System and method for electrochemical synthesis of calcium hydroxide based on machine learning assisted her / hor coupling
Patent Information
- Application Number
- CN202610706568.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-21
- Publication Date
- 2026-08-28
AI Technical Summary
但该现有技术存在致命缺点：依赖常规水电解反应产生pH梯度，导致电压需求极高（如在仅 6 mA 的微小电流下工作电压就高达 2.5V），将其规模化应用于工业制造的电能成本不可估量
[0025](1) Extremely low operating voltage and ultra-high current efficiency: This invention couples the hydrogen evolution reaction and the hydrogen oxidation reaction, so that even under a large current of 150 mA (25 times the current in the literature), the total operating voltage can still be stably maintained at an extremely low level of about 1.70 V. At the same time, the calcium carbonate consumption after 3 hours of reaction is close to the theoretical value, achieving an ultra-high current efficiency of nearly 100%.
Smart Images
Figure CN122648972A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of calcium hydroxide synthesis technology, and specifically to a system and method for the electrochemical synthesis of calcium hydroxide based on machine learning-assisted HER / HOR coupling. Background Technology
[0002] Silicate cement is the most widely used building material globally. Traditional processes heavily rely on the high-temperature calcination of limestone into cement precursor Ca(OH)2 in a rotary kiln at approximately 1500°C. Of this, 40% of carbon emissions originate from the decomposition of CaCO3, and another 40% from the combustion of fossil fuels. While alternatives such as high-belite cement or magnesium-based cement have been proposed, they face practical application drawbacks such as slower early strength development and instability when exposed to water.
[0003] Although Chiang et al. pioneered the idea of using an electrochemical pH gradient to dissolve calcium carbonate at room temperature to prepare Ca(OH)2, achieving decarbonization (see Leah D. Ellis, Andres F. Badel, Miki L. Chiang, Richard J.-Y. Park, and Yet-Ming Chiang. Toward electrochemicalsynthesis of cement-an electrolyzer-based process for decarbonating CaCO3 while producing useful gas streams. Proceedings of the National Academy of Sciences, 117(23):12584-12591, 2020), this existing technology has a fatal flaw: it relies on conventional water electrolysis to generate a pH gradient, resulting in extremely high voltage requirements (e.g., the operating voltage is as high as 2.5V at a tiny current of only 6 mA), making the energy cost of scaling it up for industrial manufacturing incalculable.
[0004] Furthermore, multidimensional parameter optimization relies on costly trial-and-error methods: current research on electrochemical synthesis of cement precursors lacks standardized datasets. Faced with extremely complex networks of variables such as current density, electrode activity, electrolyte concentration in each chamber, and gas-liquid flow rate, researchers can only rely on time-consuming trial-and-error methods, which greatly reduces research efficiency and increases R&D costs.
[0005] Therefore, how to reduce the energy consumption of electrochemical synthesis of calcium hydroxide and optimize the parameters of the electrochemical system remains an urgent technical problem to be solved. Summary of the Invention
[0006] Due to the aforementioned deficiencies in existing technologies, this invention provides a system and method for the electrochemical synthesis of calcium hydroxide based on machine learning-assisted HER / HOR coupling. This aims to overcome the bottlenecks in existing electrochemical decarbonization technologies for calcium hydroxide synthesis, which rely on traditional water splitting reactions, resulting in excessively high electrolysis voltage and energy consumption. Furthermore, this invention introduces an integrated deep learning model with a multilayer perceptron (MLP) as the framework and a physical information neural network (PINN) as the physical enhancement, addressing the challenge of manually co-controlling and optimizing multidimensional parameters (operating current density, electrolyte concentration, gas-liquid flow rate, etc.) in complex multi-compartment electrolysis systems in the electrochemical and chemical engineering field.
[0007] To achieve the above objectives, on the one hand, the present invention provides a system for the HER / HOR coupled electrochemical synthesis of calcium hydroxide based on machine learning assistance, comprising gas diffusion electrodes as anode and cathode respectively, wherein a cation exchange membrane and an anion exchange membrane are sequentially arranged between the anode and cathode, dividing the electrochemical system into three chambers: an anode chamber, an intermediate chamber, and a cathode chamber; the anode chamber is filled with CaCO3;
[0008] Hydrogen evolution occurs in the anode chamber, and hydrogen oxidation occurs in the cathode chamber; the Ca in the anolyte... 2+ OH- in the cathode electrolyte passes through the cation exchange membrane. − Passing through the anion exchange membrane, the two converge in the intermediate chamber to form calcium hydroxide precipitate;
[0009] Sensitivity analysis of system process characteristic parameters, including operating current, electrolyte concentration, and hydrogen flow rate, is performed using an integrated deep learning model based on MLP as the framework and PINN as the physical enhancement. A regularized linear regression algorithm based on response surface methodology is introduced to confirm the optimal combination of system process parameters, including operating current, electrolyte concentration, and hydrogen flow rate. The integrated deep learning model predicts the yield of calcium hydroxide based on the input multidimensional process characteristic data.
[0010] The above technical solution proposes an innovative system combining low-energy hardware and machine learning software, achieving the following breakthrough improvements: 1. Coupled HER / HOR reaction, significantly reducing operating voltage: This invention optimizes the original electrochemical system by coupling the hydrogen evolution reaction (HER) and hydrogen oxidation reaction (HOR) as electrode reactions. Under experimental conditions of 150 mA (25 times the literature current conditions), the total voltage of this system can still be stably maintained at an extremely low level of approximately 1.70 V. 2. High-purity CO2 by-product, aiding carbon capture: The carbon dioxide released at the anode in this process has extremely high purity, enabling direct and simple collection. This provides a pure CO2 gas stream for subsequent carbon capture, utilization, and storage, representing a significant advantage in reducing carbon emissions from cement production. 3. Introducing an integrated deep learning model based on MLP as the framework and PINN as the physical enhancement to deeply mine limiting factors and predict yields can significantly reduce the trial-and-error costs of process parameters.
[0011] Furthermore, the gas diffusion electrode is a carbon cloth supported on a Pt-C catalyst; the anode supplies hydrogen-containing gas, and the cathode collects the hydrogen-containing gas for recycling back to the anode. This enables in-situ self-recycling of hydrogen: the hydrogen directly generated at the cathode can be recovered to the anode for reuse, acting similarly to a catalyst (i.e., participating in the electrode reaction without net consumption).
[0012] Furthermore, the multilayer perceptron includes an input layer, a hidden layer, and an output layer. It extracts multidimensional process features, including operating current, electrolyte concentration, and hydrogen flow rate, to predict the calcium hydroxide formation rate and quantitatively identify the influence weights of system process feature parameters on the calcium hydroxide formation rate. The physical information neural network reuses the network layer structure of the multilayer perceptron, incorporating physical constraint loss terms, including Faraday's law constraints, water ion product constraints, solubility product constraints, and anode pH stability constraints, into the model's global loss function. In this way, the model training process automatically penalizes predictions that violate physicochemical laws, ensuring that the output conforms to electrochemical principles.
[0013] Furthermore, the total loss function of the integrated deep learning model is calculated as: data fitting loss + weight coefficients × physical constraint loss. The model is trained to minimize the total loss, simultaneously satisfying both data fitting accuracy and physicochemical laws, thereby achieving accurate prediction of Ca(OH)2 generation rate, real-time optimization of process parameters, and identification of system performance limiting factors.
[0014] Furthermore, the data fitting loss is the mean squared error loss function, and the weighting coefficient is 0.1.
[0015] Furthermore, the input dimension of the multilayer perceptron (MLP) is 12-dimensional features, including: operating current, electrolyte concentration, hydrogen flow rate, anode potential, cathode potential, cation exchange membrane voltage, anion exchange membrane voltage, total voltage, initial pH, pH of the anode electrolyte after reaction, pH of the intermediate chamber electrolyte after reaction, and pH of the cathode electrolyte after reaction. The high-dimensional feature fitting capability of the MLP can handle multi-parameter coupled inputs.
[0016] Furthermore, the hidden layer of the multilayer perceptron consists of two layers: the first layer includes 16 neurons, and the activation function is the ReLU function; the second layer includes 4 neurons, and the activation function is the ReLU function.
[0017] On the other hand, the present invention provides a method for synthesizing calcium hydroxide using machine learning-assisted HER / HOR coupled electrochemical methods, employing the system described above, and including the following steps:
[0018] S1. Construct a HER / HOR coupled three-chamber electrochemical system and the integrated deep learning model;
[0019] S2. Train the integrated deep learning model using the model global loss function: The model global loss function includes a data fitting loss term and a physical constraint loss term including Faraday law constraints, water ion product constraints, solubility product constraints, and anodic pH stability constraints;
[0020] S3. Optimize the system process characteristic parameter boundaries, including operating current, electrolyte concentration, and hydrogen flow rate, through the trained integrated deep learning model.
[0021] S4. Calcium hydroxide is synthesized using optimized process characteristic parameters.
[0022] Furthermore, in the HER / HOR coupled three-chamber electrochemical system, the anode supplies hydrogen-containing gas, and the cathode collects hydrogen-containing gas for recycling at the anode.
[0023] Furthermore, the collected calcium hydroxide is first placed in a vacuum oven for vacuum treatment, utilizing the intense vacuum boiling of internal moisture to break down the flocculent structure, followed by a rinsing and drying process. Since the generated Ca(OH)₂ flocculent precipitate readily encapsulates electrolyte salts (which, if not removed, can lead to alkali-silica gel reaction inside the cement or corrosion of the reinforcing steel) and is difficult to remove by simple water washing, the above-mentioned technical solution makes the subsequent rinsing and desalination process extremely efficient and easy.
[0024] Compared with the prior art, the above invention has the following advantages or beneficial effects:
[0025] (1) Extremely low operating voltage and ultra-high current efficiency: This invention couples the hydrogen evolution reaction and the hydrogen oxidation reaction, so that even under a large current of 150 mA (25 times the current in the literature), the total operating voltage can still be stably maintained at an extremely low level of about 1.70 V. At the same time, the calcium carbonate consumption after 3 hours of reaction is close to the theoretical value, achieving an ultra-high current efficiency of nearly 100%.
[0026] (2) Achieving in-situ hydrogen recycling and zero net consumption: This invention not only reduces energy consumption but also cleverly designs a self-circulating hydrogen loop. The H2 generated at the cathode can theoretically be directly collected and recycled to the anode for reuse. Throughout the electrochemical reaction, hydrogen acts as an electron mediator similar to a catalyst, participating in the electrode reaction without net consumption, greatly reducing operating costs.
[0027] (3) High-purity CO2 by-product, which is beneficial for direct carbon capture and storage (CCUS): CO2 released by traditional calcination methods is often mixed with exhaust gas from fuel combustion, making it extremely difficult to separate. The CO2 released by the ambient temperature electrochemical decarbonization process of this invention has extremely high purity and can be collected directly and simply. This provides a high-quality pure gas stream for subsequent carbon capture, utilization and storage, which is a huge advantage for truly realizing decarbonization in the cement industry.
[0028] (4) Facing a complex parameter network encompassing multiple dimensions such as current density, electrolyte concentration, and gas-liquid flow rate, this invention innovatively introduces computational methods into experimental analysis. By constructing a multilayer perceptron and a physical information neural network, and incorporating Faraday's law and K... sp K w By incorporating stringent physical constraints into the loss function, this system can not only predict reaction yields but also accurately identify the system's main limiting factors through feature sensitivity analysis. This hardware-software hybrid architecture significantly reduces the high trial-and-error costs during industrial scale-up and greatly improves R&D efficiency.
[0029] (5) Innovative vacuum flash boiling desalination post-treatment process: Addressing the engineering challenge that the generated Ca(OH)2 flocculent precipitate easily encapsulates electrolyte salts (which, if not removed, can lead to alkali-silica gel reaction inside the cement or steel corrosion) and is difficult to remove by simple water washing, this invention proposes a highly practical physical destruction method. The precipitate is placed in a vacuum oven for only tens of seconds, utilizing the intense vacuum boiling of the internal moisture to instantly disintegrate the flocculent structure, thus making the subsequent rinsing and desalination process efficient and easy. Attached Figure Description
[0030] The invention, its features and advantages will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.
[0031] Figure 1 This is a schematic diagram of the electrochemical system design for synthesizing Ca(OH)2 in one embodiment of the present invention;
[0032] Figure 2 This is a structural diagram of a multilayer perceptron (including an input layer, a hidden layer, and an output layer) in one embodiment of the present invention.
[0033] Figure 3 This is a feature sensitivity heatmap under different current levels based on sensitivity analysis in one embodiment of the present invention: the darker the color, the greater the impact on the prediction result;
[0034] Figure 4 The time-domain electrochemical performance of the electrochemical system for synthesizing Ca(OH)2 in one embodiment of the present invention is shown.
[0035] Figure 5 This is a step-by-step observation diagram (from left to right) of the calcium hydroxide precipitation process in the intermediate chamber in one embodiment of the present invention. Detailed Implementation
[0036] The structure of the present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the invention. Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It will also be understood that, unless expressly defined herein, terms such as those defined in a general dictionary shall be interpreted as having the meaning consistent with their meaning in the relevant field context, and not as having an ideal or overly formal meaning.
[0037] See Figure 1 This invention discloses a machine learning-assisted HER / HOR coupled electrochemical synthesis system for calcium hydroxide, comprising gas diffusion electrodes as anode and cathode, respectively, with a cation exchange membrane and an anion exchange membrane sequentially disposed between the anode and cathode, dividing the electrochemical system into three chambers: an anode chamber, an intermediate chamber, and a cathode chamber; CaCO3 is introduced into the anode chamber; a hydrogen evolution reaction occurs in the anode chamber, and a hydrogen oxidation reaction occurs in the cathode chamber; the Ca in the anode electrolyte... 2+ OH- in the cathode electrolyte passes through the cation exchange membrane. −Passing through the anion exchange membrane, the two molecules converge in the intermediate chamber to form calcium hydroxide precipitate. Sensitivity analysis of system process characteristic parameters, including operating current, electrolyte concentration, and hydrogen flow rate, is performed using an integrated deep learning model based on a multilayer perceptron as the framework and a physical information neural network as the physical enhancement. A regularized linear regression algorithm based on response surface methodology is introduced to confirm the optimal combination of system process parameters, including operating current, electrolyte concentration, and hydrogen flow rate. The integrated deep learning model also predicts the calcium hydroxide yield based on the input multidimensional process characteristic data.
[0038] This invention innovatively employs a three-chamber electrolytic cell architecture, coupling HER and HOR as electrode reactions, replacing the energy-intensive water splitting (oxygen evolution) reaction. HOR occurs at the anode: hydrogen gas is oxidized to protons at the gas diffusion electrode (GDE), i.e., H2 → 2H+. + + 2e - H + Diffusion enters the solution and dissolves CaCO3, releasing Ca 2+ And high-purity, easily captured CO2 gas. HER occurs at the cathode: a water reduction reaction produces hydroxide ions (OH-). − ) and hydrogen gas, i.e., 2H2O + 2e - → 2OH - + H2↑. Precipitation reaction: Kinetically, Ca... 2+ OH passes through the cation exchange membrane (CEM). - It passes through the anion exchange membrane (AEM), collects in the intermediate chamber, and forms a Ca(OH)2 precipitate.
[0039] The gas diffusion electrode is exemplified as carbon cloth supported on a Pt-C catalyst; it is understood that it could also be a porous conductive electrode supported on other HER / HOR reaction catalysts.
[0040] The anode supplies hydrogen-containing gas, and the cathode collects the hydrogen-containing gas for recycling back to the anode. The entire process involves no net hydrogen consumption; hydrogen merely acts as an electron mediator, similar to a catalyst. An example of hydrogen-containing gas is Ar-8 mol% H2 gas; it is understood that other hydrogen-containing industrial waste gases that do not poison the electrode catalyst can also be used.
[0041] The multilayer perceptron includes an input layer, a hidden layer, and an output layer. It extracts multidimensional process features, including operating current, electrolyte concentration, and hydrogen flow rate, to predict the calcium hydroxide formation rate and quantitatively identify the influence weights of system process feature parameters on the calcium hydroxide formation rate. The physical information neural network reuses the network layer structure of the multilayer perceptron, incorporating physical constraint loss terms, including Faraday's law constraints, water ion product constraints, solubility product constraints, and anode pH stability constraints, into the model's global loss function. The total loss function of the integrated deep learning model = data fitting loss + weight coefficients × physical constraint loss. The model is trained to minimize the total loss, simultaneously satisfying data fitting accuracy and physicochemical laws, achieving accurate prediction of Ca(OH)2 formation rate, real-time optimization of process parameters, and identification of system performance limiting factors.
[0042] The following examples illustrate many specific details to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that well-known algorithms and models (e.g., regularized linear regression algorithms) are not shown in detail to avoid obscuring the gist of the invention; and that the techniques not detailed in the following examples are readily available prior art.
[0043] Example
[0044] This embodiment provides a system and method for the electrochemical synthesis of calcium hydroxide based on machine learning-assisted HER / HOR coupling.
[0045] 1. Construction of the electrochemical device: See Figure 1 The electrochemical device uses LANCYTOM® CT-4 as the CEM and AHT as the AEM to separate the anode, intermediate chamber, and cathode chamber into a three-chamber structure, each with a volume of approximately 20 mL. GDEs prepared using carbon cloth supported on a Pt-C catalyst serve as the anode and cathode.
[0046] 2. Build and train an integrated deep learning model with a multilayer perceptron as the basic framework and a physical information neural network as the physical enhancement.
[0047] See Figure 2 The specific architecture and working principle of the Multilayer Perceptron (MLP):
[0048] (1) Model Architecture
[0049] Input layer: The input dimension consists of 12 features, including: operating current, electrolyte concentration, hydrogen flow rate, anode potential, cathode potential, CEM membrane voltage, AEM membrane voltage, total voltage, initial pH, pH of the anode electrolyte after reaction, pH of the intermediate chamber electrolyte after reaction, and pH of the cathode electrolyte after reaction.
[0050] Hidden layers: 2 fully connected layers in total
[0051] First layer: 16 neurons, using ReLU activation function.
[0052] Second layer: 4 neurons, using ReLU activation function.
[0053] Output layer: 1 neuron, outputting the Ca(OH)2 generation rate (mg / h).
[0054] (2) Working principle
[0055] Forward propagation: Input process parameters into the network, perform weighted summation and activation transformation in the hidden layer, and output the predicted yield;
[0056] Backpropagation: Iteratively optimizes network weights and biases with the goal of minimizing the loss function;
[0057] Model output: Under given conditions such as current, electrolyte concentration, and gas flow rate, the model can quickly and stably predict the calcium hydroxide formation rate.
[0058] MLP provides a complete network structure, which is used to perform feature extraction, numerical fitting and yield prediction functions. It includes an input layer, hidden layers and an output layer, and is the basic architecture of the entire neural network.
[0059] The specific architecture and working principle of Physical Information Neural Network (PINN):
[0060] (1) Model Architecture
[0061] Network layer: Fully reuses the input layer, 2 hidden layers, and output layer structure of the above MLP;
[0062] Physical constraint module: Embedded in the global loss function, containing 4 core physicochemical constraints:
[0063] Faraday's law constraint: the predicted yield shall not exceed the theoretical electrochemical maximum value;
[0064] Water ion product constraint: In each compartment of the solution, [H+] + ]×[OH - ]=K w ;
[0065] Solubility product constraint: Ca 2+ With OH - The concentration product satisfies the K of Ca(OH)2 sp ;
[0066] Anode pH stability constraint: The anode pH is maintained within a reasonable reaction range.
[0067] Loss function: Total loss = Data fitting loss (mean squared error) + 0.1 × Physical constraint loss.
[0068] (2) Working principle
[0069] It inherits the high-dimensional feature fitting capability of MLP and handles multi-parameter coupled inputs;
[0070] The training process automatically penalizes predictions that violate physical and chemical laws, ensuring that the output conforms to electrochemical principles;
[0071] It supports feature sensitivity analysis to quantitatively identify the influence weights of factors such as current, electrolyte concentration, and membrane voltage on system performance;
[0072] It enables accurate yield prediction, process parameter boundary optimization, and identification of key system limiting factors.
[0073] PINN is a feature of MLP networks, indicating that prior physical knowledge is introduced into the network's output during training to improve the physical interpretability of the output. Coupling is achieved by embedding physical constraints in the loss function layer. The coupling method is as follows: preserving the data fitting ability of MLPs, and forcibly incorporating Faraday's law and Kinematics into the global loss function. sp K w The physical and chemical equations constrain the model training, ensuring both accurate data fitting and adherence to physical laws. This loss function is used for training both the multilayer perceptron and the physical information neural network, serving as a shared core constraint for both.
[0074] The workflow coupling first involves the MLP (Multilayer Perceptron) learning and preliminary prediction of process parameters. Then, the physical laws coupled within the loss function correct the results for physical compliance, outputting reliable predicted values that conform to electrochemical principles. This also achieves parameter optimization and limiting factor identification. This physically constrained multilayer perceptron falls under the category of physical information neural networks. (See also...) Figure 3 Based on the response surface methodology and a regularized linear regression algorithm, sensitivity analysis of feature parameters was performed to ultimately determine the optimal combination of system process parameters, including operating current, electrolyte concentration, and hydrogen flow rate. This deep learning model can identify the main system limiting factors affecting prediction results through feature sensitivity analysis, thereby potentially significantly reducing traditional trial-and-error costs and improving the efficiency of research and industrial scale-up.
[0075] 3. Optimize the system process characteristic parameter boundaries, including operating current, electrolyte concentration, and hydrogen flow rate, using the trained integrated deep learning model; finally, synthesize calcium hydroxide using the optimized process characteristic parameters.
[0076] The operating conditions in this embodiment are as follows: 1.5 mol / L NaCl solution is used as the electrolyte, and Ar-8 mol% H2 gas is supplied to the hydrophobic side of the anode GDE at a flow rate of 40 ml / min via an electronic mass flow meter. A peristaltic pump is used to circulate the liquid within the anode chamber at a flow rate of 50 ml / min.
[0077] See Figure 4 Under a constant high current of up to 150 mA, the total system voltage remained stably low at ~1.70 V (25 times the current reported in the literature, yet the voltage dropped significantly), and the current efficiency approached 100% after 3 hours of reaction. See also Figure 5 When approximately 20 mL of the anode chamber solution is added dropwise to an equal volume of the cathode side solution, a white flocculent precipitate gradually forms, eventually giving the mixture a milky white appearance.
[0078] In some embodiments, the cathode collects hydrogen-containing gas to supply the anode for recycling. The collected calcium hydroxide is first placed in a vacuum oven for vacuum treatment, where the intense vacuum boiling of the internal moisture breaks down the flocculent structure, followed by a rinsing and drying process. Since the generated Ca(OH)2 flocculent precipitate easily encapsulates electrolyte salts (which, if not removed, can lead to alkali-silica gel reaction inside the cement or steel corrosion) and is difficult to remove by simple water washing, the above-mentioned technical solution makes the subsequent rinsing and desalination process extremely efficient and easy.
[0079] Those skilled in the art should understand that variations can be implemented by combining existing technology with the above embodiments, which will not be elaborated here. Such variations do not affect the essence of the present invention, and will not be elaborated here either.
[0080] The preferred embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and the devices and structures not described in detail should be understood as being implemented in a conventional manner in the art. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the present invention. This does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the present invention's technical solutions still fall within the protection scope of the present invention.
Claims
1. A system for the HER / HOR coupled electrochemical synthesis of calcium hydroxide based on machine learning assistance, comprising gas diffusion electrodes as anode and cathode respectively, characterized in that, A cation exchange membrane and an anion exchange membrane are sequentially arranged between the anode and the cathode, dividing the electrochemical system into three chambers: an anode chamber, an intermediate chamber, and a cathode chamber; CaCO3 is introduced into the anode chamber. Hydrogen evolution occurs in the anode chamber, and hydrogen oxidation occurs in the cathode chamber; the Ca in the anolyte... 2+ OH- in the cathode electrolyte passes through the cation exchange membrane. - Passing through the anion exchange membrane, the two converge in the intermediate chamber to form calcium hydroxide precipitate; A sensitivity analysis of system process characteristic parameters, including operating current, electrolyte concentration, and hydrogen flow rate, is performed using an integrated deep learning model based on a multilayer perceptron as the framework and a physical information neural network as the physical enhancement. A regularized linear regression algorithm based on response surface methodology is introduced to confirm the optimal combination of system process parameters, including operating current, electrolyte concentration, and hydrogen flow rate. The integrated deep learning model also predicts the yield of calcium hydroxide based on the input multidimensional process characteristic data.
2. The system according to claim 1, characterized in that, The gas diffusion electrode is a carbon cloth supported on a Pt-C catalyst; the anode supplies hydrogen-containing gas, and the cathode collects hydrogen-containing gas to be recycled back to the anode.
3. The system according to claim 1, characterized in that, The multilayer perceptron includes an input layer, a hidden layer, and an output layer. It extracts multidimensional process features, including operating current, electrolyte concentration, and hydrogen flow rate, to predict the calcium hydroxide formation rate and quantitatively identify the influence weights of system process feature parameters on the calcium hydroxide formation rate. The physical information neural network reuses the network layer structure of the multilayer perceptron and introduces physical constraint loss terms, including Faraday law constraints, water ion product constraints, solubility product constraints, and anode pH stability constraints, into the model's global loss function.
4. The system according to claim 3, characterized in that, The total loss function of the integrated deep learning model = data fitting loss + weight coefficient × physical constraint loss.
5. The system according to claim 4, characterized in that, The data fitting loss is the mean squared error loss function, and the weighting coefficient is 0.
1.
6. The system according to claim 3, characterized in that, The input dimension of the multilayer sensor is 12-dimensional features, including: operating current, electrolyte concentration, hydrogen flow rate, anode potential, cathode potential, cation exchange membrane voltage, anion exchange membrane voltage, total voltage, initial pH, pH of the anode electrolyte after reaction, pH of the intermediate chamber electrolyte after reaction, and pH of the cathode electrolyte after reaction.
7. The system according to claim 3, characterized in that, The hidden layers of the multilayer perceptron consist of two layers: the first layer includes 16 neurons, with the ReLU function as the activation function; the second layer includes 4 neurons, with the ReLU function as the activation function.
8. A method for the electrochemical synthesis of calcium hydroxide based on machine learning-assisted HER / HOR coupling, characterized in that, The synthesis of calcium hydroxide using the system described in any one of claims 1 to 7 includes the following steps: S1. Construct a HER / HOR coupled three-chamber electrochemical system and the integrated deep learning model; S2. Train the integrated deep learning model using the model global loss function: The model global loss function includes a data fitting loss term and a physical constraint loss term including Faraday law constraints, water ion product constraints, solubility product constraints, and anodic pH stability constraints; S3. Optimize the system process characteristic parameter boundaries, including operating current, electrolyte concentration, and hydrogen flow rate, through the trained integrated deep learning model. S4. Calcium hydroxide is synthesized using optimized process characteristic parameters.
9. The method according to claim 8, characterized in that, The HER / HOR coupled three-chamber electrochemical system supplies hydrogen-containing gas to the anode and collects hydrogen-containing gas at the cathode for recycling to the anode.
10. The method according to claim 8, characterized in that, The collected calcium hydroxide is first placed in a vacuum oven for vacuum treatment, where the intense vacuum boiling of the internal moisture breaks down the flocculent structure, followed by rinsing and drying.