Experimental control method and system for development of electrochemical catalyst
Through automated experimental control systems and intelligent optimization algorithms, efficient and accurate screening of electrochemical catalyst development is achieved, and the problems of low efficiency and insufficient accuracy in traditional methods are solved, which significantly improves the efficiency and reliability of catalyst development.
Patent Information
- Application Number
- CN202510576973.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-06
- Publication Date
- 2025-09-02
AI Technical Summary
Traditional electrochemical catalyst development efficiency is low, relying on manual trial and error methods, insufficient parameter space exploration capabilities, lack real-time feedback mechanisms, and existing intelligent systems are poor in special environments, making it difficult to improve development accuracy.
An automated experimental control system is adopted, combined with Bayesian optimization, SNOBFIT algorithm and neural network model, a closed-loop process of synthesis-test-optimization is realized. Through multi-dimensional stop conditions and automated cleaning modules, the experimental conditions are ensured and the data purity is purified, experimental parameters are generated dynamically, and the liquid distribution, reaction and electrochemical modules are integrated.
It significantly improves the efficiency and accuracy of electrochemical catalyst development, shortens the development cycle, reduces resource waste, improves the reliability and predictability of experimental results, and achieves efficient and accurate catalyst screening.
Smart Images
Figure CN120578054A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of automated experiments, and in particular to an experimental control method and system for electrochemical catalyst development. Background Art
[0002] Traditional electrochemical catalyst development relies on trial-and-error methods, which are limited by long experimental cycles, limited ability to explore parameter space, and a lack of real-time feedback mechanisms. Researchers must manually adjust synthesis conditions (such as metal component ratios, reaction temperature, and potential) and test performance offline. This results in a time-consuming and tedious high-dimensional parameter optimization process that is prone to falling into local optimal solutions.
[0003] In addition, although existing intelligent experimental systems (such as mobile robot platforms) have certain automation capabilities, their complex path planning, high-precision operation requirements and frequent sample transfer links significantly increase costs and time consumption, and are difficult to flexibly deploy in special environments (such as glove boxes), which seriously restricts their practical application in the research and development of electrochemical catalysts.
[0004] From the above, we can see that how to improve the development efficiency and accuracy of electrochemical catalysts remains to be solved. Summary of the Invention
[0005] In order to improve the development efficiency and accuracy of electrochemical catalysts, the present application provides an experimental control method and system for electrochemical catalyst development.
[0006] In a first aspect, the present application provides an experimental control method for electrochemical catalyst development, which adopts the following technical solution:
[0007] An experimental control method for electrochemical catalyst development, comprising:
[0008] Determining the corresponding experimental objectives and initial parameter space based on a predetermined experimental type, generating a first round of experimental combinations based on the initial parameter space, and automatically performing liquid preparation, reaction, sample synthesis, and electrochemical performance testing based on the initial parameter space to obtain corresponding first round experimental result data, wherein the initial parameter space includes metal ion concentration, voltage window, and reaction time;
[0009] Inputting the first round of experimental result data into an algorithm module for performing data analysis and modeling on the experimental results, obtaining the first round of analysis result data output by the algorithm module, dynamically generating a next round of experimental parameter plan based on the first round of analysis result data of the optimization algorithm subsystem, and driving the platform to execute the experiment again based on the next round of experimental parameter plan, wherein the optimization algorithm module includes a Bayesian optimization algorithm, a SNOBFIT algorithm, and a neural network prediction model;
[0010] Repeat the steps of obtaining the next round of experimental parameter plan, and determine whether the preset stop condition is met based on the new next round of experimental parameter plan. If not, proceed to the next round of experiments, otherwise stop the experiment. The experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module, and an electrochemical module.
[0011] Optionally, the stopping condition includes at least one of the following: obtaining performance difference data corresponding to two consecutive experiments, and judging whether the performance difference data is less than a set threshold; obtaining the total number of experiments, and judging whether the total number of experiments reaches a preset upper limit; obtaining a corresponding performance response surface graph based on the first round of analysis result data, and judging whether the gradient of the current optimal solution is less than a preset gradient threshold based on the performance response surface graph.
[0012] Optionally, the optimization algorithm module adopts a dynamic combination strategy of Bayesian optimization and SNOBFIT algorithm. The method also includes: in the global exploration stage, giving priority to using the Bayesian optimization algorithm for efficient search of the parameter space; in the local convergence stage, switching to the SNOBFIT algorithm to improve the accuracy of the local optimal solution. The switching condition for dynamic balanced exploration and convergence is that the performance standard deviation of the current experimental round is lower than a preset threshold.
[0013] Optionally, in the analysis process based on the experimental result data, the method also includes: generating a component-performance three-dimensional response surface diagram based on the experimental result data to visualize the relationship between different metal ratios and catalytic performance; drawing a sampling distribution diagram to display the coverage density of the current experimental parameters in real time; and outputting the formula parameters of the current optimal solution and the corresponding performance indicators.
[0014] Optionally, the method also includes: the metal ion concentration range in the initial parameter space is 0M / L to 1.0M / L, the voltage window range is -5.0V to +5.0V, the reaction time range is 10 seconds to 1800 seconds, and the metal ions include at least three elements of nickel, cobalt, iron, copper, manganese, molybdenum, bismuth, gold, platinum, palladium, ruthenium or rhodium.
[0015] Optionally, the method also includes: after each round of experiments, using deionized water and cleaning fluid to thoroughly clean the pipetting cleaning module, reaction module and electrodes; the cleaning process is automatically controlled by the host computer according to a preset program, and the cleaning time is dynamically adjusted according to the type and concentration of the solution used in the previous round of experiments; after the cleaning is completed, the system automatically generates a cleaning report, recording the amount of cleaning fluid used, cleaning time and cleaning effect evaluation indicators.
[0016] Optionally, the method further includes: the liquid preparation module realizes precise proportioning of the metal precursor liquid through a multi-channel liquid injection channel, and supports automatic generation of gradient concentration, wherein the step length of the gradient concentration is adjustable, and the minimum step length is 1% of the total concentration.
[0017] In a second aspect, the present application provides an experimental control system for electrochemical catalyst development, which adopts the following technical solution:
[0018] An experimental control system for electrochemical catalyst development, comprising:
[0019] a first-round experimental result data acquisition module, which determines the corresponding experimental objectives and initial parameter space based on a predetermined experimental type, generates a first-round experimental combination based on the initial parameter space, and automatically performs liquid preparation, reaction, sample synthesis, and electrochemical performance testing based on the initial parameter space to obtain the corresponding first-round experimental result data; wherein the initial parameter space includes metal ion concentration, voltage window, and reaction time;
[0020] A module for generating a parameter plan for the next round of experiments inputs the result data of the first round of experiments into an algorithm module for performing data analysis and modeling on the experimental results, obtains the first round analysis result data output by the algorithm module, and dynamically generates a parameter plan for the next round of experiments based on the first round analysis result data of the optimization algorithm subsystem, and drives the platform to execute the experiment again based on the parameter plan for the next round of experiments. The optimization algorithm module includes a Bayesian optimization algorithm, a SNOBFIT algorithm, and a neural network prediction model;
[0021] The repeated execution module is used to repeatedly execute the steps of obtaining the parameter plan for the next round of experiments, and judge whether the preset stop conditions are met based on the new parameter plan for the next round of experiments. If not, the next round of experiments is carried out; otherwise, the experiment is stopped. The experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module and an electrochemical module.
[0022] In a third aspect, the present application provides an experimental control system for electrochemical catalyst development, which adopts the following technical solution:
[0023] An experimental control system for electrochemical catalyst development includes a processor in which a program of any one of the above-mentioned experimental control methods for electrochemical catalyst development runs.
[0024] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:
[0025] A storage medium stores a program of any one of the above-mentioned experimental control methods for electrochemical catalyst development.
[0026] In summary, this application includes at least one of the following beneficial technical effects:
[0027] First, by integrating automated experimental hardware and intelligent optimization algorithms, the efficiency of electrochemical catalyst development has been significantly improved. The core lies in the closed-loop process of "synthesis-testing-optimization": the experimental execution subsystems (liquid preparation, reaction, electrochemical module) are seamlessly connected, avoiding the inefficient links of manual operation, sample transfer and offline testing in traditional methods. For example, in the screening of water electrolysis catalysts, the full process experiment of 66 sets of formulas can be completed within 90 hours. The optimization algorithm subsystem (Bayesian optimization, SNOBFIT, etc.) dynamically generates experimental parameters and quickly locks the optimal solution in the high-dimensional parameter space through intelligent switching between global exploration and local convergence. For example, only 100 iterative experiments are needed to approach the optimal performance in the literature, and redundant testing is avoided through multi-dimensional stopping conditions (such as the performance difference threshold between two consecutive experiments or a preset upper limit on the number of experiments), shortening the development cycle several times and achieving efficient use of resources and time.
[0028] Furthermore, the accuracy of catalyst development is guaranteed through multi-dimensional technical details: the liquid preparation module supports gradient generation with a concentration step of 1%, and the temperature control accuracy of the reaction module reaches ±0.5°C, ensuring the high repeatability of experimental conditions; data-driven optimization strategies (such as component-performance three-dimensional response surface diagrams and sampling distribution diagrams) intuitively present the correlation between parameters and performance, combined with neural network model predictions to reduce reliance on human experience. In addition, the automated cleaning module eliminates cross-contamination by dynamically adjusting cleaning parameters (such as cleaning time and liquid type), and cooperates with cleaning report records to ensure the purity of each round of experimental data; full-process digital records are synchronized with digital twin technology to achieve experimental traceability and minimize errors. Not only has the scientific nature of catalyst formulation design been improved, but the reliability of experimental results and the predictability of catalyst performance have been significantly improved through precise parameter control and pollution avoidance. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 The figure is a flow chart showing an experimental control method for electrochemical catalyst development according to an exemplary embodiment.
[0030] Figure 2 is a structural block diagram of an experimental control system for electrochemical catalyst development according to an exemplary embodiment.
[0031] Figure 3 This is a block diagram of an experimental control system for electrochemical catalyst development according to an exemplary embodiment. DETAILED DESCRIPTION
[0032] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.
[0033] Throughout this specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
[0034] The present invention discloses an experimental control method for the development of electrochemical catalysts. Figure 1 ,include:
[0035] S100, based on the predetermined experimental type, determines the corresponding experimental objectives and initial parameter space, generates the first round of experimental combinations based on the initial parameter space, and automatically performs liquid preparation, reaction, sample synthesis and electrochemical performance testing based on the initial parameter space to obtain the corresponding first round of experimental result data. The initial parameter space includes metal ion concentration, voltage window, and reaction time.
[0036] Among them, it is necessary to first determine the experimental type and goal. According to the pre-set experimental type (such as water electrolysis, urea oxidation reaction, etc.), the specific goal of the experiment can be clarified, such as optimizing the catalytic performance of the oxygen evolution reaction (OER) or hydrogen evolution reaction (HER); then, the initial parameter space is set. In the embodiment of the present application, the metal ion concentration range is set to 0M / L to 1.0M / L, including at least three elements of nickel, cobalt, iron, copper, manganese, molybdenum, bismuth, gold, platinum, palladium, ruthenium or rhodium. For example, in an experiment, the Ni, Co, Fe ternary system may be selected; voltage window range: set the voltage window range to -5.0V to +5.0V to cover the needs of different electrochemical reaction conditions; reaction time range: set the reaction time to 10 seconds to 1800 seconds to ensure sufficient testing at different reaction rates.
[0037] Furthermore, a first round of experimental combinations is generated: based on the above initial parameter space, the system automatically generates a set of experimental formula combinations, for example, using a gradient concentration step of 1% of the total concentration to generate multiple ternary layered hydroxide catalyst formula combinations with different proportions.
[0038] Finally, enter the automatic execution of the experiment process:
[0039] Liquid preparation: According to the generated experimental formula combination, the required solution is accurately prepared through the multi-channel injection channel. For example, a peristaltic pump and solenoid valve are used to control the solution transfer to ensure the consistency of each experimental condition. Reaction: The prepared solution is transferred to the reaction vessel, and the reaction temperature is monitored and adjusted in real time by a temperature sensor to maintain it within the set value ±0.5 ° C. For example, a constant voltage electrodeposition method is used to directly generate Ni-Co-Fe ternary layered hydroxide on a carbon paper electrode. Sample synthesis and electrochemical performance testing: The catalyst performance is tested by electrochemical analysis methods supported by the electrochemical workstation (such as CV, LSV), and detailed data are recorded. For example, the linear sweep voltammetry curve (LSV) of the catalyst electrode in the HER and OER reactions is measured respectively.
[0040] By quickly generating and executing the first round of experiments in a preset parameter space, human intervention is reduced, significantly improving the efficiency of experiment startup. At the same time, it ensures that the experiment widely covers potential optimal solutions, providing a sufficient data basis for subsequent optimization, and ensures the high repeatability of each round of experiments by precisely controlling the liquid preparation and reaction conditions, thereby greatly improving the reliability and accuracy of the experimental results, thereby achieving efficient, comprehensive and accurate experimental process startup and execution.
[0041] S200, input the first round of experimental result data into the algorithm module used for data analysis and modeling of the experimental results, obtain the first round of analysis result data output by the algorithm module, dynamically generate the next round of experimental parameter plan based on the first round of analysis result data of the optimization algorithm subsystem, and drive the platform to execute the experiment again based on the next round of experimental parameter plan. The optimization algorithm module includes Bayesian optimization algorithm, SNOBFIT algorithm, and neural network prediction model.
[0042] Among them, the first round of experimental result data (such as overpotential, current density, etc.) is input into the algorithm module, which contains a variety of intelligent optimization algorithms such as Bayesian optimization algorithm, SNOBFIT algorithm and neural network prediction model; the algorithm module analyzes the experimental data and constructs a performance prediction model. For example, the Bayesian optimization algorithm is used to calculate the expected improvement value of each formula and identify high-potential areas in the current parameter space; based on the analysis results, the system dynamically generates the next set of experimental parameter schemes. Priority is given to areas with higher performance predictions for further exploration. For example, the Bayesian optimization algorithm is used in the global search stage, and the SNOBFIT algorithm is switched to in the local convergence stage to improve accuracy; based on the generated new round of experimental parameter schemes, the system automatically adjusts the liquid preparation and reaction conditions, and re-executes the experimental process to obtain new experimental result data.
[0043] Through data analysis and modeling, we avoid blind exploration, focus on potential optimal areas, and significantly improve experimental efficiency. At the same time, we dynamically switch optimization algorithms according to the needs of different stages, balance global exploration and local convergence, and ensure that the optimal solution is found within a limited number of experiments.
[0044] S210, in this embodiment of the present application, the optimization algorithm module adopts a dynamic combination strategy of Bayesian optimization and SNOBFIT algorithm, and the method further includes:
[0045] In the global exploration phase, the Bayesian optimization algorithm is preferentially used for efficient search of the parameter space; in the local convergence phase, the SNOBFIT algorithm is switched to improve the accuracy of the local optimal solution. The switching condition for dynamic balanced exploration and convergence is that the performance standard deviation of the current experimental round is lower than the preset threshold.
[0046] In the early stages of the experiment, the Bayesian optimization algorithm is preferred for efficient search of the parameter space. The Bayesian optimization algorithm estimates the performance of unknown areas by constructing a surrogate model (such as a Gaussian process), thereby guiding the design of subsequent experiments.
[0047] When the system detects that the performance standard deviation of the current experimental round is lower than the preset threshold, it switches to the SNOBFIT algorithm; the SNOBFIT algorithm is suitable for local fine search and can more accurately find the local optimal solution, thereby improving the accuracy of the final catalyst performance.
[0048] By dynamically combining Bayesian optimization and SNOBFIT algorithms, a wide range of parameter ranges can be quickly covered in the global exploration phase, allowing for fine-tuning in the local convergence phase to ensure that the global optimal solution is found within a limited number of experiments. Algorithms are dynamically switched according to the progress of the experiment to achieve a balance between global and local search, avoiding falling into local optimality or over-exploring invalid areas.
[0049] S220, in the embodiment of the present application, during the analysis process based on the experimental result data, the method further includes:
[0050] Generate a component-performance three-dimensional response surface plot based on the experimental results data to visualize the relationship between different metal ratios and catalytic performance; draw a sampling distribution diagram to display the coverage density of the current experimental parameters in real time; and output the formulation parameters of the current optimal solution and the corresponding performance indicators.
[0051] Among them, the experimental result data is used to generate a component-performance three-dimensional response surface diagram to intuitively show the relationship between different metal ratios and catalytic performance; for example, in the screening of water electrolysis catalysts, the performance response surface diagrams of the oxygen evolution reaction (OER) and hydrogen evolution reaction (HER) of Ni-Co-Fe ternary layered hydroxide are drawn.
[0052] The coverage density of the current experimental parameters is displayed in real time, helping researchers understand which parameter intervals have been fully explored and which areas require further testing. For example, in the optimization of urea oxidation reaction (UOR) catalysts, a sampling distribution map is drawn to evaluate the coverage of the parameter space.
[0053] The system automatically outputs the formula parameters of the current optimal solution and its corresponding performance indicators (such as overpotential or charge transfer amount). For example, in the water electrolysis experiment, it outputs the optimal catalyst formula.
[0054] The three-dimensional response surface plot and sampling distribution diagram intuitively present the relationship between parameters and performance, helping researchers quickly understand experimental results and make decisions. By outputting the formulation parameters and performance indicators of the current optimal solution, real-time feedback is provided to ensure that each round of experiments can be optimized and adjusted based on the latest data, thereby improving overall experimental efficiency and result reliability.
[0055] S230, in this embodiment of the present application, the method further includes:
[0056] The first step is automatic cleaning: After each experiment, the system automatically initiates a cleaning program, using deionized water and cleaning fluid to thoroughly clean the pipetting and cleaning modules, reaction modules, and electrodes. The cleaning process is controlled by the host computer according to a preset program, including parameters such as cleaning time, liquid type, and flow rate.
[0057] The second step is to dynamically adjust cleaning parameters: the cleaning time is dynamically adjusted based on the type and concentration of the solution used in the previous experiment. For example, for highly concentrated metal solutions or strong acidic / alkaline solutions, the cleaning time is extended to ensure complete removal of residues.
[0058] The third step is to generate a cleaning report: After cleaning is completed, the system automatically generates a cleaning report, recording information such as the amount of cleaning fluid used, cleaning time, cleaning effect evaluation indicators (such as conductivity changes, surface cleanliness test results), etc., for subsequent experimental reference and traceability.
[0059] Automatic cleaning ensures the independence of each experiment and the purity of the data, avoiding errors caused by residues; dynamic adjustment of cleaning parameters to adapt to different experimental conditions ensures consistent cleaning effects and enhances the repeatability of experimental results; automatic generation and storage of cleaning reports supports complete traceability of experimental data, meeting the requirements of auditability and standardization in scientific research and industrial scenarios.
[0060] By integrating automated hardware and intelligent algorithms, a closed-loop "synthesis-testing-optimization" process has been constructed, enabling efficient exploration and precise optimization in electrochemical catalyst development. Step S200 dynamically generates the next round of experimental parameters through data analysis and modeling. Combined with the dynamic combination strategy of Bayesian optimization and the SNOBFIT algorithm in S210, this balances global exploration with local convergence, significantly improving the scientific nature of the experimental path. The component-performance three-dimensional response surface plot and sampling distribution diagram in S220 intuitively present parameter correlations, assisting in real-time decision-making. The automated cleaning module and report generation in S230 ensure consistency in experimental conditions and data reliability. Through data-driven adaptive optimization, the overall process significantly shortens the R&D cycle, reduces resource waste, and enables precise formulation screening in high-dimensional parameter spaces, addressing the pain points of traditional trial-and-error methods, such as low efficiency and high robot dependence.
[0061] S300, repeatedly executing the step of obtaining the next round of experimental parameter plan, judging whether the preset stop condition is met based on the new next round of experimental parameter plan, if not, proceeding to the next round of experiments, otherwise stopping the experiment.
[0062] Among them, after each experiment, the system automatically inputs the latest experimental results into the optimization algorithm module (such as Bayesian optimization or SNOBFIT) to dynamically generate the next round of experimental parameter plans; based on the newly generated parameter plans, the platform is driven to re-execute the full process experiment of liquid preparation, reaction, sample synthesis and electrochemical performance testing.
[0063] In the embodiment of the present application, the stopping conditions are judged as follows: Condition one, the performance difference threshold between two consecutive experiments: the system calculates the performance difference between the current experiment and the previous experiment (such as the difference in overpotential). If the difference is less than the set threshold (such as 10mV or 1C), it is determined that the experiment has converged to the vicinity of the optimal solution. Condition two, the upper limit of the total number of experiments: the maximum number of experiments is preset (such as 100 times). If the current round reaches the upper limit, the experiment is forcibly terminated. Condition three, the gradient threshold of the performance response surface graph: the component-performance three-dimensional response surface graph is generated through the first round of analysis result data, and the gradient value near the current optimal solution is calculated. If the gradient value is less than the preset threshold (such as 0.1mV / unit concentration), it indicates that there is little room for performance improvement and the experiment can be stopped.
[0064] In the embodiment of the present application, it is decided whether to continue the experiment: if any of the above conditions is not met, the next round of experiments is continued based on the newly generated parameter scheme; if all conditions are met, the experiment is terminated and the final result is output.
[0065] By scientifically setting stopping conditions (such as performance difference threshold, total number of times upper limit, gradient threshold), we can prevent blind exploration when approaching the optimal solution, save resources and shorten the R&D cycle. For example, in the optimization of water electrolysis catalysts, only 100 experiments are needed to approach the optimal performance in the literature, which reduces the number of experiments by more than 90% compared with traditional methods.
[0066] When the experimental results tend to be stable, terminate the experiment promptly to avoid misjudgment due to local fluctuations and ensure that the obtained optimal solution has high credibility. For example, if the performance difference between two consecutive experiments is less than 10mV, it can be confirmed that the catalytic performance has stabilized.
[0067] By comprehensively judging the multi-dimensional stopping conditions, we can avoid missing out on optimal solutions due to premature termination and prevent resource waste caused by excessive iterations. For example, combining gradient analysis with performance response surface plots can accurately identify the global optimal region.
[0068] It should be pointed out here that the experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module and an electrochemical module.
[0069] Liquid preparation module: According to the dynamically generated parameter scheme, the metal precursor solution (such as Ni 2+ 、Co 2+ 、Fe 3+ solution), supports automatic generation of gradient concentration (the step size of gradient concentration is adjustable, and the minimum step size is 1% of the total concentration); it can ensure that the initial conditions of each round of experiments are highly controllable, eliminate human errors, and provide a consistent raw material basis for subsequent reactions.
[0070] Liquid handling and cleaning module: After each round of experiments, deionized water and cleaning solution are used to automatically clean the pipes, reaction vessels and electrodes. The cleaning time is dynamically adjusted according to the type and concentration of the solution used in the previous round (for example, high-concentration solution prolongs the cleaning time); it can prevent cross-contamination, ensure the independence of experimental conditions, and ensure the purity and repeatability of each round of experimental data.
[0071] Reaction module: Controls reaction temperature (±0.5°C), stirring rate, and voltage window (-1.5V to +0.7V) to synthesize catalyst samples through electrodeposition or liquid-phase reaction. It provides a stable reaction environment, ensures the consistency of sample synthesis, and directly affects the activity and stability of the catalyst.
[0072] Electrochemical module: Through linear sweep voltammetry (LSV), chronoamperometry (IT) and other testing methods, the electrochemical properties of the catalyst (such as overpotential η 10 , charge transfer amount); it can quickly feedback experimental results, provide high-quality data support for optimization algorithms, and accelerate performance evaluation and parameter adjustment.
[0073] By intelligently judging multi-dimensional indicators such as the performance difference threshold, the upper limit of the number of experiments, and the response surface gradient, a dynamic termination mechanism is constructed to scientifically determine when to stop the experiment, avoiding redundant operations and waste of resources. At the same time, relying on the highly integrated design of hardware modules such as liquid preparation, cleaning, reaction, and electrochemistry, the experimental conditions are precisely controlled and data reliability is guaranteed, achieving a balance between efficient exploration and local convergence within a limited number of experiments, and ultimately screening out a high-precision optimal solution, significantly improving the success rate and economy of electrochemical catalyst development.
[0074] Based on the technical solution of this application, the role of the solution will be further demonstrated through the following cases:
[0075] Case 1:
[0076] A research team used the automated experimental system described in this application to conduct a formulation screening experiment for a Ni-Co-Fe ternary layered hydroxide catalyst. Their goal was to identify the optimal catalyst formulation by systematically testing the effects of varying metal ratios on the performance of the oxygen evolution reaction (OER) and hydrogen evolution reaction (HER). The experiment encompassed two phases: catalyst synthesis and electrochemical performance testing, all performed independently by the automated platform without the need for human intervention.
[0077] 1. Parameter Setting and Combination Generation: Targeting the water electrolysis reaction, the total concentration of the three metal elements (Ni, Co, and Fe) was set to 0.3 M / L, with the concentration of each metal ranging from 0–0.3 M / L in steps of 0.03 M / L. The system automatically generated 66 combinations of metal ratios, covering the entire parameter space. For example, typical formulas such as Ni:Co:Fe = 0.7:0:0.3 and 0:0.7:0.3 were included in the first round of experiments.
[0078] 2. Automated Experiment Execution:
[0079] Each round of experiments includes 16 standardized steps, including solution preparation, deposition, cleaning, stability testing and performance evaluation. For example, the precursor solution is prepared and the catalyst is generated by constant voltage electrodeposition (-1V, 60 seconds); the electrode surface is stabilized by cyclic voltammetry (CV) and the OER (0–0.7V) and HER (-0.8–
[0080] The system performed a linear sweep voltammetry (LSV) curve at -1.5 V; and electroacid hydrolysis (0 V, 600 seconds) to dissolve the catalyst and restore the electrode, allowing for multiple rounds of repeated experiments. The system automatically completed 66 experiments in continuous operation, taking a total of 90 hours and 34 minutes.
[0081] 3. High-performance regional screening and local optimization:
[0082] Global scan: The performance response surface diagram of OER and HER was drawn based on 66 sets of experimental data. The results showed that the optimal OER performance was Ni:Co:Fe=0.7:0:0.3(η 10 =233mV), the HER performance is optimal when Ni:Co:Fe=0:0.7:0.3(η 10 =242mV).
[0083] Local convergence: For the OER high performance region (Ni: 0.55–0.85, Fe: 0.15–0.45), the step size was reduced to 3% of the total concentration, generating 66 sets of fine combination experiments; for the HER high performance region (Co: 0.3–0.8), the step size was set to 5%, generating 35 sets of combination experiments. The system automatically performed all experiments again to further confirm the optimal formula: OER optimal: Ni 67 Fe 33 (OH)x(η 10 =233mV, close to the optimal value reported in the literature); HER optimal: Ni 67 Fe 35 Co 45 (OH)x(η 10 =242mV).
[0084] 4. Full-process assurance and data reliability: Hardware coordinated control ensures precise control of reaction temperature within ±0.5°C and voltage accuracy within ±0.01V. The pipetting module features a minimum gradient concentration step of 1% to ensure formulation consistency. Automatic cleaning and traceability dynamically adjusts the cleaning time after each round of experiment (e.g., extending it to 304 seconds for high Fe concentrations), and cumulatively records cleaning solution usage and effect evaluation indicators to avoid cross-contamination.
[0085] The traditional trial-and-error method requires several months to complete ≥300 experiments. This solution completed 167 sets of experiments (including global and local optimization) in just 10 days (229 hours), shortening the R&D cycle by more than 90%. 10 =233mV, approaching the best-known level in the literature for this type of material system, demonstrating the effectiveness of automated screening. By dynamically combining Bayesian optimization with the SNOBFIT algorithm, high-potential regions are prioritized, reducing the number of ineffective experiments (167 global scans + local convergences compared to ≥300 with traditional methods), saving reagents and energy.
[0086] This case study fully demonstrates the advantages of this application's automated experimental system in electrochemical catalyst development: through intelligent parameter space partitioning, global-local joint optimization strategies, and hardware closed-loop control, it enables efficient and accurate catalyst screening. The system not only significantly shortens R&D cycles but also replaces empirical reliance with data-driven decision-making, providing a scalable solution for industrial-scale catalyst development.
[0087] Case 2:
[0088] A research team conducted independent optimization experiments on a Ni-Co-Fe ternary layered hydroxide catalyst for water electrolysis, using the automated experimental system described in this application and a Bayesian optimization algorithm. The core objective of the experiment was to dynamically adjust the metal ratio based on real-time feedback from electrocatalytic performance data, enabling iterative performance-driven material development. Unlike traditional full-combinatorial screening, this approach employed a closed-loop "experiment-test-optimize" strategy, completing the entire optimization process without manual intervention.
[0089] In the experimental design, the system first randomly generates 3 groups of initial formulas (such as Ni:Co:Fe=
[0090] 0.4:0.2:0.4, 0.3:0.5:0.2, 0.6:0.1:0.3), the catalyst samples were synthesized by constant voltage electrodeposition, and the solution preparation, deposition, cleaning and OER performance testing (linear sweep voltammetry LSV) were automatically performed. The current density was 10mA / cm in each round of experiment. 2 The overpotential η under 10 As a performance indicator. Then, the (metal ratio, η 10 ) data is input into the Bayesian optimization algorithm (Gaussian process proxy model + UCB acquisition function) to generate the recommended formula for the fourth experiment (such as Ni:Co:Fe=0.7:0.05:0.25). The system automatically executes the full process experiment of this formula to obtain the new η 10 The training dataset is continuously updated. As the iteration progresses, the algorithm gradually focuses on the low overpotential region (such as Ni:Co:Fe≈0.65:0.08:0.27), while retaining a small amount of sampling in the low performance region to avoid the local optimal trap. 10 The process was terminated when the difference was <10mV or the number of experiments reached 100. Finally, in the 91st experiment, the optimal formula was found to be Ni 63 Fe 37 (OH)x, its OER performance η 10 =237mV, which is only 4mV different from the optimal result (233mV) of the full combination screening in Example 1, while the number of experiments is reduced from 167 to 100 times, and the efficiency is improved by about 40%.
[0091] This solution achieves a balance between global exploration and local convergence through the Bayesian optimization algorithm. In the early stage, the algorithm quickly covers the parameter space (for example, the first 30 rounds of exploration have a wide range), and then concentrates on the high-performance area (such as Ni>0.6, Fe>0.3), and finally locks the optimal solution through small step tuning. This "wide first, then fine" model is significantly better than the blindness of the traditional trial and error method. At the hardware level, the system accurately controls the reaction temperature by ±0.5°C and the voltage accuracy by ±0.01V, and eliminates cross contamination through a dynamic cleaning module (extending the cleaning time to 304 seconds for high Fe concentrations), ensuring consistency of experimental conditions in each round. In addition, this method has made breakthroughs in the fields of photocatalysis, battery electrolytes, etc. For example, the Zhejiang Normal University team used Bayesian optimization to screen out organic photocatalysts with performance close to iridium-based catalysts from 560 candidate molecules; the Beihang University team optimized the methane combustion catalyst PtPd@CeZrOx core-shell nanospheres in only 18 experiments, verifying the universality of the intelligent optimization algorithm in complex chemical space.
[0092] Compared with the traditional method which takes 6 months to complete ≥300 experiments, this solution only takes 3 weeks to complete 100 experiments, shortening the R&D cycle by 95%. 10 =237mV) is 12–14% lower than reported values in the literature, while parameter space coverage is increased by over 10-fold, and data reliability error is less than 1%. This achievement not only demonstrates the deep integration of Bayesian optimization and automated systems but also provides an efficient and cost-effective solution for industrial-scale catalyst development. By replacing empirical reliance with a scientific search strategy, the system can accurately locate the optimal formulation in a high-dimensional parameter space, promising broad application prospects.
[0093] Case 3:
[0094] Based on the automated experimental system described in this application, a research team combined the Bayesian optimization algorithm to conduct independent optimization experiments on Ni-Cu-Fe ternary layered hydroxide catalysts for urea oxidation reaction (UOR). As a key step in low-potential water electrolysis to produce hydrogen, UOR has a theoretical potential significantly lower than that of the oxygen evolution reaction (OER), and is expected to significantly improve energy conversion efficiency. However, UOR involves a complex six-electron transfer process, which places higher demands on catalyst design. Through the collaboration of the automation platform and intelligent algorithms, the team successfully achieved efficient exploration and performance optimization.
[0095] 1. Experimental Design and Execution:
[0096] The first step is parameter setting and recipe generation: Taking the Ni-Cu-Fe ternary system as the target, the total metal concentration is set to 0.3M / L and the step size is 0.03M / L. The system automatically generates the initial recipe by permutation and combination, and dynamically adjusts the Cu element ratio to replace the traditional Co-based system. For example, the first round randomly generates Ni:Cu:Fe=
[0097] Combinations such as 0.5:0.2:0.3, 0.6:0.1:0.3, 0.4:0.3:0.3, etc.
[0098] The second step involves performance testing and feedback mechanisms. Chronoamperometry (iT) testing was performed using a 1M KOH + 0.05M urea solution as the electrolyte at a constant voltage of 0.8V. The total charge transfer (in C) over the first 120 seconds was recorded. This metric directly reflects catalytic activity; higher values indicate better performance. The system extracts data in real time and feeds it into a Bayesian optimization algorithm, dynamically updating the proxy model (Gaussian process + UCB acquisition function) to generate the next set of recommended recipes.
[0099] The third step is closed-loop iteration and termination conditions: the experiment completed 35 rounds of iterations, and each round automatically executed the entire process of liquid preparation, deposition, cleaning and performance testing. The Bayesian optimization strategy quickly covers the parameter space (such as Ni>0.6, Fe>0.25 areas) in the early stage, and then focuses on the high-performance neighborhood (such as Ni:Fe≈0.7:0.3), while retaining a small amount of sampling of the low-performance area to prevent local convergence. The process is terminated when the charge difference between two consecutive experiments is <0.5C or the preset number of experiments is reached. Finally, in the 7th experiment, the optimal formula was found to be Ni 72 Fe 28 (OH) x The total charge transfer reaches 13.7C, which is significantly better than the 12.5C level reported in the literature.
[0100] The Bayesian optimization algorithm extensively explored the Ni-Cu-Fe ratio in the early stage (the response surface plot showed that the first 10 rounds covered the Ni = 0.4–0.8, Fe = 0.1–0.4 range), then concentrated on the high-charge region (such as Ni>0.7, Fe>0.25), and finally locked the optimal solution through small step tuning.
[0101] The system accurately controls the reaction temperature by ±0.5°C and the voltage accuracy by ±0.01V, and eliminates cross contamination through a dynamic cleaning module (extending the cleaning time to 304 seconds for high Cu concentrations); for example, in the 15th round of experiments, the use of a high-Cu solution resulted in pipe residues, and the system automatically extended the cleaning time to 350 seconds to ensure the consistency of subsequent experimental data. Compared with traditional methods that take several months to complete ≥100 experiments, this solution only takes 1 week to complete 35 experiments, shortening the R&D cycle by 95%. At the same time, the optimal performance (13.7C) is 10% higher than the literature value, and the parameter space coverage is increased by more than 5 times. By replacing empirical dependence with a scientific search strategy, the system can accurately locate the optimal formula in a high-dimensional parameter space, providing an efficient solution for the development of industrial-grade UOR catalysts. This achievement not only verifies the deep integration capabilities of Bayesian optimization and automation systems, but also lays the foundation for the industrialization of low-potential water electrolysis hydrogen production technology.
[0102] The present application discloses an experimental control system for electrochemical catalyst development, referring to Figure 2 ,include:
[0103] The first round of experimental result data acquisition module 001 determines the corresponding experimental objectives and initial parameter space based on the predetermined experimental type, generates the first round of experimental combinations based on the initial parameter space, and automatically performs liquid preparation, reaction, sample synthesis and electrochemical performance testing based on the initial parameter space to obtain the corresponding first round of experimental result data, wherein the initial parameter space includes metal ion concentration, voltage window, and reaction time;
[0104] The next round of experimental parameter plan generation module 002 inputs the first round of experimental result data into the algorithm module for data analysis and modeling of the experimental results, obtains the first round of analysis result data output by the algorithm module, and uses it to dynamically generate the next round of experimental parameter plan based on the first round of analysis result data of the optimization algorithm subsystem. The platform is driven to execute the experiment again based on the next round of experimental parameter plan. The optimization algorithm module includes the Bayesian optimization algorithm, the SNOBFIT algorithm, and the neural network prediction model;
[0105] Repeat execution module 003 is used to repeatedly execute the step of obtaining the next round of experimental parameter scheme, and judge whether the preset stop condition is met based on the new next round of experimental parameter scheme. If not, the next round of experiment is carried out, otherwise the experiment is stopped. Figure 3 The experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module, and an electrochemical module.
[0106] An embodiment of the present application further discloses an experimental control system for electrochemical catalyst development, comprising a processor in which a program of any one of the above-described experimental control methods for electrochemical catalyst development is run.
[0107] An embodiment of the present application further discloses a storage medium storing a program of any one of the above-mentioned experimental control methods for electrochemical catalyst development.
[0108] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. An experimental control method for electrochemical catalyst development, characterized in that: include: Determining the corresponding experimental objectives and initial parameter space based on a predetermined experimental type, generating a first round of experimental combinations based on the initial parameter space, and automatically performing liquid preparation, reaction, sample synthesis, and electrochemical performance testing based on the initial parameter space to obtain corresponding first round experimental result data, wherein the initial parameter space includes metal ion concentration, voltage window, and reaction time; Inputting the first round of experimental result data into an algorithm module for performing data analysis and modeling on the experimental results, obtaining the first round of analysis result data output by the algorithm module, dynamically generating a next round of experimental parameter plan based on the first round of analysis result data of the optimization algorithm subsystem, and driving the platform to execute the experiment again based on the next round of experimental parameter plan, wherein the optimization algorithm module includes a Bayesian optimization algorithm, a SNOBFIT algorithm, and a neural network prediction model; Repeat the steps of obtaining the next round of experimental parameter plan, and determine whether the preset stop condition is met based on the new next round of experimental parameter plan. If not, proceed to the next round of experiments, otherwise stop the experiment. The experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module, and an electrochemical module.
2. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: The stop condition includes at least one of the following: Obtain performance difference data corresponding to two consecutive experiments, and determine whether the performance difference data is less than a set threshold; Get the total number of experiments and determine whether the total number of experiments reaches the preset upper limit; A corresponding performance response surface graph is obtained based on the first round analysis result data, and based on the performance response surface graph, it is determined whether the gradient of the current optimal solution is less than a preset gradient threshold.
3. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: The optimization algorithm module adopts a dynamic combination strategy of Bayesian optimization and SNOBFIT algorithm, and the method further includes: In the global exploration phase, the Bayesian optimization algorithm is preferentially used for efficient search of the parameter space; in the local convergence phase, the SNOBFIT algorithm is switched to improve the accuracy of the local optimal solution. The switching condition for dynamic balanced exploration and convergence is that the performance standard deviation of the current experimental round is lower than the preset threshold.
4. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: In the analysis process based on the experimental result data, the method further includes: Generate a component-performance three-dimensional response surface plot based on the experimental results data to visualize the relationship between different metal ratios and catalytic performance; draw a sampling distribution diagram to display the coverage density of the current experimental parameters in real time; and output the formulation parameters of the current optimal solution and the corresponding performance indicators.
5. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: The method also includes: The metal ion concentration range in the initial parameter space is 0M / L to 1.0M / L, the voltage window range is -5.0V to +5.0V, the reaction time range is 10 seconds to 1800 seconds, and the metal ions include at least three elements of nickel, cobalt, iron, copper, manganese, molybdenum, bismuth, gold, platinum, palladium, ruthenium or rhodium.
6. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: The method also includes: After each round of experiments, deionized water and cleaning solution are used to thoroughly clean the pipetting cleaning module, reaction module and electrodes. The cleaning process is automatically controlled by the host computer according to the preset program, and the cleaning time is dynamically adjusted according to the type and concentration of the solution used in the previous round of experiments. After the cleaning is completed, the system automatically generates a cleaning report, recording the amount of cleaning solution used, cleaning time and cleaning effect evaluation indicators.
7. The experimental control method for electrochemical catalyst development according to claim 1, characterized in that: The method also includes: The liquid preparation module realizes the precise proportioning of the metal precursor solution through a multi-channel liquid injection channel and supports the automatic generation of gradient concentration, wherein the step length of the gradient concentration is adjustable, and the minimum step length is 1% of the total concentration.
8. An experimental control system for electrochemical catalyst development, characterized in that: include: a first-round experimental result data acquisition module, which determines the corresponding experimental objectives and initial parameter space based on a predetermined experimental type, generates a first-round experimental combination based on the initial parameter space, and automatically performs liquid preparation, reaction, sample synthesis, and electrochemical performance testing based on the initial parameter space to obtain the corresponding first-round experimental result data; wherein the initial parameter space includes metal ion concentration, voltage window, and reaction time; A module for generating a parameter plan for the next round of experiments inputs the result data of the first round of experiments into an algorithm module for performing data analysis and modeling on the experimental results, obtains the first round analysis result data output by the algorithm module, and dynamically generates a parameter plan for the next round of experiments based on the first round analysis result data of the optimization algorithm subsystem, and drives the platform to execute the experiment again based on the parameter plan for the next round of experiments. The optimization algorithm module includes a Bayesian optimization algorithm, a SNOBFIT algorithm, and a neural network prediction model; The repeated execution module is used to repeatedly execute the steps of obtaining the parameter plan for the next round of experiments, and judge whether the preset stop conditions are met based on the new parameter plan for the next round of experiments. If not, the next round of experiments is carried out; otherwise, the experiment is stopped. The experimental process integrates a liquid preparation module, a pipetting and cleaning module, a reaction module and an electrochemical module.
9. An experimental control system for electrochemical catalyst development, characterized in that: The method comprises a processor in which a program of the experimental control method for electrochemical catalyst development according to any one of claims 1 to 7 is run.
10. A storage medium, characterized in that: A program storing the experimental control method for electrochemical catalyst development according to any one of claims 1 to 7 is stored.