Water electrolysis hydrogen production hydrogen purification pressure optimization system and control strategy

Abstract

The invention discloses a hydrogen purification pressure optimization system and a control strategy for hydrogen production through electrolysis of water, and relates to the technical field of hydrogen production through electrolysis of water, and the hydrogen purification system comprises a hydrogen purification unit which comprises a plurality of adsorption towers arranged in parallel and is used for adsorbing and removing impurities in hydrogen; the pressure optimization system comprises a pressure sensor, a pressure regulating valve and a controller, the pressure sensor is used for monitoring the pressure in the adsorption tower in real time, the pressure regulating valve regulates the pressure according to an instruction of the controller, and the controller is used for regulating the pressure according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process. Setting pressure control parameters to adjust the pressure in the adsorption tower in real time; and the control unit is used for performing coordinated management on the temperature and the flow of the pressure optimization system, and forming closed-loop feedback regulation of parameters by comparing the actual purity of the purified hydrogen with a preset purity standard. The hydrogen purity is improved, the adsorbent works under the optimal pressure by optimizing the pressure control strategy, the hydrogen purity is effectively improved, and the energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen production by water electrolysis, and more particularly to a hydrogen purification pressure optimization system and control strategy for hydrogen production by water electrolysis. Background Art

[0002] With the growing demand for clean energy, hydrogen production by water electrolysis has attracted increasing attention as a green and sustainable method of hydrogen production. However, pressure control is crucial in the purification process after hydrogen preparation. In the process of hydrogen production by water electrolysis, hydrogen purification is a key link, which is directly related to the quality of hydrogen and the effect of subsequent applications. Traditional pressure control methods are often relatively extensive and cannot be adjusted in real time according to different working conditions, resulting in problems such as unstable hydrogen purity and high energy consumption. For example, when the gas flow fluctuates and the operating state of the electrolyzer changes, a fixed pressure setting is difficult to guarantee the purification effect, which not only affects the quality of hydrogen for subsequent high-end applications, but also causes unnecessary energy waste. For example, pressure swing adsorption (PSA) has certain deficiencies in pressure control, energy consumption and purification efficiency. The adsorption performance of the adsorbent will change under different pressures, affecting the purification effect.

[0003] Therefore, how to propose a hydrogen purification pressure optimization system and control strategy for hydrogen production by water electrolysis, a pressure control strategy, a temperature control strategy and a flow control strategy based on the adsorption characteristics of the adsorbent, so as to achieve precise control of the hydrogen purification process and improve the purity of hydrogen and the operating efficiency of the system is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0004] In view of this, the present invention provides a hydrogen purification pressure optimization system and control strategy for hydrogen production by electrolysis of water. The pressure control strategy, temperature control strategy and flow control strategy based on the adsorption characteristics of the adsorbent can achieve precise control of the hydrogen purification process, improve the purity of hydrogen and the operating efficiency of the system. In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0005] A hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water, comprising: a hydrogen purification unit, a pressure optimization system and a control unit;

[0006] The hydrogen purification unit comprises a plurality of adsorption towers arranged in parallel, each of which is filled with an adsorbent for adsorbing and removing impurities in the hydrogen;

[0007] The pressure optimization system includes a pressure sensor, a pressure regulating valve and a controller, wherein the pressure sensor is used to monitor the pressure in the adsorption tower in real time, the pressure regulating valve regulates the pressure according to the instruction of the controller, and the controller is used to set the pressure control parameters according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process to regulate the pressure in the adsorption tower in real time;

[0008] The control unit is used to coordinate the temperature and flow of the pressure optimization system, and to form a closed-loop feedback adjustment of the parameters by comparing the actual purity of the purified hydrogen with the preset purity standard.

[0009] Optionally, regulating valves are provided between the adsorption towers arranged in parallel, and the adsorption towers are interconnected through pipelines to form an integrated purification system.

[0010] Optionally, a temperature sensor is also included, which is installed in the hydrogen purification unit and is used to collect the real-time temperature inside the adsorption tower.

[0011] Optionally, a flow valve is further included, wherein the flow valve is placed in the pipeline between the adsorption towers, and the flow data is detected through the flow valve.

[0012] Optionally, a control strategy for a pressure optimization system for hydrogen purification by water electrolysis includes:

[0013] Obtain pressure data, temperature data and flow data of the purification process;

[0014] Perform feature extraction according to the hydrogen purification process and determine the key feature parameters corresponding to the purification process;

[0015] Determining the influence index of the key characteristic parameters on hydrogen purification;

[0016] Determine the target characteristic parameter whose influence index is greater than the influence index threshold;

[0017] The target characteristic parameters are input into a pre-trained deep convolutional neural network to obtain the degree of purification of the hydrogen output by the pre-trained deep convolutional neural network.

[0018] Optionally, after determining that the influence index is greater than the target characteristic parameter of the influence index threshold, the method further includes:

[0019] The neural network pre-trained with the standard dataset is trained on the collected historical dataset;

[0020] Use the trained neural network to predict the target characteristic parameters of the degree of purification, obtain the characteristic vectors of the target characteristic parameters in the time series, and construct the target characteristic parameter matrix;

[0021] The coupling vector between the target characteristic parameters, i.e., the spatial characteristics of hydrogen purification, is calculated through the obtained target characteristic parameter matrix;

[0022] Using the spatial features of hydrogen purification as input to a pre-trained deep convolutional neural network;

[0023] The degree of purification of the hydrogen is obtained through the mapping relationship between the degree of purification of the hydrogen and its spatial characteristics.

[0024] Optionally, before obtaining the characteristic vector of the target characteristic parameter in the time series and constructing the target characteristic parameter matrix, it also includes: standardizing the target characteristic parameter to map the value of the target characteristic parameter to a preset value range.

[0025] Optionally, determining the influence index of the key characteristic parameter on hydrogen purification includes: determining the correlation coefficient between the key characteristic parameter and hydrogen purification, and the contribution rate of the key characteristic parameter to the hydrogen purification; and taking the product of the contribution rate and the correlation coefficient as the influence index of the hydrogen purification.

[0026] Optionally, constructing the target feature parameter matrix includes:

[0027] The combined signal formed by the pressure signal, temperature signal and flow signal indicates the degree of purification under the time series. The signal intensity of different time series is collected, and the degree of purification at the time point to be measured is expressed as:

[0028]

[0029] in, represents the degree of purification measured at the tth time, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th sensor, and the degree of purification at the i-th measured time point is obtained through the j-th sensor to construct the purification degree fingerprint library

[0030]

[0031] Where m is the number of sensors, n is the number of time series points, and the i-th row vector ψ i =[φ i1 ,φ i2 ,...,φ im ], which is the fingerprint feature value of the measured point measured by m sensors. Different signals are measured by different types of sensors to form a multi-dimensional feature dimension. The information measured by each sensor at the measured time point is the degree of hydrogen purification.

[0032] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a pressure optimization system and control strategy for hydrogen purification by water electrolysis, which has the following beneficial effects:

[0033] The present invention proposes a hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water, comprising: a hydrogen purification unit, a pressure optimization system and a control unit; the hydrogen purification unit comprises a plurality of adsorption towers arranged in parallel, each adsorption tower is filled with an adsorbent for adsorbing and removing impurities in hydrogen; the pressure optimization system comprises a pressure sensor, a pressure regulating valve and a controller, the pressure sensor is used for real-time monitoring of the pressure in the adsorption tower, the pressure regulating valve regulates the pressure according to the instruction of the controller, the controller is used for setting the pressure control parameters according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process to regulate the pressure in the adsorption tower in real time; the control unit is used for coordinated management of the temperature and flow of the pressure optimization system, and forms a closed-loop feedback regulation of the parameters by comparing the actual purity of the purified hydrogen with the preset purity standard. The present invention improves the purity of hydrogen. By optimizing the pressure control strategy, the adsorbent works at the optimal pressure, effectively improving the purity of hydrogen and meeting the needs of different application scenarios for high-purity hydrogen. It reduces energy consumption. The optimized system can reduce unnecessary energy consumption, such as switching the adsorption tower in time when the adsorption capacity of the adsorbent reaches saturation, avoiding energy waste caused by excessive adsorption. It extends the life of the equipment. Reasonable pressure and temperature control are helpful to protect the adsorbent and equipment, extend their service life, and reduce maintenance costs. It has strong adaptability. The system can adapt to different hydrogen production processes and hydrogen purification needs, and has good versatility and flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 A schematic diagram of the structure of a hydrogen purification pressure optimization system for producing hydrogen through water electrolysis provided by the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] The embodiment of the present invention discloses a hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water, such as Figure 1 As shown, it includes: a hydrogen purification unit, a pressure optimization system and a control unit;

[0038] The hydrogen purification unit comprises a plurality of adsorption towers arranged in parallel, each of which is filled with an adsorbent for adsorbing and removing impurities in the hydrogen;

[0039] The pressure optimization system includes a pressure sensor, a pressure regulating valve and a controller, wherein the pressure sensor is used to monitor the pressure in the adsorption tower in real time, the pressure regulating valve regulates the pressure according to the instruction of the controller, and the controller is used to set the pressure control parameters according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process to regulate the pressure in the adsorption tower in real time;

[0040] The control unit is used to coordinate the temperature and flow of the pressure optimization system, and to form a closed-loop feedback adjustment of the parameters by comparing the actual purity of the purified hydrogen with the preset purity standard.

[0041] Furthermore, regulating valves are provided between the adsorption towers arranged in parallel, and the adsorption towers are interconnected through pipelines to form an integrated purification system.

[0042] Furthermore, it also includes a temperature sensor, which is installed in the hydrogen purification unit and is used to collect the real-time temperature inside the adsorption tower.

[0043] Furthermore, it also includes a flow valve, which is placed in the pipeline between the adsorption towers, and the flow data is detected through the flow valve.

[0044] Furthermore, a control strategy for a hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water includes:

[0045] Obtain pressure data, temperature data and flow data of the purification process;

[0046] Perform feature extraction according to the hydrogen purification process and determine the key feature parameters corresponding to the purification process;

[0047] Determining the influence index of the key characteristic parameters on hydrogen purification;

[0048] Determine the target characteristic parameter whose influence index is greater than the influence index threshold;

[0049] The target characteristic parameters are input into a pre-trained deep convolutional neural network to obtain the degree of purification of the hydrogen output by the pre-trained deep convolutional neural network.

[0050] In a specific embodiment, before inputting the target characteristic parameter into a pre-trained deep convolutional neural network to obtain the degree of purification of the hydrogen output by the pre-trained deep convolutional neural network, the method further includes:

[0051] Sort each of the target characteristic parameters in chronological order to construct a corresponding time series characteristic matrix; determine the characteristic parameter proportion of each characteristic parameter in the time series characteristic matrix; determine the information entropy of each characteristic parameter according to the characteristic parameter proportion; take the difference between the constant 1 and the information entropy as the coefficient of variation; determine the sum of the coefficients of variation of each characteristic parameter, and take the ratio of the coefficient of variation of each characteristic parameter to the sum as the weight of each characteristic parameter; perform time series analysis on the weight of each characteristic parameter to determine the time series analysis result, wherein the time series analysis result is used to represent the changing trend of the weight over time; set a weight threshold based on the time series analysis result, and adjust the weight of the characteristic parameter when the weight of the characteristic parameter exceeds the weight threshold.

[0052] Furthermore, after determining that the influence index is greater than the target characteristic parameter of the influence index threshold, the method further includes:

[0053] The neural network pre-trained with the standard dataset is trained on the collected historical dataset;

[0054] Use the trained neural network to predict the target characteristic parameters of the degree of purification, obtain the characteristic vectors of the target characteristic parameters in the time series, and construct the target characteristic parameter matrix;

[0055] The coupling vector between the target characteristic parameters, i.e., the spatial characteristics of hydrogen purification, is calculated through the obtained target characteristic parameter matrix;

[0056] Using the spatial features of hydrogen purification as input to a pre-trained deep convolutional neural network;

[0057] The degree of purification of the hydrogen is obtained through the mapping relationship between the degree of purification of the hydrogen and its spatial characteristics.

[0058] In a specific implementation, the neural network pre-trained with the standard data set is trained on the collected historical data set, including:

[0059] 1) Obtain the network weights of YOLOv3 as the initialization data of the neural network by downloading the YOLOv3 training data from the open source website. The data is required to be trained and saved using pytorch, and target detection can be performed on a standard data set, which is the normal operation data of the hydrogen purification process;

[0060] 2) Use pytorch to extract the weights corresponding to the first two outputs in the network parameters, that is, the outputs obtained by 32-fold downsampling and 16-fold downsampling, delete the weights of the last 1*1 convolutional layer and match it with the established neural network model, and randomly initialize the network parameters of the last convolutional layer to obtain the initialization weights of all layers of the neural network;

[0061] 3) The training set is used to complete the training process of the neural network. After all the training sets are trained once, the validation set is used for validation to prevent overfitting of the network. The Adam optimizer in pytorch is used for training, and the hyperparameters are set with reference to YOLOv3 to complete the training of the network.

[0062] In a specific implementation, the loss function and network structure of the neural network are determined:

[0063] (1) Neural network loss function: In a neural network, the loss function includes probability loss and noise loss of feature points. A larger weight is assigned to the noise loss of normal feature points, and a smaller weight is assigned to the noise loss of fault feature points. Finally, the neural network loss function is obtained.

[0064] (2) Network structure: The network structure is the structure corresponding to the first two output layers of the YOLOv3 network structure, that is, the outputs obtained by 32-fold downsampling and 16-fold downsampling. The output dimension of the last 1*1 convolutional layer of the network is modified to 20, that is, the output layer of the neural network, and the network structure branch corresponding to the 8-fold downsampling is deleted, which is the final neural network structure.

[0065] Furthermore, before obtaining the characteristic vector of the target characteristic parameter in the time series and constructing the target characteristic parameter matrix, it also includes: standardizing the target characteristic parameter to map the value of the target characteristic parameter to a preset value range.

[0066] In a specific embodiment, the standardization process includes:

[0067] The data of each target characteristic parameter is standardized, and the sample values ​​of each characteristic parameter are compressed within the [0-1] interval to eliminate the influence of unit differences on the calculation results, so that each characteristic parameter can be compared at the same scale. The standardization formula can be referred to as shown in the following formula:

[0068]

[0069] Among them, X′ ij is the standardized data matrix; X ij is the original data matrix; i is the i-th sample, j is the j-th feature parameter, m is the number of samples, and n is the number of feature parameters. The above formula (1) is the normalized formula for positive feature parameters, i.e., the higher the value, the better the feature parameter; the above formula (2) is the normalized formula for negative feature parameters, i.e., the lower the value, the better the feature parameter.

[0070] Furthermore, determining the influence index of the key characteristic parameter on hydrogen purification includes: determining the correlation coefficient between the key characteristic parameter and hydrogen purification, and the contribution rate of the key characteristic parameter to the hydrogen purification; and taking the product of the contribution rate and the correlation coefficient as the influence index of the hydrogen purification.

[0071] In a specific implementation, the correlation coefficient R between each characteristic parameter and hydrogen purification is calculated using a correlation analysis method, and the contribution rate analysis method is used to calculate the contribution rate C of each characteristic parameter to hydrogen purification. The correlation index and the contribution rate index are integrated to form a new evaluation index, namely, the influence index I. The calculation formula of the influence is as follows:

[0072] I i =R i *C i ; Among them, I i Represents the influence (index) of the i-th feature parameter, R i represents the correlation coefficient of the i-th characteristic parameter, C i Represents the contribution rate of the i-th characteristic parameter.

[0073] Furthermore, the constructing of the target feature parameter matrix includes:

[0074] The combined signal formed by the pressure signal, temperature signal and flow signal indicates the degree of purification under the time series. The signal intensity of different time series is collected, and the degree of purification at the time point to be measured is expressed as:

[0075]

[0076] in, represents the degree of purification measured at the tth time, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th sensor, and the degree of purification at the i-th measured time point is obtained through the j-th sensor to construct the purification degree fingerprint library

[0077]

[0078] Where m is the number of sensors, n is the number of time series points, and the i-th row vector ψ i =[φ i1 ,φ i2 ,...,φ im ], which is the fingerprint feature value of the measured point measured by m sensors. Different signals are measured by different types of sensors to form a multi-dimensional feature dimension. The information measured by each sensor at the measured time point is the degree of hydrogen purification.

[0079] In a specific embodiment, the pressure control parameters are set according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process to adjust the pressure in the adsorption tower in real time. The specific steps include:

[0080] S1: Construct a GRNN network, use the Parzen window to simulate the probability distribution function of the GRNN network, and use the Holdout method to optimize the smoothing factor of the GRNN network to obtain a pressure regulation model;

[0081] S2: Iteratively train the pressure regulation model through the purification degree fingerprint library and the adsorption characteristics of the adsorbent to obtain the optimal pressure regulation model;

[0082] S3: Input the real-time data into the optimal pressure regulation model to obtain the optimal pressure.

[0083] Specifically, the GRNN network consists of an input layer, a pattern layer, a summation layer and an output layer;

[0084] The number of neurons in the input layer is the dimension D of the input vector;

[0085] The activation function output of the pattern layer is p i =exp(-D n 2 / 2σ 2 ), is the square of the Mahalanobis distance between the training data and the prediction point, and σ is the smoothness parameter;

[0086] There are two types of neurons in the summation layer. The activation function of the first type of neuron is The activation function of the second neuron is Among them, p i is the activation function output of the pattern layer, y ij is the weight parameter from the pattern layer to the summation layer, n is the number of nodes in the pattern layer, and the outputs of the two types of neurons in the summation layer are used as the denominator and numerator respectively;

[0087] The activation function output of the output layer is y j =s Nj / s D .

[0088] Specifically, the GRNN network adopts the Parzen window to simulate the probability distribution function, which includes:

[0089] Construct the GRNN network and define X n is the purification degree fingerprint library; by measuring the value X = {X n} for the dependent variable Y = {Y n} to predict the value;

[0090] Define the probability distribution function of the random variable vector X and the random variable scalar Y as f(X,Y), and output the conditional expectation E[Y|X] of the prediction result Y for the given measurement value X;

[0091] The non-parametric estimation method of Parzen window is used to simulate the probability distribution function f(X,Y) and output the Parzen window probability density function;

[0092] The Gaussian kernel function is used to optimize the Parzen window density function, and the optimized GRNN network is obtained by inverting the conditional expectation E[Y|X] through the optimized Parzen window density function.

[0093] Specifically, the optimized GRNN network is g(x):

[0094]

[0095] Where x is the new measurement value for testing, is the square of the Mahalanobis distance between the training data and the prediction point, n is the summation calculation sequence, σ is the smoothness parameter, and N is the number of samples.

[0096] Specifically, the Holdout method is used to optimize the smoothing factor of the GRNN network to obtain the pressure regulation model, including: after initializing the smoothing factor, a group of samples are selected and removed from the training set in each iteration process to be used as a test set, the difference between the output and the original label is calculated using the mean square error MSE, and the minimum smoothing factor is selected:

[0097]

[0098] Where N is the number of samples, y i and are the verification sample values ​​and GRNN output values ​​respectively.

[0099] Specifically, the steps of obtaining the optimal pressure regulation model are: using the purification degree fingerprint library and the adsorption characteristics of the adsorbent as the training set of the pressure regulation model, using the pressure corresponding to the generated relevant signal as the target output, using the multiple signal quantities at the time point to be tested and the characteristic vector of the adsorption characteristics of the adsorbent as the characteristic values ​​of the pressure regulation model, generating changes in signal intensity based on time changes, and mapping the relationship between signal intensity and pressure through the GRNN network.

[0100] In a specific embodiment, the closed-loop feedback adjustment of the parameters formed by comparing the actual purity of the purified hydrogen with the preset purity standard includes: adaptive dynamic adjustment: according to the results of the timing analysis, a threshold is set, and when the weight of a characteristic parameter exceeds the threshold, its weight is automatically adjusted. Adaptive dynamic adjustment of weights is an intelligent decision-making method based on data analysis results, which aims to automatically optimize system performance according to changes in real-time or historical data. In an embodiment of the present invention, the weights of the characteristic parameters are analyzed in a timing manner using appropriate methods (such as trend analysis, cyclic fluctuation analysis, etc.) to understand their changes over time. According to the results of the timing analysis, for each time point, combined with the purification requirements and system goals, one or more thresholds are set for each characteristic parameter. These thresholds can be absolute values ​​or percentage changes relative to historical data.

[0101] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0102] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water, characterized in that: include: Hydrogen purification unit, pressure optimization system and control unit; The hydrogen purification unit comprises a plurality of adsorption towers arranged in parallel, each of which is filled with an adsorbent for adsorbing and removing impurities in the hydrogen; The pressure optimization system includes a pressure sensor, a pressure regulating valve and a controller, wherein the pressure sensor is used to monitor the pressure in the adsorption tower in real time, the pressure regulating valve regulates the pressure according to the instruction of the controller, and the controller is used to set the pressure control parameters according to the adsorption characteristics of the adsorbent and the requirements of the hydrogen purification process to regulate the pressure in the adsorption tower in real time; The control unit is used to coordinate the temperature and flow of the pressure optimization system, and to form a closed-loop feedback adjustment of the parameters by comparing the actual purity of the purified hydrogen with the preset purity standard.

2. A hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water according to claim 1, characterized in that: Regulating valves are arranged between the adsorption towers arranged in parallel, and the adsorption towers are interconnected through pipelines to form an integrated purification system.

3. A hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water according to claim 1, characterized in that: It also includes a temperature sensor, which is installed in the hydrogen purification unit and is used to collect the real-time temperature inside the adsorption tower.

4. A hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water according to claim 3, characterized in that: It also includes a flow valve, which is placed in the pipeline between the adsorption towers and detects flow data through the flow valve.

5. A control strategy for a hydrogen purification pressure optimization system for hydrogen production by water electrolysis, characterized in that: include: Obtain pressure data, temperature data and flow data of the purification process; Perform feature extraction according to the hydrogen purification process and determine the key feature parameters corresponding to the purification process; Determining the influence index of the key characteristic parameters on hydrogen purification; Determine the target characteristic parameter whose influence index is greater than the influence index threshold; The target characteristic parameters are input into a pre-trained deep convolutional neural network to obtain the degree of purification of the hydrogen output by the pre-trained deep convolutional neural network.

6. A control strategy for a hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water according to claim 5, characterized in that: After determining that the influence index is greater than the target characteristic parameter of the influence index threshold, the method further includes: The neural network pre-trained with the standard dataset is trained on the collected historical dataset; Use the trained neural network to predict the target characteristic parameters of the degree of purification, obtain the characteristic vectors of the target characteristic parameters in the time series, and construct the target characteristic parameter matrix; The coupling vector between the target characteristic parameters, i.e., the spatial characteristics of hydrogen purification, is calculated through the obtained target characteristic parameter matrix; Using the spatial features of hydrogen purification as input to a pre-trained deep convolutional neural network; The degree of purification of the hydrogen is obtained through the mapping relationship between the degree of purification of the hydrogen and its spatial characteristics.

7. A control strategy for a hydrogen purification pressure optimization system for producing hydrogen from water electrolysis according to claim 6, characterized in that: Before obtaining the characteristic vector of the target characteristic parameter in the time series and constructing the target characteristic parameter matrix, the method further includes: performing standardization processing on the target characteristic parameter to map the value of the target characteristic parameter to a preset value range.

8. A control strategy for a hydrogen purification pressure optimization system for producing hydrogen by electrolysis of water according to claim 5, characterized in that: Determining the influence index of the key characteristic parameter on hydrogen purification includes: determining the correlation coefficient between the key characteristic parameter and hydrogen purification, and the contribution rate of the key characteristic parameter to the hydrogen purification; and taking the product of the contribution rate and the correlation coefficient as the influence index of the hydrogen purification.

9. A control strategy for a hydrogen purification pressure optimization system for producing hydrogen from water electrolysis according to claim 6, characterized in that: The constructing target feature parameter matrix comprises: The combined signal formed by the pressure signal, temperature signal and flow signal indicates the degree of purification under the time series. The signal intensity of different time series is collected, and the degree of purification at the time point to be measured is expressed as: in, represents the degree of purification measured at the tth time, k represents the number of measurements, i represents the i-th measured time point, j represents the j-th sensor, and the degree of purification at the i-th measured time point is obtained through the j-th sensor to construct a purification degree fingerprint library Where m is the number of sensors, n is the number of time series points, and the i-th row vector ψ i =[φ i1 ,φ i2 ,...,φ im ], which is the fingerprint feature value of the measured point measured by m sensors. Different signals are measured by different types of sensors to form a multi-dimensional feature dimension. The information measured by each sensor at the measured time point is the degree of purification of hydrogen.

Images