Electrolytic water hydrogen production monitoring and control method with dynamic fluctuation adaptability

Abstract

This invention provides a dynamic fluctuation-adaptive monitoring and control method for hydrogen production via water electrolysis, relating to the field of hydrogen production technology. It includes: real-time acquisition of a multi-dimensional set of operating parameters of the electrolyzer through a multi-source sensor array, and establishment of an electrolysis process prediction model based on LSTM and CNN neural networks to determine the predicted electrolysis state. Based on the prediction results, a dynamic parameter adjustment matrix is ​​constructed, and fuzzy PID control is used to initially adjust the electrolysis current density, electrolyte concentration, and other key parameters of the electrolyzer. After adjustment, the multi-dimensional operating parameter set is updated, the health index of the electrolyzer is calculated, and the relationship between this index and a preset health index threshold is monitored in real time. A second adjustment is implemented based on the monitoring results to ensure the safe and efficient operation of the electrolyzer under dynamic fluctuation conditions. This invention effectively solves the problems of slow response to load changes, insufficient monitoring capabilities, and reliance on manual maintenance in existing technologies for electrolyzers.

Description

Technical Field

[0001] This invention relates to the field of hydrogen production technology, and in particular to a dynamic fluctuation-adaptive method for monitoring and controlling hydrogen production through water electrolysis. Background Technology

[0002] With the rapid development of renewable energy, especially the increasingly widespread application of wind and solar power, water electrolysis for hydrogen production has gradually become an important method for hydrogen production. Hydrogen, as a clean energy source, plays a crucial role in reducing greenhouse gas emissions and addressing global climate change. Therefore, improving the efficiency and stability of hydrogen production has a profound impact on the future application of renewable energy.

[0003] Among various water electrolysis hydrogen production technologies, anion exchange membrane (AEM) water electrolysis has attracted much attention due to its high efficiency, simple operation, and good durability. Compared with traditional proton exchange membrane (PEM) water electrolysis hydrogen production systems, AEM systems have lower energy consumption and cost advantages, and are more adaptable to power fluctuations. This makes AEM water electrolysis hydrogen production technology well-suited for applications with a high proportion of renewable energy, and its application is playing an increasingly important role in peak shaving and load balancing of renewable energy power generation systems.

[0004] However, existing monitoring technologies for hydrogen production via water electrolysis still have shortcomings. For example, patent CN117071000B discloses a remote safety monitoring system for PEM electrolyzers, including an equipment parameter analysis module, a membrane status analysis module, an impurity anomaly analysis module, an early warning assessment analysis module, a cleaning treatment analysis module, and a maintenance database. Although this patent uses sensors to collect current, voltage, and temperature parameters in real time and generates an equipment operation assessment index to determine whether the equipment is in normal working condition, it still has significant deficiencies in actual operation. First, the system lacks the ability to adjust electrolysis parameters in real time when the equipment load changes rapidly, and cannot dynamically respond to changes in the state within the electrolyzer. Especially when the output of renewable energy sources such as wind and solar power fluctuates, traditional systems react slowly, which may lead to a decrease in the stability of the electrolysis process, ultimately affecting the yield and purity of hydrogen. Second, existing monitoring technologies mainly focus on monitoring single parameters, lacking comprehensive analysis and optimization of multiple key indicators of the electrolysis process. Therefore, during equipment operation, it is impossible to effectively integrate information from various modules to make intelligent fault warnings and maintenance decisions, increasing the potential risk of equipment failure. Finally, the strong reliance on manual maintenance is also a major problem with existing monitoring systems. Current maintenance often relies on periodic manual inspections, making fully automated monitoring difficult. This not only increases operation and maintenance costs but also prevents timely adjustments or responses to emergencies, impacting overall hydrogen production efficiency and safety. In conclusion, monitoring and control of anion exchange membrane (AEM) water electrolysis for hydrogen production is crucial. However, the shortcomings of existing monitoring technologies in terms of dynamic adaptability, comprehensive monitoring capabilities, and intelligent decision-making pose significant technical challenges for the development of water electrolysis for hydrogen production. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a dynamic fluctuation-adaptive method for monitoring and controlling hydrogen production through water electrolysis. This invention solves the problems of insufficient dynamic adaptability, comprehensive monitoring capabilities, and intelligent decision-making capabilities in existing water electrolysis hydrogen production monitoring technologies.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A dynamic fluctuation-adaptive method for monitoring and controlling hydrogen production via water electrolysis includes:

[0008] The electrolytic cell's multi-dimensional operating parameters are collected in real time using a multi-source sensor array.

[0009] Based on LSTM and CNN neural networks, and according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, an electrolysis process prediction model is constructed to determine the electrolysis state prediction result of the electrolytic cell within a preset time.

[0010] A dynamic parameter adjustment matrix for the electrolytic cell is constructed based on the electrolysis state prediction results;

[0011] The electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolytic cell are adjusted for the first time using the dynamic parameter adjustment matrix via fuzzy PID control.

[0012] Based on the adjusted electrolytic cell, the multi-dimensional operating parameter set is updated to obtain an updated parameter set;

[0013] The health index of the electrolytic cell is calculated using the updated parameter set;

[0014] The health index of the electrolytic cell is monitored in real time according to a preset health index threshold. The monitoring results are obtained, and the electrolytic cell is adjusted a second time based on the monitoring results.

[0015] Preferably, the multi-dimensional operating parameter set includes:

[0016] Current density, voltage fluctuation, temperature gradient, pressure change, electrolyte flow rate, hydrogen purity, and anion exchange membrane stress parameters.

[0017] Preferably, the method for constructing the electrolysis process prediction model is as follows:

[0018] The multi-dimensional operating parameter set is obtained and data preprocessed to obtain the preprocessed multi-dimensional operating parameter set.

[0019] The preprocessed multi-dimensional operating parameter set is constructed into three-dimensional input data according to time steps;

[0020] Construct a CNN sub-model and an LSTM sub-model, wherein the CNN sub-model includes an input layer, a convolutional layer, a pooling layer and an output layer, and the LSTM sub-model has 128 LSTM units;

[0021] The three-dimensional input data is input into the CNN sub-model to obtain the input data of the LSTM sub-model, and then input into the LSTM sub-model to obtain the first output result;

[0022] A fully connected layer is added to the LSTM sub-model, and the first output result is converted into a second output result;

[0023] The predicted results of the electrolysis process are determined using the linear activation function and the second output result;

[0024] The initial electrolysis process prediction model is determined based on the electrolysis process prediction results and the constructed CNN sub-model and LSTM sub-model.

[0025] The initial electrolysis process prediction model is trained to obtain the final electrolysis process prediction model.

[0026] Preferably, the expression for the electrolysis process prediction model is:

[0027]

[0028] Among them, X t-N:t The input is a multi-dimensional set of runtime parameters, with a time window of tN:t. The electrolysis state prediction result is given at time step T, where σ is the linear activation function and W is the value of W. FC For a fully connected layer, b FC This is a bias term.

[0029] Preferably, the expression for the dynamic parameter adjustment matrix is:

[0030]

[0031] Among them, M adjust Adjust the matrix for dynamic parameters. These are the predicted current density, concentration change rate, temperature gradient, and pressure fluctuation amplitude, respectively. The time derivative of the prediction parameters, To predict the short-term fluctuations of the parameters, For the spatial gradient of the predicted parameters, α * β * γ * δ * These are the first coupling weight coefficient, the second coupling weight coefficient, the third coupling weight coefficient, and the fourth coupling weight coefficient, respectively.

[0032] Preferably, the first adjustment of the electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolytic cell using the dynamic parameter adjustment matrix via fuzzy PID includes:

[0033] Extract the current electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolyzer based on a multi-dimensional set of operating parameters;

[0034] Design a fuzzy PID controller and calculate its output;

[0035] The adjustment factor is determined based on the dynamic parameter adjustment matrix;

[0036] The current electrolysis current density, electrolyte concentration, operating temperature, and pressure are adjusted for the first time based on the output of the PID controller and the adjustment factor.

[0037] Preferably, the design of the fuzzy PID controller and the calculation of the PID controller output include:

[0038] Fuzzy control rules are established based on the characteristics of the input variables, wherein the input variables are electrolysis current density error, concentration error, temperature error and pressure error;

[0039] Based on the fuzzy control rules, define fuzzy sets and design corresponding membership functions;

[0040] Based on the fuzzy set, a fuzzy control rule base is constructed;

[0041] The actual error value is converted into a fuzzy value and integrated with the fuzzy output through a fuzzy inference mechanism based on the fuzzy control rule base.

[0042] The fuzzy output is defuzzified to obtain the output control signal.

[0043] Preferably, the formula for calculating the health index is:

[0044]

[0045] Among them, HI t For the health index, ξ i,t Let ξ be the real-time value of the i-th monitoring parameter. i,ref Let Δξ be the reference value of the i-th parameter under standard operating conditions. i,t Let ω be the short-term fluctuation of the i-th parameter. i λ is the parameter weighting coefficient. i Let M be the parameter sensitivity attenuation factor, η be the dynamic control compensation coefficient, and ||M adjust || F The Frobenius norm of the dynamic parameter adjustment matrix, ||M adjust,0 || F Adjust the Frobenius norm of the matrix for the initial state.

[0046] A dynamic fluctuation-adaptive monitoring and control system for hydrogen production via water electrolysis includes:

[0047] The data acquisition module is used to acquire a set of multi-dimensional operating parameters of the electrolyzer in real time through a multi-source sensor array;

[0048] An electrolysis state prediction module is used to construct an electrolysis process prediction model based on LSTM neural network and CNN neural network, according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, to determine the electrolysis state prediction result of the electrolytic cell within a preset time.

[0049] The adjustment matrix determination module is used to construct a dynamic parameter adjustment matrix for the electrolyzer based on the electrolysis state prediction results.

[0050] The first parameter adjustment module is used to make the first adjustment of the electrolytic current density, electrolyte concentration, operating temperature and pressure of the electrolytic cell by using the dynamic parameter adjustment matrix through fuzzy PID;

[0051] The update module is used to update the multi-dimensional operating parameter set according to the adjusted electrolytic cell, so as to obtain the updated parameter set;

[0052] A health index calculation module is used to calculate the health index of the electrolytic cell using the updated parameter set.

[0053] The second parameter adjustment module is used to monitor the health index of the electrolytic cell in real time according to a preset health index threshold, obtain the monitoring results, and make a second adjustment to the electrolytic cell based on the monitoring results.

[0054] The present invention discloses the following technical effects:

[0055] This invention provides a dynamic fluctuation-adaptive method for monitoring and controlling hydrogen production through water electrolysis, characterized by comprising: real-time acquisition of a multi-dimensional set of operating parameters of an electrolyzer via a multi-source sensor array; constructing an electrolysis process prediction model based on an LSTM neural network and a CNN neural network, according to the multi-dimensional set of operating parameters and historical data of the electrolyzer, to determine the predicted electrolysis state of the electrolyzer within a preset time period; constructing a dynamic parameter adjustment matrix for the electrolyzer based on the predicted electrolysis state; using the dynamic parameter adjustment matrix to perform a first adjustment on the electrolysis current density, electrolyte concentration, operating temperature, and pressure of the electrolyzer through fuzzy PID control; updating the multi-dimensional set of operating parameters based on the adjusted electrolyzer to obtain an updated parameter set; calculating the health index of the electrolyzer using the updated parameter set; and monitoring the health index of the electrolyzer in real time according to a preset health index threshold to obtain monitoring results and performing a second adjustment on the electrolyzer based on the monitoring results. This invention achieves immediate and accurate state perception and prediction of the electrolyzer by real-time acquisition of multi-dimensional operating parameters through a multi-source sensor array and data analysis using LSTM and CNN neural networks, improving the system's dynamic adjustment capability under load fluctuations. It can respond promptly to changes in external conditions, thereby optimizing the electrolysis process and improving the stability and efficiency of hydrogen production. The electrolysis process prediction model is constructed based on a multi-dimensional set of operating parameters and historical data. Through a dynamic parameter adjustment matrix, it comprehensively regulates the electrolyzer, solving the problem of single-parameter blind spots in existing monitoring technologies. It enables real-time monitoring and optimization of multiple key parameters such as current density, electrolyte concentration, operating temperature, and pressure, allowing for a more comprehensive and scientific evaluation of the electrolyzer's performance during operation. Using the dynamic parameter adjustment matrix, fuzzy PID control is employed for intelligent regulation of the electrolyzer, while a health index is calculated for real-time monitoring, reducing reliance on manual maintenance and improving the level of maintenance automation. In case of anomalies, the system can react quickly and make secondary adjustments, thereby reducing the risk of equipment failure and improving the safety and efficiency of hydrogen production. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 A flowchart of a dynamic fluctuation-adaptive water electrolysis hydrogen production monitoring and control method provided in an embodiment of the present invention;

[0058] Figure 2This is a schematic diagram of a dynamic fluctuation-adaptive electrolytic water hydrogen production monitoring and control system provided in an embodiment of the present invention.

[0059] Explanation of reference numerals in the attached figures:

[0060] 1-Data acquisition module, 2-Electrolysis state prediction module, 3-Adjustment matrix determination module, 4-First parameter adjustment module, 5-Update module, 6-Health index calculation module, 7-Second parameter adjustment module. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] like Figure 1 As shown, this invention provides a dynamic fluctuation-adaptive method for monitoring and controlling hydrogen production through water electrolysis, comprising:

[0064] Step 100: Real-time acquisition of multi-dimensional operating parameters of the electrolyzer using a multi-source sensor array;

[0065] Step 200: Based on LSTM neural network and CNN neural network, according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, construct an electrolysis process prediction model to determine the electrolysis state prediction result of the electrolytic cell within a preset time.

[0066] Step 300: Construct the dynamic parameter adjustment matrix of the electrolytic cell based on the electrolysis state prediction results;

[0067] Step 400: Using the dynamic parameter adjustment matrix, the electrolytic current density, electrolyte concentration, operating temperature and pressure of the electrolytic cell are adjusted for the first time through fuzzy PID control;

[0068] Step 500: Update the multi-dimensional operating parameter set according to the adjusted electrolytic cell to obtain an updated parameter set;

[0069] Step 600: Calculate the health index of the electrolytic cell using the updated parameter set;

[0070] Step 700: Monitor the health index of the electrolytic cell in real time according to the preset health index threshold, obtain the monitoring results, and make a second adjustment to the electrolytic cell based on the monitoring results.

[0071] Furthermore, the multi-dimensional operating parameter set includes:

[0072] Current density, voltage fluctuation, temperature gradient, pressure change, electrolyte flow rate, hydrogen purity, and anion exchange membrane stress parameters.

[0073] Specifically, the selection of sensors includes:

[0074] Current sensor: Used to measure the current density of an electrolytic cell.

[0075] Voltage sensor: Used to monitor voltage fluctuations in the electrolytic cell.

[0076] Temperature sensor: Used to detect the temperature distribution inside the electrolytic cell.

[0077] Pressure sensor: Used to monitor pressure changes inside the electrolytic cell.

[0078] Flow sensor: Used to measure the flow rate of electrolyte and hydrogen.

[0079] Gas analyzer: used to detect the purity of hydrogen gas.

[0080] Strain sensor: Used to monitor the stress of anion exchange membranes.

[0081] Based on the structure of the electrolytic cell, sensors should be arranged rationally to ensure that each sensor can accurately monitor its required parameters. The following arrangement principles should be followed for the electrolytic cell:

[0082] Uniform distribution: Ensure that each sensor covers different areas of the electrolyzer to obtain comprehensive monitoring data.

[0083] Proximity to critical components: The gas analyzer should be located near the hydrogen outlet, while the current and voltage sensors should be located near the electrodes.

[0084] Furthermore, the method for constructing the electrolysis process prediction model is as follows:

[0085] The multi-dimensional operating parameter set is obtained and data preprocessed to obtain the preprocessed multi-dimensional operating parameter set.

[0086] The preprocessed multi-dimensional operating parameter set is constructed into three-dimensional input data according to time steps;

[0087] Specifically, the time series data is constructed into three-dimensional input data according to a fixed-length window (e.g., a window length of 10, meaning every 10 time points are used as an input sample). The shape of the three-dimensional data is (number of samples, time step, number of features). For example, if there are 1000 time point data points, using a window of length 10: the input data shape = (991, 10, number of features). Assuming the number of features is 7, the final input shape is (991, 10, 7).

[0088] Construct a CNN sub-model and an LSTM sub-model, wherein the CNN sub-model includes an input layer, a convolutional layer, a pooling layer and an output layer, and the LSTM sub-model has 128 LSTM units;

[0089] The three-dimensional input data is input into the CNN sub-model to obtain the input data of the LSTM sub-model, and then input into the LSTM sub-model to obtain the first output result;

[0090] Specifically, the constructed 3D input data is fed into the CNN sub-model. The CNN will extract higher-level features through convolution and pooling operations. The feature output from the CNN is fed into the LSTM sub-model. The LSTM will use historical information from the time series for further analysis to capture the dynamic changes in the electrolysis process.

[0091] A fully connected layer is added to the LSTM sub-model, and the first output result is converted into a second output result;

[0092] Specifically, a fully connected layer is added after the output of the LSTM sub-model, which maps the LSTM output to the final prediction result.

[0093] The predicted results of the electrolysis process are determined using the linear activation function and the second output result;

[0094] The initial electrolysis process prediction model is determined based on the electrolysis process prediction results and the constructed CNN sub-model and LSTM sub-model.

[0095] The initial electrolysis process prediction model is trained to obtain the final electrolysis process prediction model.

[0096] Furthermore, the expression for the electrolysis process prediction model is as follows:

[0097]

[0098] Among them, X t-N:t The input is a multi-dimensional set of runtime parameters, with a time window of tN:t. The electrolysis state prediction result is given at time step T, where σ is the linear activation function and W is the value of W. FC For a fully connected layer, bFC This is a bias term.

[0099] Expanding the above equation, we get:

[0100] Input Feature Extraction (CNN Part)

[0101] F CNN =ReLU(MaxPool(Conv3D(X) t-N:t ,K conv )));

[0102] Temporal Feature Modeling (LSTM Part)

[0103] H t =LSTM θ (F CNN H t-1 );

[0104] Predicted output (fully connected layer)

[0105]

[0106] Among them, F CNN For the high-dimensional feature tensor output by the CNN, capture the spatial coupling relationship of multiple parameters, K conv H represents the parameters of the CNN convolution kernel, used to extract spatial local features (such as the spatial correlation between temperature gradient and current density). t The hidden states of the LSTM are encoded with temporal dynamic features (such as the hysteretic effect of pressure fluctuations on membrane stress).

[0107] In the above formula, CNN extracts the spatial correlation of multiple parameters (such as the local coupling of "temperature gradient-membrane stress"), and LSTM captures the temporal evolution law (such as voltage fluctuation trend). This overcomes the limitation of traditional single models that only focus on time or space. The prediction error is reduced by about 18% compared with the single LSTM model (verified by simulation data). In particular, the prediction stability is significantly improved in high-fluctuation scenarios (such as sudden wind speed changes).

[0108] Furthermore, the expression for the dynamic parameter adjustment matrix is:

[0109]

[0110] Among them, M adjust Adjust the matrix for dynamic parameters. These are the predicted current density, concentration change rate, temperature gradient, and pressure fluctuation amplitude, respectively. The time derivative of the prediction parameters, To predict the short-term fluctuations of the parameters, For the spatial gradient of the predicted parameters, α * β *γ * δ * These are the first coupling weight coefficient, the second coupling weight coefficient, the third coupling weight coefficient, and the fourth coupling weight coefficient, respectively.

[0111] Furthermore, the initial adjustment of the electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolytic cell using the dynamic parameter adjustment matrix via fuzzy PID includes:

[0112] Extract the current electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolyzer based on a multi-dimensional set of operating parameters;

[0113] Design a fuzzy PID controller and calculate its output;

[0114] The adjustment factor is determined based on the dynamic parameter adjustment matrix;

[0115] The current electrolysis current density, electrolyte concentration, operating temperature, and pressure are adjusted for the first time based on the output of the PID controller and the adjustment factor.

[0116] Specifically, based on the calculated PID controller output and the value of the dynamic adjustment factor, the electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolyzer are adjusted for the first time.

[0117] Electrolysis current density: Adjust the current of the electrolytic cell to ensure production efficiency.

[0118] Electrolyte concentration: Adjust the composition of the electrolyte as needed to optimize reaction efficiency.

[0119] Operating temperature: The temperature inside the electrolytic cell is adjusted by cooling or heating measures.

[0120] Pressure: The amount of gas emitted or input will be dynamically increased or decreased based on electrolysis requirements and safety standards.

[0121] Furthermore, the design method of the fuzzy PID controller includes:

[0122] Fuzzy control rules are established based on the characteristics of the input variables, wherein the input variables are electrolysis current density error, concentration error, temperature error and pressure error;

[0123] Define fuzzy sets and design corresponding membership functions, for example:

[0124] The membership functions set for the electrolysis current density error include "negative large" (NB), "negative small" (NS), "zero" (ZE), "positive small" (PS), and "positive large" (PB), which correspond to different error ranges.

[0125] Define membership functions similarly for output quantities (such as adjustments to electrolysis current density, concentration, temperature, and pressure) to ensure that different output quantities have corresponding fuzzy sets.

[0126] Construct a fuzzy control rule base, for example:

[0127] If the current density error is "negatively large" and the concentration error is "negatively small", then the adjustment amount of the electrolysis current density is "positively large".

[0128] If the temperature error is "positive small" and the pressure error is "zero", then the adjustment amount for the working temperature is "negative small".

[0129] In the fuzzy inference process, fuzzy control rules are used to fuzzify the input, transforming the actual error value into a fuzzy value;

[0130] The fuzzy output obtained through fuzzy inference mechanism is finally defuzzified to obtain a specific control output signal, which is used to guide the parameter adjustment of the electrolytic cell.

[0131] Furthermore, the health index is a comprehensive indicator calculated by taking into account various operating parameters of the electrolyzer (such as current density, temperature, pressure, hydrogen purity, etc.) and is used to assess the current status and overall health level of the electrolyzer.

[0132] The calculated health index is compared with a preset threshold to classify the electrolytic cell into a healthy state or an alarm state.

[0133] If the monitoring results show an abnormal health index, first analyze the key parameters that caused the change in the health index (e.g., electrolyte concentration too low, temperature too high, etc.).

[0134] Adjustment measures:

[0135] Based on the analysis results, a strategy adjustment was determined, and the adjustment measures included:

[0136] Electrolysis current density: Increase or decrease the electrolysis current density in response to the health status.

[0137] Electrolyte concentration: Adjust the electrolyte ratio to ensure that its concentration is maintained within the ideal range.

[0138] Temperature regulation: Adjust the temperature inside the tank by increasing cooling or heating to restore it to a suitable range.

[0139] Pressure adjustment: Increase or decrease the system pressure as needed to stabilize the electrolysis reaction.

[0140] Furthermore, the formula for calculating the health index is as follows:

[0141]

[0142] Among them, HI t For the health index, ξ i,t Let ξ be the real-time value of the i-th monitoring parameter. i,ref Let Δξ be the reference value of the i-th parameter under standard operating conditions. i,t Let ω be the short-term fluctuation of the i-th parameter. i λ is the parameter weighting coefficient. i Let M be the parameter sensitivity attenuation factor, η be the dynamic control compensation coefficient, and ||M adjust || F The Frobenius norm of the dynamic parameter adjustment matrix, ||M adjust,0 || F Adjust the Frobenius norm of the matrix for the initial state.

[0143] like Figure 2 As shown, the present invention also provides a dynamic fluctuation-adaptive monitoring and control system for hydrogen production via water electrolysis, comprising:

[0144] Data acquisition module 1 is used to acquire a set of multi-dimensional operating parameters of the electrolytic cell in real time through a multi-source sensor array;

[0145] Electrolysis state prediction module 2 is used to construct an electrolysis process prediction model based on LSTM neural network and CNN neural network, according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, to determine the electrolysis state prediction result of the electrolytic cell within a preset time.

[0146] The adjustment matrix determination module 3 is used to construct the dynamic parameter adjustment matrix of the electrolytic cell based on the electrolysis state prediction results;

[0147] The first parameter adjustment module 4 is used to make the first adjustment of the electrolytic current density, electrolyte concentration, working temperature and pressure of the electrolytic cell by using the dynamic parameter adjustment matrix through fuzzy PID;

[0148] Update module 5 is used to update the multi-dimensional operating parameter set according to the adjusted electrolytic cell, so as to obtain an updated parameter set;

[0149] Health index calculation module 6 is used to calculate the health index of the electrolytic cell using the updated parameter set;

[0150] The second parameter adjustment module 7 is used to monitor the health index of the electrolytic cell in real time according to a preset health index threshold, obtain the monitoring results, and make a second adjustment to the electrolytic cell based on the monitoring results.

[0151] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0152] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for monitoring and controlling hydrogen production via water electrolysis with dynamic fluctuation adaptability, characterized in that, include: The electrolytic cell's multi-dimensional operating parameters are collected in real time using a multi-source sensor array. Based on LSTM and CNN neural networks, and according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, an electrolysis process prediction model is constructed to determine the electrolysis state prediction result of the electrolytic cell within a preset time. A dynamic parameter adjustment matrix for the electrolytic cell is constructed based on the electrolysis state prediction results; The electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolytic cell are adjusted for the first time using the dynamic parameter adjustment matrix via fuzzy PID control. Based on the adjusted electrolytic cell, the multi-dimensional operating parameter set is updated to obtain an updated parameter set; The health index of the electrolytic cell is calculated using the updated parameter set; The health index of the electrolytic cell is monitored in real time according to a preset health index threshold, the monitoring results are obtained, and the electrolytic cell is adjusted a second time based on the monitoring results. The method for constructing the electrolysis process prediction model is as follows: The multi-dimensional operating parameter set is obtained and data preprocessed to obtain the preprocessed multi-dimensional operating parameter set. The preprocessed multi-dimensional operating parameter set is constructed into three-dimensional input data according to time steps; Construct a CNN sub-model and an LSTM sub-model, wherein the CNN sub-model includes an input layer, a convolutional layer, a pooling layer and an output layer, and the LSTM sub-model has 128 LSTM units; The three-dimensional input data is input into the CNN sub-model to obtain the input data of the LSTM sub-model, and then input into the LSTM sub-model to obtain the first output result; A fully connected layer is added to the LSTM sub-model, and the first output result is converted into a second output result; The predicted results of the electrolysis process are determined using the linear activation function and the second output result; The initial electrolysis process prediction model is determined based on the electrolysis process prediction results and the constructed CNN sub-model and LSTM sub-model. The initial electrolysis process prediction model is trained to obtain the final electrolysis process prediction model; The expression for the electrolysis process prediction model is: Among them, X t-N:t The input is a multi-dimensional set of runtime parameters, with a time window of tN:t. The electrolysis state prediction result is given at time step T, where σ is the linear activation function and W is the value of W. FC For a fully connected layer, b FC For bias terms; The expression for the dynamic parameter adjustment matrix is: Among them, M adjust Adjust the matrix for dynamic parameters. These are the predicted current density, concentration change rate, temperature gradient, and pressure fluctuation amplitude, respectively. The time derivative of the prediction parameters, To predict the short-term fluctuations of the parameters, For the spatial gradient of the predicted parameters, α * β * γ * δ * These are the first coupling weight coefficient, the second coupling weight coefficient, the third coupling weight coefficient, and the fourth coupling weight coefficient, respectively.

2. The method for monitoring and controlling hydrogen production via water electrolysis with dynamic fluctuation adaptability according to claim 1, characterized in that, The multi-dimensional operating parameter set includes: Current density, voltage fluctuation, temperature gradient, pressure change, electrolyte flow rate, hydrogen purity, and anion exchange membrane stress parameters.

3. The method for monitoring and controlling hydrogen production via water electrolysis with dynamic fluctuation adaptability according to claim 2, characterized in that, The first adjustment of the electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolytic cell using the dynamic parameter adjustment matrix via fuzzy PID includes: Extract the current electrolytic current density, electrolyte concentration, operating temperature, and pressure of the electrolyzer based on a multi-dimensional set of operating parameters; Design a fuzzy PID controller and calculate its output; The adjustment factor is determined based on the dynamic parameter adjustment matrix; The current electrolysis current density, electrolyte concentration, operating temperature, and pressure are adjusted for the first time based on the output of the PID controller and the adjustment factor.

4. The method for monitoring and controlling hydrogen production via water electrolysis with dynamic fluctuation adaptability according to claim 3, characterized in that, The design of the fuzzy PID controller and the calculation of the PID controller output include: Fuzzy control rules are established based on the characteristics of the input variables, wherein the input variables are electrolysis current density error, concentration error, temperature error and pressure error; Based on the fuzzy control rules, define fuzzy sets and design corresponding membership functions; Based on the fuzzy set, a fuzzy control rule base is constructed; The actual error value is converted into a fuzzy value and integrated with the fuzzy output through a fuzzy inference mechanism based on the fuzzy control rule base. The fuzzy output is defuzzified to obtain the output control signal.

5. The method for monitoring and controlling hydrogen production via water electrolysis with dynamic fluctuation adaptability according to claim 3, characterized in that, The formula for calculating the health index is as follows: Among them, HI t For the health index, ξ i,t Let ξ be the real-time value of the i-th monitoring parameter. i,ref Let Δξ be the reference value of the i-th parameter under standard operating conditions. i,t Let ω be the short-term fluctuation of the i-th parameter. i λ is the parameter weighting coefficient. i Let M be the parameter sensitivity attenuation factor, η be the dynamic control compensation coefficient, and ||M adjust || F The Frobenius norm of the dynamic parameter adjustment matrix, ||M adjust,0 || F Adjust the Frobenius norm of the matrix for the initial state.

6. A dynamic fluctuation-adaptive monitoring and control system for hydrogen production via water electrolysis, characterized in that, include: The data acquisition module is used to acquire a set of multi-dimensional operating parameters of the electrolyzer in real time through a multi-source sensor array; An electrolysis state prediction module is used to construct an electrolysis process prediction model based on LSTM neural network and CNN neural network, according to the multi-dimensional operating parameter set and the historical data of the electrolytic cell, to determine the electrolysis state prediction result of the electrolytic cell within a preset time. The adjustment matrix determination module is used to construct a dynamic parameter adjustment matrix for the electrolyzer based on the electrolysis state prediction results. The first parameter adjustment module is used to make the first adjustment of the electrolytic current density, electrolyte concentration, operating temperature and pressure of the electrolytic cell by using the dynamic parameter adjustment matrix through fuzzy PID; The update module is used to update the multi-dimensional operating parameter set according to the adjusted electrolytic cell, so as to obtain the updated parameter set; A health index calculation module is used to calculate the health index of the electrolytic cell using the updated parameter set. The second parameter adjustment module is used to monitor the health index of the electrolytic cell in real time according to a preset health index threshold, obtain the monitoring results, and make a second adjustment to the electrolytic cell based on the monitoring results.

Images