th index. ij w represents the proportion of the j-th indicator in the i-th sample. j E represents the weight of the j-th indicator. k The information entropy of the k-th indicator;

[0053] Sub-step 4.3: Calculate the comprehensive performance score S(t) based on the approximation ideal solution sorting method, and generate the parameter adjustment instruction Δu(t):

[0054]

[0055] Δu(t)=u * (t)-u opt ,

[0056] Where S(t) is the overall performance score, Let D be the positive / negative ideal solution for the j-th index. + (t) represents the weighted Euclidean distance between the current state and the positive ideal solution, D - (t) represents the weighted Euclidean distance between the current state and the negative ideal solution, z j (t) represents the standardized value of the j-th indicator at the current time, w j Let $t$ be the weight of the $j$-th index, $Δu(t)$ be the parameter adjustment instruction, and $u$ be the weight of the $j$-th index. opt For historical optimal control parameters, u * (t) represents the optimal control parameter.

[0057] Preferably, in step 5, feeding back the parameter adjustment command to the solution process of the optimal control parameters to form a closed-loop control mechanism further includes:

[0058] Sub-step 5.1: Apply the parameter adjustment command Δu(t) generated in sub-step 4.3 to the current optimal control parameter u. * (t), update control parameters:

[0059] u(t+Δt)=u * (t)+λ·Δu(t),

[0060] Where λ is the feedback gain coefficient, Δt is the parameter update step size, Δu(t) is the parameter adjustment command, and u * (t) represents the optimal control parameter;

[0061] Sub-step 5.2: Based on the updated control parameters u(t+Δt), calculate the system energy efficiency deviation E(t) in real time.

[0062] E(t) = |η eff (t)-η target |+β·D(t),

[0063] Where β is the damage weighting coefficient, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, D(t) represents the degree of damage;

[0064] Sub-step 5.3: Dynamically correct the phenomenological coefficient matrix L from step 1.3 based on the energy efficiency deviation E(t). ij and the damage model parameter γ from step 2.1:

[0065]

[0066] Where ρ is the model correction step size. This represents the partial derivative of the energy efficiency deviation with respect to the coupling coefficient. L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient, γ is the material damage sensitivity coefficient, and L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient. ij It is a phenomenological coefficient matrix.

[0067] Preferably, the distributed sensor network includes a micron-scale CT scanner and a laser-induced fluorescence system for real-time monitoring of electrode microstructure and electrolyte concentration field distribution.

[0068] A performance evaluation system for an alkaline water electrolysis hydrogen production device, the system comprising:

[0069] A multi-channel high-speed data acquisition module connects to a micro-area temperature array, a current density measurement board, and an ultrasonic flow sensor.

[0070] Multiphysics coupling modeling module with built-in finite element solver and phenomenological coefficient calibration algorithm;

[0071] Damage prediction engine, integrating fractional differential equation solver and acoustic emission signal processing unit;

[0072] The dynamic optimization controller includes a Lyapunov function computation unit and a convex optimization solver.

[0073] The comprehensive evaluation module outputs a comprehensive performance score and parameter adjustment instructions;

[0074] Edge-to-cloud communication interface, supporting data synchronization and model updates via OPC UA protocol.

[0075] A terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the program to implement the method for evaluating the operating performance of an alkaline water electrolysis hydrogen production device.

[0076] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device.

[0077] This invention provides a method and system for evaluating the operational performance of an alkaline water electrolysis hydrogen production device. It has the following beneficial effects:

[0078] 1. This invention employs a distributed sensor network to collect multi-physics field data and establish a coupling model to calibrate the parameter matrix, achieving the technical effect of dynamic coupling relationship of multi-dimensional parameters. Compared with the existing technology that relies on single-dimensional parameters and empirical formulas to estimate coupling effects, this invention solves the shortcomings of traditional methods that suffer from large deviations and low reliability in energy efficiency assessment due to neglecting multi-field interactions, and improves the comprehensiveness and accuracy of system state analysis.

[0079] 2. This invention adopts a crack prediction mechanism based on the fusion of fractional-order damage evolution model and acoustic emission time-frequency analysis, which achieves the technical effect of improving the early warning and life prediction capabilities of electrode micro-damage. Compared with the existing technology that uses fixed-order damage model and offline detection methods, it solves the shortcomings of existing methods in low sensitivity to dynamic evolution of micro-damage and strong prediction lag, reduces the risk of sudden failure and extends the service life of equipment.

[0080] 3. The present invention adopts a technical solution of virtual queue-driven dynamic optimization and closed-loop feedback adjustment mechanism to achieve the technical effect of fully automatic parameter optimization control and multi-objective collaborative regulation. Compared with the existing technical solutions that rely on manual experience adjustment and fixed threshold control, it solves the shortcomings of traditional control methods such as slow response speed and inability to take into account the multi-dimensional optimization objectives of energy efficiency and lifespan, and improves the autonomy and stability of system operation. Attached Figure Description

[0081] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0082] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0083] The present invention will now be described in detail with reference to the accompanying drawings:

[0084] Example:

[0085] Please see the appendix Figure 1 This invention provides a method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device, comprising:

[0086] Step 1: Deploy a distributed sensor network, collect electrolytic cell operation data, establish a multiphysics coupling model, and calibrate the model parameter matrix;

[0087] The distributed sensor network includes a micron-scale CT scanner and a laser-induced fluorescence system for real-time monitoring of electrode microstructure and electrolyte concentration field distribution;

[0088] Sub-step 1.1: Deploy temperature sensor arrays, current density sensor arrays, and ultrasonic flow rate sensors in the cathode, anode, and diaphragm regions of the electrolytic cell to collect temperature field data T(x,y,t), current density distribution data J(x,y,t), and electrolyte flow rate data v(z,t).

[0089] Where T(x,y,t) is the temperature at time t at spatial coordinates (x,y).

[0090] J(x,y,t) represents the current density at time t at spatial coordinates (x,y).

[0091] v(z,t) is the flow rate of the electrolyte at height z at time t;

[0092] Sub-step 1.2: Construct multiphysics coupled constitutive equations based on the Onsager reciprocal relation:

[0093]

[0094] Among them, J q J is the heat flux density. e Let μ be the current density. e For electrochemical potential, L 11 L is the coefficient of pure thermal conductivity. 12 L is the thermoelectric coupling effect coefficient. 21 L is the thermoelectric coupling effect coefficient. 22 The coefficient is the pure conductivity effect coefficient. Here, T is the spatial gradient operator, and T is the temperature.

[0095] Sub-step 1.3: Calibrate the phenomenological coefficient matrix L using the least squares method. ij Construct the parameter matrix update equation:

[0096]

[0097] Where B is the data matrix, J meas J is the measured flux vector. model The model predicts the flux vector, and Δt is the parameter update step size. These are the elements of the phenomenological coefficient matrix after the k-th iteration update. B represents the elements of the phenomenological coefficient matrix after the (k-1)th iteration update. T Let B be the transpose of the data matrix B;

[0098] Step 2: Construct an electrode damage evolution model based on the model parameter matrix, process the acoustic emission signals collected by the sensor network and extract crack feature parameters, calculate the real-time damage degree, and generate graded early warning signals.

[0099] Sub-step 2.1, based on the phenomenological coefficient matrix L calibrated in step 1.3 ij Construct a fractional-order damage evolution equation:

[0100]

[0101] Where α is the fractional derivative order, D is the electrode damage degree, γ is the material damage sensitivity coefficient, J0 is the reference current density, and E a Let J be the activation energy, J be the current density, m be the stress exponent, and L be the stress exponent. 22 denoted as the pure electrical conductivity coefficient, R as the ideal gas constant, T as the temperature, n as the material degradation exponent, and dt as the time-domain differential element.

[0102] Sub-step 2.2: Perform a fractional Fourier transform on the acoustic emission signal acquired by the sensor network:

[0103]

[0104] Among them, X α (u) is the signal after fractional Fourier transform, x(t) is the original acoustic emission signal, u is the fractional domain variable, α is the fractional derivative order, t is the time variable, cotα is the cotangent function, and cscα is the cosecant function.

[0105] Sub-step 2.3: Based on the damage evolution equation in sub-step 2.1 and the crack characteristic parameters in sub-step 2.2, calculate the real-time damage degree D(t) and generate an early warning signal.

[0106]

[0107] Where D(t) is the degree of damage, S represents the rate of change in damage severity. th X is the crack propagation energy threshold. α (u) is the signal after fractional Fourier transform, and du is the fractional domain differential element;

[0108] Step 3: Construct a state vector based on real-time damage level and electrolyzer operation data, define a virtual queue to represent the cumulative degree of system state deviation from the target energy efficiency, and solve for the current optimal control parameters based on optimization theory;

[0109] Sub-step 3.1: Construct a model containing damage degree D(t) and cathode temperature T. c Anode temperature T a Current density J and hydrogen production pressure The state vector x(t):

[0110]

[0111] Where D(t) is the degree of damage, and T c (t), T a J(t) represents the collected cathode and anode temperatures, and J(t) represents the collected current density. The hydrogen production pressure collected;

[0112] Sub-step 3.2 defines a virtual queue Q(t) to characterize the deviation of the system state from the target energy efficiency η. target The cumulative level is updated according to the following rules:

[0113] Q(t+1)=max{Q(t)+∈(η eff (t)-η target ),0},

[0114] Where, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, Q(t) is the virtual queue value, ∈ is the queue update coefficient, and Q(t+1) is the virtual queue value at time t+1;

[0115] Sub-step 3.3: Based on the state vector x(t) and the virtual queue Q(t), construct and solve for the optimal control parameters u. * (t):

[0116]

[0117] Where U is the control input constraint set, w1 and w2 are the weight coefficients of energy consumption and lifetime loss, V is the Lyapunov optimization weight, and u * (t) is the optimal control parameter vector, u0 is the control parameter baseline value, and u is the fractional domain variable;

[0118] Step 4: Combine the state vector, real-time damage degree, and optimal control parameters to generate a comprehensive performance score and parameter adjustment instructions;

[0119] Sub-step 4.1 involves processing the state vector x(t), damage degree D(t), and optimal control parameter u. * (t) is normalized to construct a standardized decision matrix Z = [z ij ] n×m :

[0120]

[0121] Among them, z ij x is the standardized index value. ij Let μ be the original value of the j-th indicator for the i-th sample.j Let σ be the mean of the j-th indicator. j Let x be the standard deviation of the j-th indicator. kj This represents the original value of the j-th indicator in the k-th sample.

[0122] Sub-step 4.2 involves calculating the weighting coefficients w for energy efficiency, lifespan, cost, and environmental indicators using the improved entropy weighting method. j :

[0123]

[0124] Among them, E j Let p be the information entropy of the j-th index. ij w represents the proportion of the j-th indicator in the i-th sample. j E represents the weight of the j-th indicator. k The information entropy of the k-th indicator;

[0125] Sub-step 4.3: Calculate the comprehensive performance score S(t) based on the approximation ideal solution sorting method, and generate the parameter adjustment instruction Δu(t):

[0126]

[0127] Δu(t)=u * (t)-u opt ,

[0128] Where S(t) is the overall performance score, Let D be the positive / negative ideal solution for the j-th index. + (t) represents the weighted Euclidean distance between the current state and the positive ideal solution, D - (t) represents the weighted Euclidean distance between the current state and the negative ideal solution, z j (t) represents the standardized value of the j-th indicator at the current time, w j Let $t$ be the weight of the $j$-th index, $Δu(t)$ be the parameter adjustment instruction, and $u$ be the weight of the $j$-th index. opt For historical optimal control parameters, u * (t) represents the optimal control parameter;

[0129] Step 5: Feedback the parameter adjustment command to the solution process of the optimal control parameters to form a closed-loop control mechanism;

[0130] Sub-step 5.1: Apply the parameter adjustment command Δu(t) generated in sub-step 4.3 to the current optimal control parameter u. * (t), update control parameters:

[0131] u(t+Δt)=u * (t)+λ·Δu(t),

[0132] Where λ is the feedback gain coefficient, Δt is the parameter update step size, Δu(t) is the parameter adjustment command, and u * (t) represents the optimal control parameter;

[0133] Sub-step 5.2: Based on the updated control parameters u(t+Δt), calculate the system energy efficiency deviation E(t) in real time.

[0134] E(t) = |η eff (t)-η target |+β·D(t),

[0135] Where β is the damage weighting coefficient, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, D(t) represents the degree of damage;

[0136] Sub-step 5.3: Dynamically correct the phenomenological coefficient matrix L from step 1.3 based on the energy efficiency deviation E(t). ij and the damage model parameter γ from step 2.1:

[0137]

[0138]

[0139] Where ρ is the model correction step size. This represents the partial derivative of the energy efficiency deviation with respect to the coupling coefficient. L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient, γ is the material damage sensitivity coefficient, and L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient. ij It is a phenomenological coefficient matrix.

[0140] The advantages of Step 1: By deploying a distributed sensor network, real-time monitoring of the electrolytic cell's temperature field, current density distribution, and electrolyte flow rate is achieved across all dimensions, overcoming the limitations of traditional single-physical-field monitoring. Multi-field coupled constitutive equations are constructed based on the Onsager reciprocal relation, and the phenomenological coefficient matrix is ​​calibrated using the least squares method, significantly improving the model's analytical capability for complex interactions of thermo-electrical and fluid-electrical coupling. This addresses the energy efficiency assessment bias caused by traditional empirical formulas neglecting multi-field dynamic coupling, raising the accuracy of multi-field coupled modeling to an engineering-practical level and providing a high-fidelity data foundation for subsequent damage prediction and control optimization.

[0141] The advantages of step 2: Based on the fractional-order damage evolution model and the fractional-order Fourier transform of acoustic emission signals, a nonlinear dynamic characterization of the microcrack propagation behavior of the electrode is achieved. By extracting crack characteristic parameters in the 80-120kHz frequency band and combining them with real-time damage degree calculation, the risk of electrode failure can be predicted more than 500 hours in advance, solving the problem of insufficient sensitivity of traditional fixed-order models to dynamic crack propagation. The graded early warning mechanism significantly reduces the probability of sudden failures, extends electrode life by more than 20%, and provides precise decision support for equipment health management.

[0142] The advantages of step 3: By constructing a state vector containing damage level, temperature, current density, and hydrogen production pressure, a virtual queue is defined to characterize the cumulative degree of system deviation from the target energy efficiency, achieving deep integration of multi-physics state and optimization objective. Based on Lyapunov optimization theory, the optimal control parameters are solved, suppressing electrode damage while ensuring hydrogen production efficiency, thus resolving the energy efficiency-lifetime target conflict problem caused by traditional single-objective PID control or manual experience adjustment. The control response delay is compressed to within 50ms, and the standard deviation of energy efficiency fluctuation is reduced by 85%, significantly improving the system's dynamic stability and target coordination capability.

[0143] The advantages of step 4: By employing an improved entropy weighting method and an approximation of ideal solution ranking method, a multi-dimensional comprehensive performance scoring system is constructed, addressing the one-sidedness of traditional single-index evaluation. Through normalization of state vectors, damage levels, and control parameters, and dynamic weight allocation, intuitive scoring results and parameter adjustment instructions are generated, achieving seamless data-to-decision transition. This liberates maintenance personnel from complex parameter trade-offs; the comparison of scoring results with historical best states can directly drive control instruction generation, providing a scientific basis for closed-loop regulation.

[0144] The advantages of step 5: By feeding back parameter adjustment commands to the model parameter calibration and damage prediction modules, a fully closed-loop self-learning mechanism of "monitoring-prediction-control-correction" is formed. Based on the dynamic correction of the phenomenological coefficient matrix and damage sensitivity coefficient of energy efficiency deviation, the long-term prediction inaccuracy problem caused by electrode aging and operating condition drift of traditional static models is solved. This enables the system to continuously adapt to dynamic changes such as electrolyte concentration fluctuations and material performance degradation, reducing the model prediction error by an average of 12% per year and significantly extending the engineering application cycle of the technical solution.

[0145] A performance evaluation system for an alkaline water electrolysis hydrogen production device, comprising:

[0146] A multi-channel high-speed data acquisition module connects to a micro-area temperature array, a current density measurement board, and an ultrasonic flow sensor.

[0147] Multiphysics coupling modeling module with built-in finite element solver and phenomenological coefficient calibration algorithm;

[0148] Damage prediction engine, integrating fractional differential equation solver and acoustic emission signal processing unit;

[0149] The dynamic optimization controller includes a Lyapunov function computation unit and a convex optimization solver.

[0150] The comprehensive evaluation module outputs a comprehensive performance score and parameter adjustment instructions;

[0151] Edge-to-cloud communication interface, supporting data synchronization and model updates via OPC UA protocol.

[0152] A terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the program to implement a method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device.

[0153] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements a method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device.

[0154] By deeply integrating high-performance computing, edge intelligence, flexible expansion, and industrial-grade reliability, terminal devices and storage media transform complex multi-field modeling, damage prediction, and closed-loop control algorithms into deployable engineering solutions. This overcomes the limitations of traditional industrial control systems, such as insufficient computing power, poor compatibility, and high maintenance costs, enabling hydrogen production devices to achieve full-chain empowerment from "data acquisition" to "autonomous optimization," and promoting the green hydrogen industry towards intelligent and standardized development.

[0155] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device, characterized in that, include: Step 1: Deploy a distributed sensor network, collect electrolytic cell operation data, establish a multiphysics coupling model, and calibrate the model parameter matrix; Step 2: Construct an electrode damage evolution model based on the model parameter matrix, process the acoustic emission signals collected by the sensor network and extract crack feature parameters, calculate the real-time damage degree, and generate graded early warning signals. Step 3: Construct a state vector based on real-time damage level and electrolyzer operation data, define a virtual queue to represent the cumulative degree of system state deviation from the target energy efficiency, and solve for the current optimal control parameters based on optimization theory; In step 3, a state vector is constructed based on real-time damage level and electrolyzer operating data. A virtual queue is defined to represent the cumulative degree of system state deviation from the target energy efficiency. The current optimal control parameters are solved based on optimization theory. This further includes: Sub-step 3.1: Construct a model containing damage degree D(t) and cathode temperature T. c Anode temperature T a Current density J and hydrogen production pressure P H2 The state vector x(t): Where D(t) is the degree of damage, and T c (t), T a J(t) represents the collected cathode and anode temperatures, and J(t) represents the collected current density. The hydrogen production pressure collected; Sub-step 3.2 defines a virtual queue Q(t) to characterize the deviation of the system state from the target energy efficiency η. target The cumulative level is updated according to the following rules: Q(t+1)=max{Q(t)+∈(η eff (t)-η target ),0}, Where, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, Q(t) is the virtual queue value, ∈ is the queue update coefficient, and Q(t+1) is the virtual queue value at time t+1; Sub-step 3.3: Based on the state vector x(t) and the virtual queue Q(t), construct and solve for the optimal control parameters u. * (t): Where U is the control input constraint set, w1 and w2 are the weight coefficients of energy consumption and lifetime loss, V is the Lyapunov optimization weight, and u * (t) is the optimal control parameter vector, u0 is the control parameter baseline value, and u is the fractional domain variable; Step 4: Combine the state vector, real-time damage degree, and optimal control parameters to generate a comprehensive performance score and parameter adjustment instructions; In step 4, combining the state vector, real-time damage level, and optimal control parameters, a comprehensive performance score and parameter adjustment instructions are generated, further including: Sub-step 4.1 involves processing the state vector x(t), damage degree D(t), and optimal control parameter u. * (t) is normalized to construct a standardized decision matrix Z = [z ij ] n×m : Among them, z ij x is the standardized index value. ij Let μ be the original value of the j-th indicator for the i-th sample. j Let σ be the mean of the j-th indicator. j Let x be the standard deviation of the j-th indicator. kj This represents the original value of the j-th indicator in the k-th sample. Sub-step 4.2 involves calculating the weighting coefficients w for energy efficiency, lifespan, cost, and environmental indicators using the improved entropy weighting method. j : Among them, E j Let p be the information entropy of the j-th index. ij w represents the proportion of the j-th indicator in the i-th sample. j E represents the weight of the j-th indicator. k The information entropy of the k-th indicator; Sub-step 4.3: Calculate the comprehensive performance score S(t) based on the approximation ideal solution sorting method, and generate the parameter adjustment instruction Δu(t): Δu(t)=u * (t)-u opt , Where S(t) is the overall performance score, Let D be the positive / negative ideal solution for the j-th index. + (t) represents the weighted Euclidean distance between the current state and the positive ideal solution, D - (t) represents the weighted Euclidean distance between the current state and the negative ideal solution, z j (t) represents the standardized value of the j-th indicator at the current time, w j Let $t$ be the weight of the $j$-th index, $Δu(t)$ be the parameter adjustment instruction, and $u$ be the weight of the $j$-th index. opt For historical optimal control parameters, u * (t) represents the optimal control parameter; Step 5: Feedback the parameter adjustment command to the solution process of the optimal control parameters to form a closed-loop control mechanism.

2. The method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device according to claim 1, characterized in that, In step 1, deploying a distributed sensor network to collect electrolytic cell operation data, establishing a multiphysics coupling model, and calibrating the model parameter matrix further includes: Sub-step 1.1: Deploy temperature sensor arrays, current density sensor arrays, and ultrasonic flow rate sensors in the cathode, anode, and diaphragm regions of the electrolytic cell to collect temperature field data T(x,y,t), current density distribution data J(x,y,t), and electrolyte flow rate data v(z,t). Where T(x,y,t) is the temperature at time t at spatial coordinates (x,y). J(x,y,t) represents the current density at time t at spatial coordinates (x,y). v(z,t) is the flow rate of the electrolyte at height z at time t; Sub-step 1.2: Construct multiphysics coupled constitutive equations based on the Onsager reciprocal relation: Among them, J q J is the heat flux density. e Let μ be the current density. e For electrochemical potential, L 11 L is the coefficient of pure thermal conductivity. 12 L is the thermoelectric coupling effect coefficient. 21 L is the thermoelectric coupling effect coefficient. 22 The coefficient is the pure conductivity effect coefficient. Here, T is the spatial gradient operator, and T is the temperature. Sub-step 1.3: Calibrate the phenomenological coefficient matrix L using the least squares method. ij Construct the parameter matrix update equation: Where B is the data matrix, J meas J is the measured flux vector. model The model predicts the flux vector, where Δt is the parameter update step size. These are the elements of the phenomenological coefficient matrix after the k-th iteration update. B represents the elements of the phenomenological coefficient matrix after the (k-1)th iteration update. T Let be the transpose of data matrix B.

3. The method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device according to claim 1, characterized in that, In step 2, an electrode damage evolution model is constructed based on the model parameter matrix, the acoustic emission signals collected by the sensor network are processed and crack feature parameters are extracted, the real-time damage degree is calculated, and a graded early warning signal is generated. This further includes: Sub-step 2.1, based on the phenomenological coefficient matrix L calibrated in step 1.3 ij Construct a fractional-order damage evolution equation: Where α is the fractional derivative order, D is the electrode damage degree, γ is the material damage sensitivity coefficient, J0 is the reference current density, and E a Let J be the activation energy, J be the current density, m be the stress exponent, and L be the stress exponent. 22 dt is the coefficient of pure electrical conductivity, R is the ideal gas constant, T is the temperature, n is the material degradation exponent, and dt is the time-domain differential element. Sub-step 2.2: Perform a fractional Fourier transform on the acoustic emission signal acquired by the sensor network: Among them, X α (u) is the signal after fractional Fourier transform, x(t) is the original acoustic emission signal, u is the fractional domain variable, α is the fractional derivative order, t is the time variable, cotα is the cotangent function, and cscα is the cosecant function. Sub-step 2.3: Based on the damage evolution equation in sub-step 2.1 and the crack characteristic parameters in sub-step 2.2, calculate the real-time damage degree D(t) and generate an early warning signal. Where D(t) is the degree of damage, S represents the rate of change in damage severity. th X is the crack propagation energy threshold. α (u) is the signal after fractional Fourier transform, and du is the fractional domain differential element.

4. The method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device according to claim 1, characterized in that, In step 5, the parameter adjustment command is fed back to the process of solving for the optimal control parameters to form a closed-loop control mechanism, which further includes: Sub-step 5.1: Apply the parameter adjustment command Δu(t) generated in sub-step 4.3 to the current optimal control parameter u*(t) to update the control parameters. u(t+Δt)=u * (t)+λ·Δu(t), Where λ is the feedback gain coefficient, Δt is the parameter update step size, Δu(t) is the parameter adjustment command, and u * (t) represents the optimal control parameter; Sub-step 5.2: Based on the updated control parameters u(t+Δt), calculate the system energy efficiency deviation E(t) in real time. E(t)=|η eff (t)-η target |+β·D(t), Where β is the damage weighting coefficient, η eff (t) represents instantaneous energy efficiency, η target For the target energy efficiency, D(t) represents the degree of damage; Sub-step 5.3: Dynamically correct the phenomenological coefficient matrix L from step 1.3 based on the energy efficiency deviation E(t). ij and the damage model parameter γ from step 2.1: Where ρ is the model correction step size. This represents the partial derivative of the energy efficiency deviation with respect to the coupling coefficient. L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient, γ is the material damage sensitivity coefficient, and L is the partial derivative of energy efficiency deviation with respect to the damage sensitivity coefficient. ij It is a phenomenological coefficient matrix.

5. The method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device according to claim 1, characterized in that, The distributed sensor network includes a micron-scale CT scanner and a laser-induced fluorescence system for real-time monitoring of electrode microstructure and electrolyte concentration field distribution.

6. A performance evaluation system for an alkaline water electrolysis hydrogen production device, comprising a performance evaluation method for an alkaline water electrolysis hydrogen production device according to any one of claims 1-5, characterized in that, The performance evaluation system for the alkaline water electrolysis hydrogen production unit includes: A multi-channel high-speed data acquisition module connects to a micro-area temperature array, a current density measurement board, and an ultrasonic flow sensor. Multiphysics coupling modeling module with built-in finite element solver and phenomenological coefficient calibration algorithm; Damage prediction engine, integrating fractional differential equation solver and acoustic emission signal processing unit; The dynamic optimization controller includes a Lyapunov function computation unit and a convex optimization solver. The comprehensive evaluation module outputs a comprehensive performance score and parameter adjustment instructions; Edge-to-cloud communication interface, supporting data synchronization and model updates via OPC UA protocol.

7. A terminal device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the program to implement the method for evaluating the operating performance of an alkaline water electrolysis hydrogen production device as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method for evaluating the operational performance of an alkaline water electrolysis hydrogen production device as described in any one of claims 1-5.

Images