Anion exchange membrane performance prediction and life evaluation method, device and medium

Abstract

The invention relates to the technical field of ion exchange membranes, in particular to an anion exchange membrane performance prediction and life evaluation method and device and a medium, and the anion exchange membrane performance prediction and life evaluation method comprises the following steps: obtaining a multi-source data set in the anion exchange membrane operation process in real time; performing space-time alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with a timestamp mark; constructing a hybrid prediction model fusing deep learning and a physical mechanism; and inputting the dynamic characteristic matrix into the trained hybrid prediction model, and outputting a real-time performance degradation index and a residual life confidence interval of the anion exchange membrane. According to the technical scheme, through the hybrid prediction model fusing deep learning and physical mechanisms, the deep learning model can capture complex laws of performance changes of the anion exchange membrane, and the physical constraint module fuses priori knowledge into the model, so that the generalization ability of the model is enhanced, and the interpretability of the model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of ion exchange membranes, and particularly to a method, device and medium for predicting the performance and evaluating the lifespan of an anion exchange membrane. Background Art

[0002] Anion exchange membranes are widely used in energy fields such as fuel cells and water electrolysis. The stability of their performance and service life are crucial for the operation of related equipment. However, during actual operation, anion exchange membranes are affected by various factors and undergo performance degradation, such as attacks by free radicals generated by electrochemical reactions and erosion by hydrated hydroxide ions, resulting in a decrease in the mechanical properties, chemical stability, and ion conduction performance of the membrane, thereby affecting the efficiency and safety of the entire energy conversion or storage system.

[0003] Traditional methods for predicting the performance and evaluating the lifespan of anion exchange membranes mainly rely on single data-driven models or pure physical models. Although data-driven models can process a large amount of experimental data, they often lack an understanding of the internal physical and chemical mechanisms of anion exchange membranes and are difficult to accurately capture their complex degradation processes. Especially when faced with new operating conditions or data changes, the generalization ability of the prediction is poor. Pure physical models, on the other hand, require accurate parameter estimation and in-depth understanding of the mechanisms. In practical applications, due to system complexity and parameter uncertainty, it is difficult to establish an accurate physical model, and the computational cost is high, making it difficult to provide performance degradation and lifespan evaluation results in real time and quickly. Summary of the Invention

[0004] To solve the problems in the related technologies, embodiments of the present disclosure provide a method, device and medium for predicting the performance and evaluating the lifespan of an anion exchange membrane.

[0005] In a first aspect, embodiments of the present disclosure provide a method for predicting the performance and evaluating the lifespan of an anion exchange membrane, including the following steps: Obtain a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; Perform spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with timestamp markings; Construct a hybrid prediction model that combines deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding, and the two modules perform joint output through residual connection; Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

[0006] According to an embodiment of the present disclosure, the electrochemistry data includes a voltage fluctuation sequence, a current density distribution, and an electrochemical impedance spectrum, the environmental monitoring data includes a temperature gradient value, a relative humidity, and its change rate, and the material characterization data includes an ion conductivity decay curve and a dynamic swelling rate change value.

[0007] According to an embodiment of the present disclosure, the generating a dynamic feature matrix with timestamp marks by performing spatio-temporal alignment and feature fusion on the multi-source data set includes: According to a unified timestamp benchmark, perform sliding window statistics on the voltage fluctuation sequence and the current density distribution, extract the root mean square error of the voltage and the peak-to-valley difference of the current as time domain features, and at the same time perform equivalent circuit model fitting on the electrochemical impedance spectrum to extract the frequency domain degradation features of the charge transfer resistance and the double-layer capacitance; Perform cubic spline interpolation alignment on the temperature gradient value, the relative humidity, and its change rate, and the ion conductivity decay curve according to the timestamp benchmark to eliminate the timing offset caused by the sampling interval difference; Perform adaptive normalization processing on the dynamic swelling rate change value based on the real-time collected relative humidity to generate a normalized swelling rate feature vector; Splice the processed time domain features, frequency domain degradation features, temperature gradient value, relative humidity and / or change rate, ion conductivity decay curve, and normalized swelling rate feature vector into a dynamic feature matrix according to the timestamp benchmark.

[0008] According to an embodiment of the present disclosure, the time series prediction module is implemented by using one or more of a long short-term memory network, a gated recurrent unit, or a convolutional neural network; the physical constraint module simulates the change process of the ion exchange rate and the free radical concentration in the anion exchange membrane by establishing a degradation kinetic equation set, where the degradation kinetic equation set includes at least one or more of a diffusion equation based on Fick's law, a Nernst-Planck equation, or a Butler-Volmer equation.

[0009] According to an embodiment of the present disclosure, the training process of the hybrid prediction model includes: The time series prediction module takes the dynamic feature matrix as input and outputs a preliminary prediction value of the membrane performance degradation trend; The physical constraint module inputs the temperature, humidity, and free radical concentration data in the dynamic feature matrix and calculates the theoretical performance degradation index; Subtract the preliminary prediction value of the time series prediction module from the theoretical degradation index output by the physical constraint module through a residual connection to obtain a residual correction term, and use the corrected performance degradation index as the final output; Using the true value of the performance degradation index in historical data as a supervision signal, jointly optimize the weight parameters of the time series prediction module and the learnable reaction rate constant in the physical constraint module by the gradient descent method.

[0010] According to an embodiment of the present disclosure, the inputting the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane includes: Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index of the anion exchange membrane; Based on the time series fitting result of the real-time performance degradation index, calculate the remaining life confidence interval in combination with the failure probability model.

[0011] According to an embodiment of the present disclosure, the method further includes: Trigger a maintenance alarm when the remaining life is lower than a preset threshold.

[0012] In a second aspect, an anion exchange membrane performance prediction and life evaluation device is provided in an embodiment of the present disclosure, including: An acquisition module configured to acquire a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; A generation module configured to perform spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with time stamp marks; A construction module configured to construct a hybrid prediction model that integrates deep learning and physical mechanisms, where the hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts the non-linear degradation pattern in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding, and the two modules perform joint output through a residual connection; A prediction module configured to input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

[0013] In a third aspect, an electronic device is provided in an embodiment of the present disclosure, including a memory and a processor. Among them, the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the method according to any item of the first aspect.

[0014] Fourthly, an embodiment of the present disclosure provides a computer-readable storage medium, on which computer instructions are stored. When the computer instructions are executed by a processor, the method described in any item of the first aspect is implemented.

[0015] The technical effects provided by the embodiments of the present disclosure may include the following beneficial effects: According to the technical solution provided by the embodiment of the present disclosure, a method for predicting the performance and evaluating the life of an anion exchange membrane includes the following steps: obtaining a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; performing spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with timestamp marks; constructing a hybrid prediction model that combines deep learning and physical mechanisms, where the hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding, and the two modules are jointly output through a residual connection; inputting the dynamic feature matrix into the trained hybrid prediction model, and outputting the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane. In the above technical solution, through the hybrid prediction model that combines deep learning and physical mechanisms, the advantages of both are fully utilized. The deep learning model can automatically extract non-linear degradation patterns from the dynamic feature matrix and capture the complex laws of the performance change of the anion exchange membrane; while the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration based on the degradation kinetic equation and integrates its prior knowledge into the model, which not only helps to improve the adaptability of the model to different working conditions and data changes, enhances the generalization ability of the model, but also increases the interpretability of the model.

[0016] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart showing a method for predicting the performance and evaluating the life of an anion exchange membrane according to an embodiment of the present disclosure.

[0018] Figure 2 A structural block diagram showing an apparatus for predicting the performance and evaluating the life of an anion exchange membrane according to an embodiment of the present disclosure.

[0019] Figure 3 A structural block diagram showing an electronic device according to an embodiment of the present disclosure.

[0020] Figure 4 A schematic structural diagram showing a computer system suitable for implementing the method according to an embodiment of the present disclosure. Detailed implementation manners

[0021] In the following, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily implement them. In addition, for clarity, parts unrelated to the description of the exemplary embodiments are omitted in the drawings.

[0022] In the present disclosure, it should be understood that terms such as "including" or "having" are intended to indicate the presence of features, numbers, steps, actions, components, parts, or combinations thereof disclosed in this specification, and are not intended to exclude the possibility of the presence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0023] In addition, it should be noted that, without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] Traditional methods for predicting the performance and evaluating the lifespan of anion exchange membranes mainly rely on a single data-driven model or a pure physical model. Although data-driven models can process a large amount of experimental data, they often lack an understanding of the intrinsic physical and chemical mechanisms of anion exchange membranes and are difficult to accurately capture their complex degradation processes. Especially when facing new working conditions or data changes, the generalization ability of the prediction is poor. On the other hand, pure physical models require accurate parameter estimation and in-depth understanding of the mechanisms. In practical applications, due to system complexity and parameter uncertainty, it is difficult to establish an accurate physical model, and the computational cost is high, making it difficult to provide performance degradation and lifespan evaluation results in real time and quickly.

[0025] Considering the above defects, the anion exchange membrane performance prediction and lifespan evaluation method provided by the embodiments of the present disclosure makes full use of the advantages of both through a hybrid prediction model that combines deep learning and physical mechanisms. The deep learning model can automatically extract non-linear degradation patterns from the dynamic feature matrix and capture the complex laws of the performance changes of anion exchange membranes; while the physical constraint module conducts mechanism modeling on the ion exchange rate and free radical concentration based on the degradation kinetic equation and integrates its prior knowledge into the model, which not only helps improve the adaptability of the model to different working conditions and data changes, enhances the generalization ability of the model, but also increases the interpretability of the model.

[0026] Figure 1 A flowchart showing the anion exchange membrane performance prediction and lifespan evaluation method according to an embodiment of the present disclosure.

[0027] As Figure 1 shown, the anion exchange membrane performance prediction and lifespan evaluation method includes the following steps: S110: Obtain a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; S120: Align the multi-source data set spatiotemporally and fuse the features to generate a dynamic feature matrix with timestamp marks; S130: Construct a hybrid prediction model that combines deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts the non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module conducts mechanism modeling on the ion exchange rate and radical concentration through differential equation embedding, and the two modules are jointly output through residual connection; S140: Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

[0028] In the embodiments of the present disclosure, a multi-source data set during the operation of the anion exchange membrane is obtained in real time through a variety of sensors and detection instruments installed around the anion exchange membrane device. Among them, the electrochemical data covers voltage fluctuation sequences, current density distributions, and electrochemical impedance spectra, which can reflect the performance changes of the membrane during the electrochemical reaction process; the environmental monitoring data includes temperature gradient values, relative humidity, and its change rate, which are used to reflect the influence of the operating environment on the membrane; the material characterization data includes the ion conductivity decay curve and the dynamic change value of the swelling rate, which can reflect the degradation trend of the membrane from the perspective of material properties.

[0029] Specifically, take the operation of the anion exchange membrane (AEM) in a fuel cell as an example. First, obtain a multi-source data set during the operation of the AEM in real time. For example, through high-precision voltage sensors and current sensors, with a sampling interval of 0.1 second, continuously collect the voltage fluctuation sequence and current density distribution data of the AEM during operation; at the same time, use an electrochemical impedance spectrometer to collect electrochemical impedance spectrum data every 10 minutes under different operating conditions. The environmental monitoring data includes temperature gradient values, relative humidity, and its change rate. By arranging multiple temperature sensors and humidity sensors around the AEM, the temperature change and humidity change in its operating environment are monitored in real time, and the sampling frequency is 1 time per second. The material characterization data includes the ion conductivity decay curve and the dynamic change value of the swelling rate. With the help of professional material testing equipment, the ion conductivity and swelling rate of the AEM are regularly tested at different stages of its operation to obtain the corresponding decay curve and dynamic change value.

[0030] In the embodiments of the present disclosure, the collected multi-source data sets are subjected to spatio-temporal alignment and feature fusion operations to generate a dynamic feature matrix with timestamp markings, so that the subsequent model can accurately perform performance prediction and life assessment based on time series.

[0031] Specifically, based on a unified timestamp benchmark (such as the running time accurate to seconds), sliding window statistics are performed on the voltage fluctuation sequence and the current density distribution. The window size is set to 10 sampling points, and the step size is 1 sampling point. The root mean square error of voltage and the peak-to-valley difference of current are extracted as time-domain features. At the same time, the electrochemical impedance spectrum is fitted with an equivalent circuit model. The commonly used Randles equivalent circuit model is adopted, and through a non-linear fitting algorithm, the frequency-domain degradation features of the charge transfer resistance and the double-layer capacitance are extracted. The temperature gradient value, relative humidity, and its change rate are aligned with the ionic conductivity decay curve by cubic spline interpolation according to the timestamp benchmark, eliminating the timing offset caused by the sampling interval differences of different sensors, so that each data sequence is consistent in the time dimension. And an adaptive normalization process based on the real-time collected relative humidity is performed on the dynamic change value of the swelling rate to generate a normalized swelling rate feature vector. Finally, the processed time-domain features, frequency-domain degradation features, temperature gradient value, relative humidity and / or change rate, ionic conductivity decay curve, and normalized swelling rate feature vector are spliced into a dynamic feature matrix according to the timestamp benchmark. Each row of this matrix represents the comprehensive feature vector of the AEM at a certain moment, containing key information from different data sources.

[0032] Specifically, the dynamic feature matrix can be expressed as:

[0033] where n represents the timestamp; represents the processed time-domain features; represents the frequency-domain degradation features; represents the temperature gradient value; represents the relative humidity; represents the relative humidity change rate; represents the ionic conductivity decay curve; represents the normalized swelling rate feature vector.

[0034] In the embodiments of the present disclosure, a hybrid prediction model integrating deep learning and physical mechanisms is constructed. The model consists of a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. The time series prediction module utilizes its powerful non-linear fitting ability to extract non-linear degradation patterns in the dynamic feature matrix and capture the complex laws of membrane performance changing over time. The physical constraint module, based on the method of embedding differential equations, conducts mechanism modeling on the ion exchange rate and free radical concentration, describes the degradation process of the membrane from the physical essence, and the two modules are jointly output through residual connection, realizing the deep integration of data-driven and physical mechanisms, and improving the accuracy and reliability of prediction.

[0035] Specifically, the time series prediction module can be implemented using a long short-term memory network (LSTM). The dynamic feature matrix is input into the LSTM network, and the network learns the sequence data through its internal gating structures (input gate, forget gate, and output gate) to extract the non-linear degradation patterns therein. The physical constraint module simulates the change process of the ion exchange rate and free radical concentration in the AEM by establishing a degradation kinetic equation set. For example, the diffusion process of ions in the AEM is described based on Fick's law diffusion equation, the influence of the electric field on ion migration is considered using the Nernst - Planck equation, and the Butler - Volmer equation is combined to characterize the kinetic behavior in the electrode reaction process. The two modules are jointly output through residual connection, that is, the theoretical performance degradation index output by the physical constraint module is subtracted from the preliminary prediction value output by the time series prediction module to obtain a residual correction term, and the corrected performance degradation index is used as the final output.

[0036] In the embodiments of the present disclosure, the processed dynamic feature matrix is input into the trained hybrid prediction model. After complex operations and analyses, the model outputs the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane, providing a scientific basis for the maintenance and replacement of the anion exchange membrane and ensuring the stable operation of related equipment.

[0037] For example, assuming that the AEM fails when the performance degradation index reaches a certain threshold, based on the failure probability distribution obtained from the current degradation trend and historical data statistics, methods such as Monte Carlo simulation are used to calculate the remaining life confidence interval of the AEM at different future time points, providing a decision-making basis for the maintenance and replacement of the AEM. When the remaining life is lower than a preset threshold (such as the remaining life is less than 100 hours), a maintenance alarm is triggered to remind relevant personnel to check and maintain the AEM in a timely manner to ensure the normal operation of the system.

[0038] According to an embodiment of the present disclosure, in step S120, performing spatio-temporal alignment and feature fusion on the multi-source dataset to generate a dynamic feature matrix with timestamp marks includes: According to a unified timestamp benchmark, perform sliding window statistics on the voltage fluctuation sequence and the current density distribution, extract the root mean square error of voltage and the peak-valley difference of current as time-domain features, and at the same time perform equivalent circuit model fitting on the electrochemical impedance spectrum to extract the frequency-domain degradation features of the charge transfer resistance and the double-layer capacitance; Perform cubic spline interpolation alignment on the temperature gradient value, relative humidity and its change rate, and the ion conductivity decay curve according to the timestamp benchmark to eliminate the timing offset caused by the sampling interval difference; Perform adaptive normalization processing on the dynamic change value of the swelling rate based on the relative humidity collected in real time to generate a normalized swelling rate feature vector; Splice the processed time-domain features, frequency-domain degradation features, temperature gradient value, relative humidity and / or change rate, ion conductivity decay curve, and normalized swelling rate feature vector into a dynamic feature matrix according to the timestamp benchmark.

[0039] In the embodiment of the present disclosure, first, according to a unified timestamp benchmark, perform sliding window statistics on the voltage fluctuation sequence and the current density distribution, and extract the root mean square error of voltage and the peak-valley difference of current as time-domain features. These time-domain features can reflect the change trends and fluctuations of voltage and current over time; at the same time, perform equivalent circuit model fitting on the electrochemical impedance spectrum to extract the frequency-domain degradation features of the charge transfer resistance and the double-layer capacitance. The frequency-domain degradation features can reflect the characteristic changes of the electrochemical impedance spectrum at different frequencies, further enriching the feature information in the electrochemical aspect. Then, perform cubic spline interpolation alignment on the temperature gradient value, relative humidity and its change rate, and the ion conductivity decay curve according to the timestamp benchmark to effectively eliminate the timing offset caused by the sampling interval difference and ensure the consistency and comparability of each data in the time dimension. Then, perform adaptive normalization processing on the dynamic change value of the swelling rate based on the relative humidity collected in real time to generate a normalized swelling rate feature vector, making the swelling rate data comparable and unified under different humidity conditions. Finally, splice the processed time-domain features, frequency-domain degradation features, temperature gradient value, relative humidity and / or change rate, ion conductivity decay curve, and normalized swelling rate feature vector into a dynamic feature matrix according to the timestamp benchmark to form a complete, systematic and orderly data set, providing high-quality input data for the subsequent hybrid prediction model and helping to improve the accuracy and reliability of model prediction.

[0040] According to an embodiment of the present disclosure, in step S130, the time series prediction module is implemented by using one or more of a long short-term memory network, a gated recurrent unit, or a convolutional neural network; the physical constraint module simulates the change process of the ion exchange rate and the free radical concentration in the anion exchange membrane by establishing a degradation kinetic equation set, where the degradation kinetic equation set includes at least one or more of a diffusion equation based on Fick's law, a Nernst-Planck equation, or a Butler-Volmer equation.

[0041] In this embodiment, the time series prediction module can be implemented by using one or more of a long short-term memory network (LSTM), a gated recurrent unit (GRU), or a convolutional neural network (CNN). Both LSTM and GRU are variants of recurrent neural networks specifically designed for processing time series data. They can effectively capture long-term dependencies in the data and perform well in processing data with time series characteristics such as the performance degradation of anion exchange membranes; while CNN can automatically extract local features in the data through convolution operations and has unique advantages in feature learning. The physical constraint module simulates the change process of the ion exchange rate and the free radical concentration in the anion exchange membrane by establishing a degradation kinetic equation set. The degradation kinetic equation set includes at least one or more of a diffusion equation based on Fick's law (Fick's law), a Nernst-Planck equation (Nernst-Planck equation), or a Butler-Volmer equation (Butler-Volmer equation). The Fick's law diffusion equation is mainly used to describe the diffusion process of substances in a medium, the Nernst-Planck equation takes into account the influence of the electric field on ion migration, and the Butler-Volmer equation can describe the charge transfer process between the electrode and the electrolyte. These equations can model the degradation mechanism of anion exchange membranes from a physical essence and provide a solid physical basis for prediction. By reasonably selecting and combining the implementation methods of the time series prediction module and the physical constraint module, the advantages of both can be fully utilized to achieve accurate prediction and evaluation of the performance of anion exchange membranes.

[0042] In the present disclosure, the Fick's law diffusion equation is as follows:

[0043] Where, C is the free radical concentration, D is the diffusion coefficient, t is the time.

[0044] The Nernst-Planck equation is as follows:

[0045] Among them, J i is the ion flux, z i is the ion charge number, ϕ is the electric potential, F is the Faraday constant, R is the gas constant, T is the temperature.

[0046] The Butler - Volmer equation is as follows:

[0047] Among them, j is the current density, j 0 is the exchange current density, η is the overpotential, α a , α c is the transfer coefficient, F is the Faraday constant, R is the gas constant, T is the temperature.

[0048] According to the embodiments of the present disclosure, the training process of the hybrid prediction model in step S130 includes: The time - series prediction module takes the dynamic feature matrix as input and outputs a preliminary prediction value of the membrane performance degradation trend; The physical constraint module inputs the temperature, humidity, and free radical concentration data in the dynamic feature matrix and calculates the theoretical performance degradation index; Subtract the theoretical degradation index output by the physical constraint module from the preliminary prediction value of the time - series prediction module through residual connection to obtain a residual correction term, and use the corrected performance degradation index as the final output; Taking the true value of the performance degradation index in historical data as the supervision signal, jointly optimize the weight parameters of the time - series prediction module and the learnable reaction rate constant in the physical constraint module by using the gradient descent method.

[0049] In the embodiments of the present disclosure, first, the time series prediction module takes the dynamic feature matrix as input, and after operations through the internal neural network structure, outputs a preliminary prediction value of the membrane performance degradation trend. Then, the physical constraint module inputs the temperature, humidity, and free radical concentration data in the dynamic feature matrix, and based on the pre-established degradation kinetic equations, calculates the theoretical performance degradation index. Next, the theoretical degradation index output by the physical constraint module is subtracted from the preliminary prediction value of the time series prediction module through residual connection to obtain a residual correction term, and the corrected performance degradation index is used as the final output. This process realizes the fusion and correction of the data-driven prediction result and the physical mechanism prediction result, which helps to improve the accuracy and reliability of the prediction. Finally, using the true value of the performance degradation index in the historical data as the supervision signal, the gradient descent method is used to jointly optimize the weight parameters of the time series prediction module and the learnable reaction rate constant in the physical constraint module. Through continuous iterative optimization, the model gradually learns the laws and physical mechanisms in the data, thereby improving the prediction performance of the model and enabling it to more accurately evaluate the performance degradation situation and remaining life of the anion exchange membrane. For example, setting the learning rate to 0.001 and the number of iterations to 1000 times, through continuous iterative optimization, a hybrid prediction model that can accurately predict the performance degradation and remaining life of the AEM is finally obtained.

[0050] In the present disclosure, the degradation kinetic model of the physical constraint module includes: Free radical concentration kinetic equation:

[0051] where [⋅ OH is the hydroxyl radical concentration, k 1, k 2 are the reaction rate constants.

[0052] Ion exchange rate equation:

[0053] where Q is the ion exchange capacity, k ( T ) is the Arrhenius temperature dependence term, E a is the activation energy.

[0054] According to the embodiments of the present disclosure, the step of inputting the dynamic feature matrix into the trained hybrid prediction model in step S140 and outputting the real-time performance degradation index and remaining life confidence interval of the anion exchange membrane includes: Inputting the dynamic feature matrix into the trained hybrid prediction model and outputting the real-time performance degradation index of the anion exchange membrane; Based on the time series fitting results of the real-time performance degradation index, combined with the failure probability model, calculate the confidence interval of the remaining life.

[0055] In the embodiments of the present disclosure, first, input the dynamic feature matrix into the trained hybrid prediction model. Based on the knowledge and rules learned, after complex operations and analyses, the model outputs the real-time performance degradation index of the anion exchange membrane, which can intuitively reflect the performance state of the membrane at the current moment. Then, based on the time series fitting results of the real-time performance degradation index, combined with a failure probability model such as the Weibull distribution model, calculate the confidence interval of the remaining life.

[0056] The time series fitting results can reflect the change trend of the performance degradation index over time. The failure probability model calculates the confidence interval of the remaining life of the anion exchange membrane according to this trend and the preset failure criterion, providing a more scientific and accurate decision-making basis for the maintenance and replacement of the anion exchange membrane, helping to plan maintenance work in advance, reduce the risk of equipment failure, and improve the reliability and economy of the system.

[0057] In the present disclosure, the Weibull distribution formula of the failure probability model is:

[0058] Where, η is the characteristic life, β is the shape parameter, t is the time.

[0059] According to the embodiments of the present disclosure, the method further includes: Trigger a maintenance alarm when the remaining life is lower than a preset threshold.

[0060] In the embodiments of the present disclosure, in practical applications, by monitoring the real-time performance degradation index and the confidence interval of the remaining life of the anion exchange membrane, when the calculated remaining life is lower than the preset threshold, the system will automatically trigger a maintenance alarm. This alarm can be issued in various ways, such as displaying alarm information on the monitoring device, sending text messages or email notifications to maintenance personnel, etc., timely reminding relevant personnel to pay attention to the operating status of the anion exchange membrane, arranging maintenance or replacement plans in advance, avoiding adverse consequences such as equipment downtime and production interruption caused by sudden failures of the anion exchange membrane, effectively improving the stability and reliability of equipment operation, reducing maintenance costs and potential loss risks, and providing a strong guarantee for the safe and efficient operation of anion exchange membrane-related equipment.

[0061] Application example: In the electrolytic water hydrogen production system, the initial thickness of the anion exchange membrane is 250 microns, the initial conductivity is 0.1 S / cm, and the initial swelling rate is 5%. When the system is operating, the set current density is 1.5 A / cm², the working temperature is maintained at 70°C, and the relative humidity is controlled at 60%.

[0062] Install sensors in the electrolytic water hydrogen production system to collect multi-source data of the anion exchange membrane: voltage fluctuation sequence, current density distribution, electrochemical impedance spectrum, temperature gradient value, relative humidity, ion conductivity decay curve, swelling rate dynamic change value. The collected data every 100 hours of operation is shown in Table 1 below:

[0063] Splice the dynamic feature matrix in the manner of the foregoing embodiments, and input the dynamic feature matrix into the trained model. The model uses a long short-term memory network to extract the non-linear degradation mode in the dynamic feature matrix. By establishing a degradation kinetic equation set based on the Nernst-Planck equation, the ion exchange rate and the change of free radical concentration are simulated, and the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane are output. The results are shown in Table 2 below:

[0064] It can be seen from this that the electrolytic water hydrogen production system can accurately predict the performance degradation and remaining life of the anion exchange membrane in advance, and the prediction accuracy reaches 96.8% on average. When the remaining life is less than 100 hours, the system automatically triggers a maintenance alarm, providing a sufficient time window for timely replacement of the membrane material, effectively improving the reliability and operation efficiency of the system.

[0065] Figure 2 The structural block diagram of the anion exchange membrane performance prediction and life evaluation device according to an embodiment of the present disclosure is shown. Among them, the device can be implemented as part or all of an electronic device through software, hardware or a combination of both.

[0066] As Figure 2 shown, the anion exchange membrane performance prediction and life evaluation device 200 includes: An acquisition module 210, configured to acquire a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; A generation module 220, configured to perform spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with time stamp marks; A building module 230, configured to build a hybrid prediction model that fuses deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding. And the two modules perform joint output through residual connection; A prediction module 240, configured to input the dynamic feature matrix into the trained hybrid prediction model and output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

[0067] The anion exchange membrane performance prediction and life evaluation device provided by the embodiments of the present disclosure gives full play to the advantages of both through a hybrid prediction model that fuses deep learning and physical mechanisms. The deep learning model can automatically extract non-linear degradation patterns from the dynamic feature matrix and capture the complex laws of the performance change of the anion exchange membrane; while the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration based on the degradation kinetic equation and incorporates its prior knowledge into the model. This not only helps to improve the adaptability of the model to different working conditions and data changes, enhances the generalization ability of the model, but also increases the interpretability of the model.

[0068] The present disclosure also discloses an electronic device, Figure 3 showing a structural block diagram of the electronic device according to the embodiments of the present disclosure.

[0069] As Figure 3 shown, the electronic device includes a memory and a processor. Among them, the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the method according to the embodiments of the present disclosure.

[0070] The anion exchange membrane performance prediction and life evaluation method includes the following steps: Obtain a multi-source data set in real time during the operation of the anion exchange membrane. The multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; Perform spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with timestamp marks; Build a hybrid prediction model that fuses deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding. And the two modules perform joint output through residual connection; Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

[0071] According to an embodiment of the present disclosure, the electrochemistry data includes a voltage fluctuation sequence, a current density distribution, and an electrochemical impedance spectrum, the environmental monitoring data includes a temperature gradient value, a relative humidity, and its change rate, and the material characterization data includes an ion conductivity decay curve and a dynamic swelling rate change value.

[0072] According to an embodiment of the present disclosure, the generating a time-stamped dynamic feature matrix by performing spatio-temporal alignment and feature fusion on the multi-source data set includes: According to a unified time-stamp benchmark, perform a sliding window statistics on the voltage fluctuation sequence and the current density distribution, extract the root mean square error of the voltage and the peak-valley difference of the current as time-domain features, and at the same time perform an equivalent circuit model fitting on the electrochemical impedance spectrum to extract the frequency-domain degradation features of the charge transfer resistance and the double-layer capacitance; Perform cubic spline interpolation alignment on the temperature gradient value, the relative humidity, and its change rate, and the ion conductivity decay curve according to the time-stamp benchmark to eliminate the time-sequence offset caused by the sampling interval difference; Perform an adaptive normalization process on the dynamic swelling rate change value based on the real-time collected relative humidity to generate a normalized swelling rate feature vector; Splice the processed time-domain features, frequency-domain degradation features, temperature gradient value, relative humidity and / or change rate, ion conductivity decay curve, and normalized swelling rate feature vector into a dynamic feature matrix according to the time-stamp benchmark.

[0073] According to an embodiment of the present disclosure, the time series prediction module is implemented by using one or more of a long short-term memory network, a gated recurrent unit, or a convolutional neural network; the physical constraint module simulates the change process of the ion exchange rate and the free radical concentration in the anion exchange membrane by establishing a degradation kinetic equation set, where the degradation kinetic equation set includes at least one or more of a diffusion equation based on Fick's law, a Nernst-Planck equation, or a Butler-Volmer equation.

[0074] According to an embodiment of the present disclosure, the training process of the hybrid prediction model includes: The time series prediction module takes the dynamic feature matrix as an input and outputs a preliminary prediction value of the membrane performance degradation trend; The physical constraint module inputs the temperature, humidity, and free radical concentration data in the dynamic feature matrix and calculates a theoretical performance degradation index; Subtract the preliminary prediction value of the time series prediction module from the theoretical degradation index output by the physical constraint module through a residual connection to obtain a residual correction term, and use the corrected performance degradation index as the final output; Taking the true value of the performance degradation index in historical data as a supervision signal, jointly optimize the weight parameters of the time series prediction module and the learnable reaction rate constant in the physical constraint module using the gradient descent method.

[0075] According to an embodiment of the present disclosure, the inputting the dynamic feature matrix into the trained hybrid prediction model and outputting the real-time performance degradation index and remaining life confidence interval of the anion exchange membrane includes: Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index of the anion exchange membrane; Based on the time series fitting result of the real-time performance degradation index, calculate the remaining life confidence interval in combination with the failure probability model.

[0076] According to an embodiment of the present disclosure, the method further includes: Trigger a maintenance alarm when the remaining life is lower than a preset threshold.

[0077] Figure 4 The structural schematic diagram of a computer system suitable for implementing the method according to an embodiment of the present disclosure is shown.

[0078] As Figure 4 shown, the computer system includes a processing unit, which can execute various methods in the above embodiments according to the program stored in the read-only memory (ROM) or the program loaded from the storage part into the random access memory (RAM). In the RAM, various programs and data required for the operation of the computer system are also stored. The processing unit, ROM, and RAM are connected to each other through a bus. The input / output (I / O) interface is also connected to the bus.

[0079] The following components are connected to the I / O interface: an input part including a keyboard, a mouse, etc.; an output part including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a LAN card, a modem, etc. The communication part performs a communication process via a network such as the Internet. The drive is also connected to the I / O interface as needed. A removable medium, such as a disk, an optical disc, a magneto-optical disc, a semiconductor memory, etc., is installed on the drive as needed, so that the computer program read from it can be installed into the storage part as needed. Among them, the processing unit can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0080] In particular, according to an embodiment of the present disclosure, the method described above can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product that includes a computer program tangibly embodied on a machine-readable medium, the computer program including program code for performing the above-described method. In such an embodiment, the computer program can be downloaded and installed from a network via a communication section, and / or installed from a removable medium.

[0081] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in an order different from that noted in the drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0082] The units or modules involved in the embodiments described in the present disclosure can be implemented in software or in programmable hardware. The units or modules described can also be provided in a processor, and the names of these units or modules do not, in some cases, constitute a limitation on the units or modules themselves.

[0083] As another aspect, the present disclosure also provides a computer-readable storage medium, which can be the computer-readable storage medium included in the electronic device or computer system in the above embodiments; or can be a computer-readable storage medium that exists separately and is not assembled into the device. The computer-readable storage medium stores one or more programs, and the one or more programs are used by one or more processors to execute the methods described in the present disclosure.

[0084] The above description is only a preferred embodiment of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) disclosed in the present disclosure that have similar functions.

Claims

1. A method for predicting the performance and evaluating the lifespan of an anion exchange membrane, characterized in that, It includes the following steps: Obtain a multi-source dataset during the operation of the anion exchange membrane in real time, where the multi-source dataset includes electrochemical data, environmental monitoring data, and material characterization data; Perform spatio-temporal alignment and feature fusion on the multi-source dataset to generate a dynamic feature matrix with timestamp markings; Construct a hybrid prediction model that combines deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts non-linear degradation patterns in the dynamic feature matrix, and the physical constraint module conducts mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding, and the two modules perform joint output through residual connection; Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

2. The method according to claim 1, characterized in that, The electrochemical data includes a voltage fluctuation sequence, a current density distribution, and an electrochemical impedance spectrum. The environmental monitoring data includes a temperature gradient value, a relative humidity, and its change rate. The material characterization data includes an ion conductivity decay curve and a dynamic change value of the swelling rate.

3. The method according to claim 2, wherein The step of performing spatio-temporal alignment and feature fusion on the multi-source dataset to generate a dynamic feature matrix with timestamp markings includes: According to a unified timestamp benchmark, perform sliding window statistics on the voltage fluctuation sequence and the current density distribution, extract the root mean square error of the voltage and the peak-to-valley difference of the current as time domain features, and at the same time perform equivalent circuit model fitting on the electrochemical impedance spectrum to extract the frequency domain degradation features of the charge transfer resistance and the double layer capacitance; Align the temperature gradient value, the relative humidity, and its change rate with the ion conductivity decay curve according to the timestamp benchmark through cubic spline interpolation to eliminate the timing offset caused by the sampling interval difference; Perform adaptive normalization processing on the dynamic change value of the swelling rate based on the real-time collected relative humidity to generate a normalized swelling rate feature vector; Splice the processed time domain features, frequency domain degradation features, temperature gradient value, relative humidity and / or change rate, ion conductivity decay curve, and normalized swelling rate feature vector into a dynamic feature matrix according to the timestamp benchmark.

4. The method according to claim 1, characterized in that The time series prediction module is implemented by using one or more of long short-term memory networks, gated recurrent units, or convolutional neural networks; the physical constraint module simulates the change process of the ion exchange rate and free radical concentration in the anion exchange membrane by establishing a degradation kinetic equation set, where the degradation kinetic equation set includes at least one or more of the diffusion equation based on Fick's law, the Nernst-Planck equation, or the Butler-Volmer equation.

5. The method according to claim 1, characterized in that The training process of the hybrid prediction model includes: The time series prediction module takes the dynamic feature matrix as input and outputs a preliminary prediction value of the membrane performance degradation trend; The physical constraint module inputs the temperature, humidity, and free radical concentration data in the dynamic feature matrix and calculates the theoretical performance degradation index; Subtract the theoretical degradation index output by the physical constraint module from the preliminary prediction value of the time series prediction module through a residual connection to obtain a residual correction term, and use the corrected performance degradation index as the final output; Taking the true value of the performance degradation index in historical data as a supervision signal, jointly optimize the weight parameters of the time series prediction module and the learnable reaction rate constant in the physical constraint module using the gradient descent method.

6. The method according to claim 1, wherein The step of inputting the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane includes: Input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index of the anion exchange membrane; Based on the time series fitting result of the real-time performance degradation index, calculate the remaining life confidence interval in combination with the failure probability model.

7. The method according to claim 1, characterized in that It further includes: Trigger a maintenance alarm when the remaining life is lower than a preset threshold.

8. An anion exchange membrane performance prediction and life assessment device, characterized in that, It includes: An acquisition module configured to acquire a multi-source data set during the operation of the anion exchange membrane in real time, where the multi-source data set includes electrochemical data, environmental monitoring data, and material characterization data; A generation module configured to perform spatio-temporal alignment and feature fusion on the multi-source data set to generate a dynamic feature matrix with timestamp markings; A construction module configured to construct a hybrid prediction model that integrates deep learning and physical mechanisms. The hybrid prediction model includes a time series prediction module and a physical constraint module coupled with a degradation kinetic equation. Among them, the time series prediction module extracts the non-linear degradation pattern in the dynamic feature matrix, and the physical constraint module performs mechanism modeling on the ion exchange rate and free radical concentration through differential equation embedding, and the two modules perform joint output through a residual connection; A prediction module configured to input the dynamic feature matrix into the trained hybrid prediction model to output the real-time performance degradation index and the remaining life confidence interval of the anion exchange membrane.

9. An electronic device, characterized in that, It includes a memory and a processor; wherein, the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1-7.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that, When the computer instructions are executed by the processor, the method according to any one of claims 1-7 is implemented.

Images