A distributed control system and method for a liquid hydrogen refueling station

Abstract

The present invention provides a distributed control system and method for a liquid hydrogen refueling station, which relates to the technical field of intelligent control and includes: a fault diagnosis and early warning module for real-time monitoring of the operating status and process parameters of each device in the liquid hydrogen refueling station, and when detecting abnormal situations or potential risks, including equipment failures and process parameter overlimits, sending out early warning signals and generating corresponding treatment measure suggestions; a distributed control node module for controlling specific devices or areas in the liquid hydrogen refueling station, including liquid hydrogen storage devices, hydrogen compression devices, hydrogen dispensers, and safety monitoring devices. By integrating monitoring, diagnosis, control, analysis, and adaptive functions, the present invention improves the operating efficiency, safety, and intelligent management ability of the liquid hydrogen refueling station, realizes the automatic generation of fault early warnings and optimization suggestions, and the automatic adjustment of control schemes.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control, and particularly to a distributed control system and method for a liquid hydrogen refueling station. Background Art

[0002] Although the traditional centralized control system can, to a certain extent, achieve unified management and control of the entire liquid hydrogen refueling station, there are also some defects that cannot be ignored.

[0003] First of all, the centralized control system has a high degree of dependence on the central controller. This dependence may lead to an increase in system vulnerability. Once the central controller fails or is interfered with, the control system of the entire liquid hydrogen refueling station may be affected or even paralyzed, thus seriously affecting the normal operation of the liquid hydrogen refueling station. For example, in a certain actual operation, due to the central controller being damaged by lightning strike, the entire liquid hydrogen refueling station could not work properly, causing a significant loss to the operation.

[0004] Secondly, with the expansion of the scale of the liquid hydrogen refueling station, the complexity and maintenance cost of the centralized control system will also increase continuously. This increase is not only reflected in the increase of hardware devices and the complexity of wiring, but also in the upgrade and maintenance of the software system. For example, when new equipment or functions need to be added to the liquid hydrogen refueling station, large-scale modification and debugging of the entire control system may be required, which not only increases the work difficulty but also may introduce new risk points. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a distributed control system and method for a liquid hydrogen refueling station, which improves the reliability and flexibility of the system.

[0006] To solve the above technical problem, the technical solution of the present invention is as follows:

[0007] In the first aspect, a distributed control system for a liquid hydrogen refueling station includes:

[0008] A central control unit module, which is used to preprocess and extract features from control instructions, and then classify them through a trained classification model to obtain a classification result; determine an execution plan according to the classification result and issue instructions, and collect the operation data of the liquid hydrogen refueling station in real time through control nodes and sensors;

[0009] A fault diagnosis and early warning module, which is used to monitor the operation status and process parameters of each device in the liquid hydrogen refueling station in real time,

[0010] When detecting abnormal situations or potential risks, including equipment failures and process parameter overlimits, issue an early warning signal and generate corresponding treatment measure suggestions;

[0011] Distributed control node module, used to control specific equipment or areas in the liquid hydrogen refueling station, including liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen dispensers, and safety monitoring equipment;

[0012] Analysis module, used to analyze real-time sensor data, equipment operation status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of the liquid hydrogen refueling station, and generate an operation efficiency report, fault prediction, and optimization suggestions based on the analysis results;

[0013] Adaptive control function module, used to automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen refueling station and external environmental changes, including temperature, humidity, and equipment aging.

[0014] Furthermore, after preprocessing and feature extraction of the control instructions, they are classified by the trained classification model to obtain the classification results; according to the classification results, the execution plan is determined and the instructions are issued, and the operation data of the liquid hydrogen refueling station are collected in real time through the control node and sensors, including:

[0015] Preprocess the control instructions and extract instruction features, including instruction length, keyword occurrence frequency, and instruction structure;

[0016] Use the historical control instruction dataset to train the preset classification model to obtain the trained classification model;

[0017] Input the extracted instruction features into the trained classification model, and classify the types of instructions according to the input feature vectors, including equipment control instructions, process flow control instructions, or safety monitoring instructions, to obtain the classified instruction set;

[0018] Formulate an instruction execution plan according to the priority, equipment load, process flow progress, and safety status of the instructions in the classified instruction set;

[0019] According to the instruction execution plan, send the control instructions to the corresponding distributed control node, and collect the operation status data of the liquid hydrogen refueling station in real time through the distributed control node and sensors, including the current operation status of the equipment, real-time parameters of the process flow, and relevant data of environmental monitoring.

[0020] Furthermore, use the historical control instruction dataset to train the preset classification model to obtain the trained classification model, including:

[0021] Obtain the historical control instruction dataset, and the instruction dataset includes equipment control instructions, process flow control instructions, and safety monitoring instructions;

[0022] Use the neural network as the preset classification model and initialize the parameters of the classification model, including setting the initial learning rate and the number of iterations;

[0023] During the training process of the classification model, the range of parameter updates is restricted by the trust region method;

[0024] Determine the initial size of the trust region, and calculate the direction and size of parameter updates based on the parameters of the current classification model and the size of the trust region;

[0025] Use the historical control instruction dataset to train the classification model and calculate the value of the objective function;

[0026] Adjust and update the parameters of the classification model according to the value of the objective function and the constraints of the trust region;

[0027] After iterative training for a preset number of times, the trained classification model is obtained.

[0028] Furthermore, the calculation formula for the value of the objective function is:

[0029] ;

[0030] where represents the loss function; represents the total number of samples; represents the sample index; represents the total number of categories; , represent the category index; represents the category frequency; represents the sample for the category weight coefficient; represents the sample belongs to the category true label; represents the original output of the classification model for the sample belonging to the category ; represents the sample for the category bias adjustment term; represents the original output of the classification model for the sample belonging to the category ; represents the sample for the category bias adjustment term; represents the regularization coefficient; represents the square of the classification model parameters; represents the base of the natural logarithm.

[0031] Furthermore, the operating status and process parameters of each device in the liquid hydrogen hydrogenation station are monitored in real time. When abnormal conditions or potential risks are detected, including equipment failures and process parameter overlimits, warning signals are issued, and corresponding treatment measure suggestions are generated, including:

[0032] Continuously analyze the real-time monitored equipment and process parameters, and detect the real-time status of the equipment and process parameters by comparing the current data with the preset normal operating range or mode;

[0033] Set a threshold range for each process parameter. When the process parameter exceeds the threshold range, it is regarded as abnormal, and the warning mechanism is triggered to generate a warning signal. The warning signal includes the abnormal type, occurrence time, relevant equipment or parameter information;

[0034] According to the abnormal type in the warning signal, determine the corresponding treatment measures from the emergency treatment plan library, and generate treatment measure suggestions, including operation steps, required tool or equipment list, and safety precautions information.

[0035] Furthermore, analyze the real-time sensor data, equipment operating status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of the liquid hydrogen hydrogenation station, and generate an operation efficiency report, fault prediction, and optimization suggestions according to the data analysis results, including:

[0036] Continuously collect real-time sensor data during the operation of the liquid hydrogen hydrogenation station, including key parameters such as temperature, pressure, flow rate, and liquid level, and record the operating status of the equipment, including startup, shutdown, fault status changes, operating time, and maintenance records;

[0037] Extract features that have an impact on the generation of operation efficiency, fault prediction, and optimization suggestions from the real-time sensor data and equipment operating status data, including the average operating time, fault frequency, and energy consumption level of the equipment;

[0038] According to the extracted features, calculate the operation efficiency indicators of the liquid hydrogen hydrogenation station, including equipment utilization rate, energy consumption efficiency, and process stability, and generate an operation efficiency report, including the calculation results of the efficiency indicators, trend analysis, and problem point identification content;

[0039] According to the operation efficiency indicators, historical fault record data, and real-time monitoring data, construct a fault prediction model to predict the future fault types and probabilities;

[0040] Take improving the operation efficiency of the liquid hydrogen hydrogenation station and reducing the failure rate as the optimization goal, define the optimization problem, and map the solution space of the optimization problem to the search space of the whale population. Each whale individual represents a solution;

[0041] By simulating the foraging behavior of whales, including surrounding prey, bubble net attacks, and searching for prey, the positions of whale individuals are continuously updated, and the fitness function is used to evaluate the quality of whale individuals.

[0042] Repeat the process of simulating the foraging behavior of whales and updating the positions of whale individuals until the preset number of iterations is reached to determine the final solution.

[0043] Based on the final solution, generate optimization suggestions, including suggestions for equipment adjustment, process flow improvement, and maintenance plan optimization.

[0044] Furthermore, the calculation formula of the fitness function is:

[0045] ;

[0046] Among them, represents the comprehensive performance of the equipment in period ; , , , represent the weight coefficients; represents the actual operating time of the equipment within period ; represents the planned operating time of the equipment within period ; represents the unplanned downtime; represents the planned operating time; represents period the theoretical output of the equipment within; represents the theoretical output; represents the actual output; represents the theoretical energy consumption of the equipment within period ; represents the input energy; represents the output energy; represents the quality coefficient of the equipment within period ; represents the number of defective products; represents the total number of products; represents period the failure rate of the equipment within; represents the severity coefficient of the failure consequence.

[0047] Furthermore, according to the actual operating conditions of the liquid hydrogen refueling station and changes in the external environment, including temperature, humidity, and equipment aging, automatically adjust the control scheme, including:

[0048] Define fuzzy sets for temperature, humidity, equipment status, and control parameter adjustment amounts;

[0049] Determine the membership function for each fuzzy set and formulate fuzzy rules;

[0050] Based on the fuzzy sets, membership functions, and fuzzy rules, infer the fuzzy value of the output variable;

[0051] Convert the real-time monitored temperature, humidity, and equipment status data into fuzzy values through the membership function, and use the fuzzy rules for inference to obtain the fuzzy value of the control parameter adjustment amount;

[0052] Convert the fuzzy value of the control parameter adjustment amount into a specific numerical value through the defuzzification method, and automatically adjust the control scheme when changes in temperature, humidity, or equipment status are detected.

[0053] Furthermore, the calculation formula of the membership function:

[0054] ;

[0055] Wherein, represents the degree to which the variable belongs to the fuzzy set ; represents the variable; represents the base value of the left endpoint; represents the offset of the left endpoint; represents the exponent of the left half; represents the vertex; represents the base value of the right endpoint; represents the offset of the right endpoint; represents the scaling coefficient of the left half; represents the scaling coefficient of the right half; represents the exponent of the right half.

[0056] In a second aspect, a distributed control method for a liquid hydrogen refueling station includes:

[0057] After preprocessing and feature extraction of the control instruction, classify it through the trained classification model to obtain the classification result; determine the execution plan according to the classification result and issue the instruction, and collect the operation data of the liquid hydrogen refueling station in real time through the control node and sensors;

[0058] Monitor the operation status and process parameters of each device in the liquid hydrogen refueling station in real time,

[0059] When detecting abnormal situations or potential risks, including equipment failures and process parameter overlimits, issue a warning signal and generate corresponding treatment measure suggestions;

[0060] Control specific devices or areas in the liquid hydrogen refueling station, including liquid hydrogen storage devices, hydrogen compression devices, hydrogen dispensers, and safety monitoring devices;

[0061] Analyze the real-time sensor data, equipment operation status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of the liquid hydrogen hydrogen refueling station. According to the analysis results of the data, generate an operation efficiency report, fault prediction, and optimization suggestions.

[0062] Automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen hydrogen refueling station and external environmental changes, including temperature, humidity, and equipment aging.

[0063] The above solutions of the present invention have at least the following beneficial effects:

[0064] After preprocessing and feature extraction of the control instructions by the central control unit module, classify them through the trained classification model to obtain the classification results; determine the execution plan according to the classification results and issue instructions, and collect the operation data of the liquid hydrogen hydrogen refueling station in real time through the control nodes and sensors to ensure the coordinated operation of each device, thereby improving the overall operation efficiency. The analysis module can generate an operation efficiency report by analyzing the real-time sensor data, equipment operation status data, etc., providing data support for optimizing the operation of the liquid hydrogen hydrogen refueling station. The fault diagnosis and early warning module can monitor the operation status and process parameters of each device in the liquid hydrogen hydrogen refueling station in real time, timely detect abnormal situations or potential risks, and issue early warning signals, which helps prevent the occurrence of safety accidents.

[0065] The distributed control node module can control specific devices or areas in the liquid hydrogen hydrogen refueling station, including safety monitoring devices, to ensure rapid response and measures can be taken in case of emergencies. Through the real-time monitoring and early warning functions of the fault diagnosis and early warning module, preventive measures can be taken before the occurrence of equipment failures, reducing the failure rate.

[0066] The analysis module can analyze according to the historical fault record data to help identify fault patterns and provide support for fault prediction and preventive maintenance. The analysis module can analyze the equipment operation status data and process flow parameter data to help formulate a more reasonable maintenance plan and reduce unnecessary maintenance costs. Through the fault prediction function, maintenance work can be arranged in advance to avoid the shutdown of the liquid hydrogen hydrogen refueling station caused by equipment failures.

[0067] The adaptive control function module can automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen hydrogen refueling station and external environmental changes (such as temperature, humidity, and equipment aging) to ensure the stable operation of the liquid hydrogen hydrogen refueling station in different environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of a distributed control system for a liquid hydrogen hydrogen refueling station provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0069] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0070] As Figure 1 shown, an embodiment of the present invention provides a distributed control system for a liquid hydrogen refueling station, including:

[0071] A central control unit module 11, configured to preprocess and extract features from control instructions, and then classify them through a trained classification model to obtain a classification result; determine an execution plan according to the classification result and issue instructions, and collect the operation data of the liquid hydrogen refueling station in real time through control nodes and sensors;

[0072] A fault diagnosis and early warning module 12, configured to monitor the operation status and process parameters of each device in the liquid hydrogen refueling station in real time, and issue an early warning signal and generate corresponding treatment measure suggestions when detecting abnormal conditions or potential risks, including equipment failures and process parameter overlimits;

[0073] A distributed control node module 13, configured to control specific devices or areas in the liquid hydrogen refueling station, including liquid hydrogen storage devices, hydrogen compression devices, hydrogen dispensers, and safety monitoring devices;

[0074] An analysis module 14, configured to analyze the real-time sensor data, device operation status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of the liquid hydrogen refueling station, and generate an operation efficiency report, fault prediction, and optimization suggestions according to the data analysis results;

[0075] An adaptive control function module 15, configured to automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen refueling station and external environmental changes, including temperature, humidity, and equipment aging.

[0076] In the embodiment of the present invention, through centralized reception, processing, and sending of control instructions, unified management and scheduling of the liquid hydrogen refueling station are achieved. Continuously monitor the operation status of the liquid hydrogen refueling station to ensure the normal operation of each device, and promptly discover and handle abnormal situations. The central control unit can quickly respond to control instructions, improving the overall operation efficiency of the liquid hydrogen refueling station. Real-time monitoring of the device operation status and process parameters can promptly discover abnormal situations or potential risks and issue early warning signals. Through early warning and treatment measure suggestions, losses caused by equipment failures or process parameter overlimits can be reduced. Effectively prevent the occurrence of safety accidents and ensure the safe operation of the liquid hydrogen refueling station.

[0077] Precisely control specific equipment or areas in the liquid hydrogen refueling station to improve control accuracy and efficiency. The distributed control nodes can be flexibly configured according to actual needs to adapt to the requirements of different liquid hydrogen refueling stations. Through distributed control, the impact of a single node failure on the entire system is reduced, and the reliability of the system is improved. Through the analysis of a large amount of data, data support is provided for the operation and management of the liquid hydrogen refueling station to achieve data-driven decision-making. An operation efficiency report is generated based on the data analysis results to help optimize the operation efficiency of the liquid hydrogen refueling station. Through the analysis of historical fault record data, fault prediction and preventive maintenance are realized to reduce the occurrence of faults.

[0078] Automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen refueling station and changes in the external environment to ensure the stable operation of the liquid hydrogen refueling station. By optimizing the control scheme, the goals of energy conservation and emission reduction can be achieved, and the environmental performance of the liquid hydrogen refueling station can be improved.

[0079] In a preferred embodiment of the present invention, after preprocessing and feature extraction of the control instruction, it is classified by a trained classification model to obtain a classification result; an execution plan is determined according to the classification result and the instruction is issued. The operation data of the liquid hydrogen refueling station can be collected in real time through the control node and sensors, including:

[0080] Preprocess the control instruction and extract instruction features, including instruction length, keyword occurrence frequency, and instruction structure;

[0081] Use the historical control instruction data set to train a preset classification model to obtain a trained classification model;

[0082] Input the extracted instruction features into the trained classification model, and classify the type of the instruction according to the input feature vector, including equipment control instruction, process flow control instruction, or safety monitoring instruction, to obtain a classified instruction set;

[0083] Formulate an instruction execution plan according to the priority, equipment load, process flow progress, and safety status of the instructions in the classified instruction set;

[0084] According to the instruction execution plan, send the control instruction to the corresponding distributed control node, and collect the operation status data of the liquid hydrogen refueling station in real time through the distributed control node and sensors, including the current operation status of the equipment, real-time parameters of the process flow, and relevant data of environmental monitoring.

[0085] In the embodiments of the present invention, an original control instruction text is received. The instruction is cleaned using text processing techniques (such as word segmentation, stop word removal, etc.). The length of the instruction, i.e., the number of characters or words, is calculated. The frequency of occurrence of keywords in the instruction is counted. The keywords can be preset words related to device control, process flow, or safety monitoring. The structure of the instruction is analyzed, such as the beginning, body, and ending parts of the instruction, and the logical relationships between them.

[0086] A historical control instruction dataset is collected and organized, including instruction samples with labeled types. A suitable classification algorithm (such as a decision tree) is selected as the preset model. The preset model is trained using the historical dataset, and the model parameters are adjusted to optimize the classification performance. The performance of the trained model on the validation set is evaluated to ensure that the model has sufficient accuracy and generalization ability.

[0087] The preprocessed instruction features are converted into the form of feature vectors. The feature vectors are input into the trained classification model. The model makes inferences based on the input feature vectors and outputs the type label of the instruction. All classified instructions are collected to form a classified instruction set.

[0088] The classified instruction set is analyzed to determine the priority of each instruction. The device load condition is monitored in real time to evaluate the device's ability to execute instructions. The progress of the process flow is tracked to ensure the coordination between the instructions and the process flow. The safety condition is considered to avoid executing instructions that may pose safety risks. Considering the above factors, a reasonable instruction execution plan is formulated.

[0089] According to the instruction execution plan, the control instruction is sent to the corresponding distributed control node. The distributed control node receives the instruction and executes it to control the corresponding device or process flow. The operating state data of the liquid hydrogen refueling station, including device status, process parameters, and environmental data, is collected in real time through sensors. The collected data is sent to the central control unit for further processing and analysis.

[0090] Suppose a control instruction is received: "Start the hydrogen compression device, adjust the pressure to 10 MPa, and monitor the device temperature." Clean the instruction text to remove irrelevant words. Calculate the instruction length: several characters. Count the keyword frequencies: "Start" 1 time, "hydrogen compression device" 1 time, "pressure" 1 time, "10 MPa" 1 time, "monitor" 1 time, "device temperature" 1 time. The instruction consists of a beginning (start the device), a body (adjust the pressure, monitor the temperature), and an ending (no specific ending). Extract the feature vector and input it into the trained classification model. The model outputs the instruction type as "device control instruction". Determine the instruction priority as high.

[0091] Check the load condition of the hydrogen compression equipment and confirm that the equipment is available. Confirm that the process flow progress allows the execution of this instruction. Evaluate the safety of executing the instruction, and there is no safety risk. Immediately start the equipment, adjust the pressure to 10 MPa, and continuously monitor the equipment temperature. Send the instruction to the distributed control node that controls the hydrogen compression equipment. The control node executes the instruction, starts the equipment, and adjusts the pressure. The sensor collects the equipment status, pressure value, and temperature data in real time.

[0092] Through preprocessing and classification, quickly identify the instruction type and reduce the instruction processing time. Develop a reasonable execution plan to ensure that the instructions are executed in an orderly manner according to the priority and actual situation. Be able to process various types of control instructions and adapt to the requirements of different equipment and process flows. Through real-time data collection and analysis, timely adjust the instruction execution plan to improve the adaptability of the system. Conduct a safety assessment before executing the instruction to avoid executing instructions that may cause safety risks. Real-time monitor the equipment status and process flow parameters, promptly discover and handle abnormal situations, and ensure the safe operation of the liquid hydrogen filling station. Develop an execution plan based on the equipment load and process flow progress, rationally utilize resources, and avoid resource waste. Through data analysis, optimize the equipment operation parameters and process flow to improve the resource utilization efficiency.

[0093] In another preferred embodiment of the present invention, training a preset classification model using a historical control instruction data set to obtain a trained classification model may include:

[0094] Obtain a historical control instruction data set, and the instruction data set includes equipment control instructions, process flow control instructions, and safety monitoring instructions;

[0095] Use a neural network as the preset classification model and initialize the parameters of the classification model, including setting the initial learning rate and the number of iterations;

[0096] During the training process of the classification model, limit the range of parameter updates through the trust region method;

[0097] Determine the initial size of the trust region, and calculate the direction and size of parameter updates based on the parameters of the current classification model and the size of the trust region;

[0098] Use the historical control instruction data set to train the classification model and calculate the value of the objective function;

[0099] Adjust and update the parameters of the classification model according to the value of the objective function and the limitations of the trust region;

[0100] After a preset number of iterative trainings, obtain the trained classification model.

[0101] In an embodiment of the present invention, a historical control instruction dataset is loaded from a database or a file system. The dataset should contain various types of instructions, such as device control instructions, process flow control instructions, and safety monitoring instructions. A neural network is selected as the classification model, and its structure (such as the number of layers, the number of neurons in each layer, etc.) is determined. The parameters of the neural network, such as weights and biases, are initialized, usually using random initialization or initialization with a pre-trained model. The learning rate is set, which is a hyperparameter that controls the speed of parameter update. The number of iterations is determined, that is, the number of times the model is updated during the training process. Through the trust region method, which is an optimization algorithm used to limit the range of parameter updates to prevent the model from deviating too much during training. The size of the trust region is defined, that is, the maximum allowable change range when updating parameters.

[0102] Set the initial size of the trust region. In each iteration, according to the current parameters of the model and the size of the trust region, calculate the direction and size of the parameter update. This involves calculating the gradient to determine the direction of parameter update.

[0103] Input the historical control instruction dataset into the classification model for training. In each iteration, calculate the value of the objective function. According to the value of the objective function and the limitations of the trust region, determine whether the parameter update is effective. If the update is effective, adjust the parameters of the model according to the calculated update direction and size. If the update is ineffective or exceeds the trust region range, adjust the size of the trust region or reject the update. Train according to the set number of iterations until the preset number of iterations is reached. Save the trained classification model for subsequent use.

[0104] Suppose there is a historical dataset containing device control instructions, process flow control instructions, and safety monitoring instructions, and a neural network classification model is to be trained to identify these instructions.

[0105] Load the historical control instruction dataset from the database, which contains 1000 instructions, including 300 device control instructions, 400 process flow control instructions, and 300 safety monitoring instructions. Select a neural network with two hidden layers as the classification model. Initialize the weights and biases of the neural network to random values.

[0106] Set the learning rate to 0.01 and the number of iterations to 1000. Set the initial size of the trust region to 0.1. In each iteration, calculate the direction and size of the parameter update to ensure that the update is within the trust region range. Input the dataset into the neural network for training. In each iteration, calculate the value of the loss function and adjust the parameters of the model according to the limitations of the trust region. After 1000 iterations of training, obtain the trained neural network classification model. The accuracy of the model on the validation set reaches 95% and can be used for actual instruction classification tasks.

[0107] By training a neural network classification model, it is possible to more accurately identify device control instructions, process control instructions, and safety monitoring instructions. The introduction of the trust region method limits the range of parameter updates, preventing overfitting and instability during model training. Through training with historical datasets, the model can learn the characteristics and patterns of instructions and also has a certain classification ability for unseen instructions. The trust region method helps the model maintain stability during training and improves the generalization ability of the model. The neural network model can improve training efficiency through parallel computing and GPU acceleration. The trust region method reduces ineffective updates and waste of computing resources by restricting the range of parameter updates. The neural network model has good scalability and maintainability and can adjust the model structure and parameters according to actual needs. The trust region method during the training process can provide certain guidance for model updates, making the model updates more stable and reliable.

[0108] In a preferred embodiment of the present invention, the calculation formula for the value of the objective function is:

[0109] ;

[0110] Wherein, represents the loss function; represents the total number of samples; represents the sample index; represents the total number of categories; , represents the category index; represents the category frequency; represents the sample for the category weight coefficient; represents the sample belongs to the category true label; represents the original output of the classification model for the sample belonging to the category ; represents the sample for the category bias adjustment term; represents the original output of the classification model for the sample belonging to the category ; represents the sample for the category bias adjustment term; represents the regularization coefficient; represents the square of the classification model parameters; represents the base of the natural logarithm.

[0111] In the embodiments of the present invention, set the total number of samples , the total number of categories . Initialize the sample index , the category index , . Prepare all the required array storage , , , , etc. Traverse all the samples and count the number of occurrences of each category . Calculate the frequency of each category , that is, the number of samples in category divided by the total number of samples . According to the preset rules, assign a weight coefficient to each sample and category . Read the true category label of each sample from the dataset and convert it into one-hot encoded form, that is, if the sample belongs to category , then , otherwise . Input the sample into the classification model to obtain the original output of the model for the sample belonging to each category . According to the model design, calculate the bias adjustment term .

[0112] Initialize the first part value of the loss function to 0. Traverse all the samples and categories , and calculate the loss term of each sample for the category . Multiply the above result by , and to obtain the loss term of the current sample for the category . Add up the loss terms of all the samples and categories and multiply by to obtain the first part value of the loss function.

[0113] Initialize the regularization term value to 0. Traverse all the parameters of the classification model and calculate their sum of squares. Multiply the sum of squares by and the regularization coefficient to obtain the regularization term value. Add the first part value of the loss function and the regularization term value to obtain the total loss function value 。

[0114] By considering the weight coefficients of samples for classes and class frequencies , the loss function can more accurately reflect the impacts of different samples and classes on model training. The bias adjustment term can further adjust the output of the model for samples and improve the classification accuracy. The regularization term can prevent the model from overfitting and enhance the generalization ability of the model. By restricting the sum of squares of model parameters, the regularization term can prompt the model to select a simpler solution, thereby improving the generalization performance of the model. The design of the loss function is flexible, and the weight coefficients , bias adjustment term and regularization coefficient can be adjusted according to actual requirements. This flexibility enables the loss function to adapt to different classification tasks and datasets. When calculating the logarithmic function, terms are used to prevent numerical instability when the input of the logarithmic function is 0 or close to 0. This handling of numerical stability can ensure that the calculation process of the loss function is more reliable and robust.

[0115] In a preferred embodiment of the present invention, the operating states and process parameters of each device in a liquid hydrogen refueling station are monitored in real time. When an abnormal situation or potential risk is detected, including equipment failures and process parameter overlimits, a warning signal is issued, and corresponding treatment measure suggestions are generated, which may include:

[0116] Continuously analyze the real-time monitored equipment and process parameters, and detect the real-time states of the equipment and process parameters by comparing the current data with the preset normal operating range or mode;

[0117] Set a threshold range for each process parameter. When the process parameter exceeds the threshold range, it is regarded as abnormal, and the warning mechanism is triggered to generate a warning signal. The warning signal includes the type of abnormality, occurrence time, relevant equipment or parameter information;

[0118] According to the type of abnormality in the warning signal, determine the corresponding treatment measures from the emergency treatment plan library, and generate treatment measure suggestions, including operation steps, a list of required tools or equipment, and safety precautions information.

[0119] In the embodiments of the present invention, the operating states and process parameter data of each device in the liquid hydrogen refueling station are obtained in real time through sensors and a data acquisition system. The original data is processed such as cleaning, denoising, and calibration to ensure the accuracy and reliability of the data. Using data analysis algorithms, the preprocessed data is analyzed in real time, and the current data is compared with the preset normal operating range or mode. Set the normal operating range or mode: Based on historical data and expert experience, set the normal operating range or mode for each device and process parameter. Compare the current data with the preset normal operating range or mode to determine whether the real-time status of the device and process parameters is normal. According to process requirements and safety standards, set a reasonable threshold range for each process parameter. Monitor the changes in process parameters in real time. When the parameter value exceeds the set threshold range, it is determined as abnormal. Once an abnormality is detected, immediately trigger an early warning mechanism and prepare to generate an early warning signal. According to the results of the anomaly detection, determine the specific type of the anomaly (such as equipment failure, process parameter overrun, etc.). Record the exact time when the anomaly occurs. Determine the device or parameter related to the anomaly and extract relevant information. Combine the anomaly type, occurrence time, and related device or parameter information into an early warning signal and prepare to send it to relevant personnel or systems.

[0120] Establish an emergency treatment plan library: Develop and store in advance emergency treatment plans for various types of anomalies. According to the anomaly type in the early warning signal, search for a matching treatment plan in the emergency treatment plan library. Extract the matching treatment plan from the emergency treatment plan library. According to the treatment plan, generate specific treatment measure suggestions, including operation steps, a list of required tools or equipment, and safety precautions information. Send the treatment measure suggestions to relevant personnel or systems so that they can take measures to handle the anomaly in a timely manner.

[0121] Suppose there is a temperature sensor in the liquid hydrogen refueling station to monitor the temperature of the reaction kettle in real time. The normal operating range is set to 20°C to 80°C. The sensor collects temperature data once every second, and the data acquisition system transmits these data to the monitoring system in real time. The monitoring system preprocesses the received data, such as denoising and calibration, and then conducts comparative analysis. Set the threshold range for the temperature parameter: the lower limit is 20°C and the upper limit is 80°C. The monitoring system monitors the temperature data in real time. When the temperature exceeds 80°C, it is determined as abnormal.

[0122] The monitoring system generates warning signals, including the type of anomaly (temperature exceeding the limit), the occurrence time (such as 12:05:30 on April 1, 2023), and the relevant equipment (the temperature sensor of the reactor). The emergency treatment plan library contains treatment plans for temperature exceeding the limit, such as immediately shutting down the heating system and turning on the cooling system. The monitoring system searches and extracts the matching treatment plan from the emergency treatment plan library according to the type of anomaly in the warning signal. It generates suggestions for treatment measures, including operation steps (such as pressing the emergency stop button to shut down the heating system, opening the cooling water valve, etc.), a list of required tools or equipment (such as wrenches, keys for the cooling water valve, etc.), and information on safety precautions (such as wearing protective gloves, avoiding electric shock, etc.). The monitoring system sends the suggestions for treatment measures to the operators or control system of the liquid hydrogen filling station so that measures can be taken in a timely manner to handle the anomaly.

[0123] Through the real-time monitoring and warning mechanism, anomalies in equipment and process parameters can be detected in a timely manner, measures can be taken to prevent accidents from occurring, and the safety of the liquid hydrogen filling station can be improved. Early detection and handling of anomalies can avoid losses such as equipment damage and production interruption, and reduce the operating costs of the liquid hydrogen filling station. The automated monitoring and warning system can respond quickly to anomalies, reduce the delay in manual monitoring and handling, and improve the operating efficiency of the liquid hydrogen filling station. The generated suggestions for treatment measures provide clear operation steps and safety precautions for the operators, enhancing the scientificity and accuracy of decision-making. The combination of the real-time monitoring and warning system and the emergency treatment plan library provides strong support for the intelligent management of the liquid hydrogen filling station, promoting the digital transformation and upgrading of the liquid hydrogen filling station.

[0124] In a preferred embodiment of the present invention, controlling specific equipment or areas within the liquid hydrogen filling station, including liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen filling machines, and safety monitoring equipment, may include:

[0125] In the embodiments of the present invention, according to the layout and equipment distribution of the liquid hydrogen refueling station, a control system architecture is designed, including a central control room, a field equipment layer, a communication network layer, etc. Select a suitable industrial control system (such as SCADA, PLC, etc.) as the control platform to ensure the stability and reliability of the system. Cooperate with the suppliers of liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen dispensers, and safety monitoring equipment to obtain the equipment interface protocols and achieve seamless integration of the equipment with the control system. Install sensors (such as temperature sensors, pressure sensors, flow sensors, etc.) and actuators (such as valves, switches, etc.) on key equipment for real-time collection of equipment status parameters and control of equipment operation. Set up a data acquisition system to collect sensor data in real time and ensure the accuracy and timeliness of the data. Establish a data monitoring interface in the central control room to display the real-time status, operating parameters, and alarm information of the equipment, facilitating the operators to understand the equipment status at any time. Analyze the control requirements of the equipment based on the functions and operation processes of the equipment, such as temperature control of liquid hydrogen storage equipment, pressure regulation of hydrogen compression equipment, hydrogen filling volume control of hydrogen dispensers, etc. Based on the control requirements, design an automated control logic, including condition judgment, logical operation, control algorithms, etc., to achieve automated control of the equipment. Program the control logic into the control system for simulation testing and on-site commissioning to ensure the correctness and effectiveness of the control logic.

[0126] Configure safety monitoring equipment in the liquid hydrogen refueling station, such as hydrogen leakage detectors, fire alarms, video surveillance cameras, etc., for real-time monitoring of potential safety hazards. Set the thresholds of safety monitoring parameters according to safety standards and equipment characteristics, such as the upper limit of hydrogen concentration, the upper limit of temperature, etc. When the safety monitoring parameters exceed the set thresholds, trigger the warning mechanism and remind the operators to handle it in time through means such as audible and visual alarms, text message notifications.

[0127] Through the communication network layer, implement the remote control function of the central control room over the field equipment, such as remotely switching on and off the equipment, adjusting the equipment parameters, etc. Design an intuitive and easy-to-use operation interface, provide equipment control buttons, parameter setting input boxes, etc., to facilitate the operators to perform remote operations. Set different levels of operation permissions to ensure that only authorized personnel can perform remote control and operations, guaranteeing the security of the system.

[0128] Regularly maintain the control system, including software upgrades, hardware inspections, data backups, etc., to ensure the stable operation of the system. Optimize and adjust the control logic, operation interface, etc. according to the system operation conditions and feedback opinions to improve the performance and user experience of the system.

[0129] In a preferred embodiment of the present invention, real-time sensor data, equipment operation status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of a liquid hydrogen refueling station are analyzed, and based on the analysis results, an operation efficiency report, fault prediction, and optimization suggestions are generated, which may include:

[0130] Continuously collect real-time sensor data during the operation of the liquid hydrogen refueling station, including key parameters such as temperature, pressure, flow rate, and liquid level, and record the operation status of the equipment, including startup, shutdown, fault status changes, operation time, and maintenance records;

[0131] Extract features that have an impact on the generation of operation efficiency, fault prediction, and optimization suggestions from the real-time sensor data and equipment operation status data, including the average operation time, fault frequency, and energy consumption level of the equipment;

[0132] According to the extracted features, calculate the operation efficiency indicators of the liquid hydrogen refueling station, including equipment utilization rate, energy consumption efficiency, and process stability, and generate an operation efficiency report, including the calculation results of the efficiency indicators, trend analysis, and identification of problem points;

[0133] Construct a fault prediction model based on the operation efficiency indicators, historical fault record data, and real-time monitoring data to predict the types and probabilities of future faults;

[0134] Take improving the operation efficiency of the liquid hydrogen refueling station and reducing the failure rate as the optimization goal, define the optimization problem, map the solution space of the optimization problem to the search space of the whale population, and each whale individual represents a solution;

[0135] By simulating the foraging behavior of whales, including surrounding prey, bubble net attack, and searching for prey, continuously update the positions of whale individuals, and evaluate the quality of whale individuals through a fitness function,

[0136] Repeat the process of simulating the foraging behavior of whales and updating the positions of whale individuals until a preset number of iterations is reached to determine the final solution;

[0137] According to the final solution, generate optimization suggestions, including suggestions for equipment adjustment, process flow improvement, and maintenance plan optimization.

[0138] In the embodiments of the present invention, sensors such as temperature, pressure, flow rate, and liquid level are deployed at key parts of the liquid hydrogen hydrogenation station to ensure accurate collection of real-time data. A data acquisition system is established to obtain data from sensors regularly or in real time and store it in a database. By integrating the device control system, information such as the startup, shutdown, change of fault status, operation time, and maintenance records of the device is recorded in real time. The collected data is processed such as cleaning, denoising, and calibration to ensure the accuracy and reliability of the data. Analyze the real-time sensor data and device operation status data to identify features that affect operation efficiency, fault prediction, and optimization suggestions. According to the identified features, calculate key indicators such as the average operation time, fault frequency, and energy consumption level of the device.

[0139] Using the extracted eigenvalue, calculate operation efficiency indicators such as device utilization rate, energy consumption efficiency, and process stability. Integrate the calculated efficiency indicators, trend analysis (such as time series analysis, trend chart, etc.), and problem point identification (such as anomaly detection, bottleneck analysis, etc.) into an operation efficiency report. Intuitively display the content of the operation efficiency report in the form of charts, tables, etc., for easy understanding by decision-makers and operators.

[0140] Integrate the operation efficiency indicators, historical fault record data, and real-time monitoring data into a dataset for the fault prediction model. According to the data characteristics and prediction requirements, select a suitable machine learning algorithm (such as random forest) to construct the fault prediction model. Use the integrated dataset to train the selected machine learning algorithm to obtain the fault prediction model. Verify the performance of the fault prediction model through methods such as cross-validation and accuracy evaluation.

[0141] Define improving the operation efficiency of the liquid hydrogen hydrogenation station and reducing the failure rate as the optimization goal. According to the optimization goal, define specific optimization problems, such as minimizing energy consumption, maximizing device utilization rate, minimizing the failure rate, etc. Map the solution space of the optimization problem to the search space of the whale swarm optimization algorithm (such as the whale optimization algorithm), and each whale individual represents a solution. Implement foraging behaviors in the whale optimization algorithm such as surrounding prey, bubble net attack, and searching for prey. According to the foraging behaviors of the whales, continuously update the positions of the whale individuals, that is, continuously search for new solutions. Define a fitness function to evaluate the quality of the whale individuals, that is, the quality of the solutions. Repeat the process of simulating the foraging behaviors of the whales and updating the positions of the whale individuals until the preset number of iterations is reached. According to the evaluation results of the fitness function, determine the final whale individual as the final solution. According to the final solution, generate specific optimization suggestions, such as device adjustment plans, process flow improvement measures, and maintenance plan optimization plans, etc. Present the optimization suggestions in the form of reports or plans for easy implementation by decision-makers and operators.

[0142] Suppose there is a hydrogen compression device in the liquid hydrogen hydrogenation station, and data analysis is required to optimize its operation efficiency and predict faults.

[0143] Collect real-time sensor data such as temperature, pressure, and flow rate of the compression equipment, as well as the operating status data of the equipment. Extract features such as the average operating time, failure frequency, and energy consumption level of the equipment. Calculate operating efficiency indicators such as the utilization rate, energy consumption efficiency, and process stability of the equipment. Generate an operating efficiency report to display the calculation results, trend analysis, and problem point identification content of the indicators. Construct a fault prediction model using the operating efficiency indicators, historical fault record data, and real-time monitoring data. Train the model and verify its performance to ensure that it can accurately predict the future fault types and probabilities. Take improving the operating efficiency of the compression equipment and reducing the failure rate as the optimization goals. Define the optimization problem and map the solution space to the search space of the whale population. Simulate the foraging behavior of whales, continuously update the positions of whale individuals, and evaluate their advantages and disadvantages through the fitness function.

[0144] Determine the optimal whale individual as the final solution through an iterative process. Generate optimization suggestions based on the final solution, such as adjusting the operating parameters of the compression equipment, improving the process flow, and optimizing the maintenance plan.

[0145] Through real-time data analysis and calculation of operating efficiency indicators, the operating efficiency of the liquid hydrogen filling station can be accurately evaluated, and efficiency bottleneck problems can be discovered and solved in a timely manner. The implementation of optimization suggestions can further improve the utilization rate and energy consumption efficiency of the equipment and reduce the operating cost. The construction of the fault prediction model can accurately predict the future fault types and probabilities, providing strong support for fault prevention. By taking preventive maintenance measures or adjusting operating parameters in advance, the occurrence of faults can be effectively avoided, reducing the downtime and maintenance cost. The operating efficiency report and optimization suggestions provide scientific data support and decision-making basis for decision-makers, helping to make more reasonable decisions. Through the optimization process driven by data analysis, the scientific nature and accuracy of decisions can be ensured, and the overall operation level of the liquid hydrogen filling station can be improved.

[0146] In a preferred embodiment of the present invention, the calculation formula of the fitness function is:

[0147] ;

[0148] where represents the comprehensive performance of the equipment in period ; , , , represent the weight coefficients; represents the actual operating time of the equipment within period ; represents the planned operating time of the equipment within period ; represents the unplanned downtime; represents the planned operating time; Indicates a period The theoretical output of the equipment within; Indicates the theoretical output; Indicates the actual output; Indicates during the period The theoretical energy consumption of the equipment within; Indicates the input energy; Indicates the output energy; Indicates during the period The quality coefficient of the equipment within; Indicates the number of defective products; Indicates the total number of products; Indicates a period The failure rate of the equipment within; Indicates the severity coefficient of the failure consequence.

[0149] In the embodiment of the present invention, the equipment operation data of the liquid hydrogen refueling station at different periods (such as daily, weekly, monthly) is collected, including the actual operation time ( ), the planned operation time ( ), the unplanned downtime ( ), the theoretical output ( ), the actual output ( ), the theoretical energy consumption ( ), the input energy ( ), the output energy ( ), the quality coefficient ( ), the number of defective products ( ), the total number of products ( ), the failure rate ( ), and the severity coefficient of the failure consequence ( ). Ensure the accuracy and integrity of the data, and appropriately process the missing or abnormal data. According to the specific operation objectives and key points of the liquid hydrogen refueling station, determine the weight coefficients , , , .

[0150] Use the collected data and the determined weight coefficients to calculate according to the formula of the fitness function. Calculate the ratio of the actual operation time to the planned operation time, and consider the impact of the unplanned downtime. Calculate the difference between the theoretical output and the actual output, and consider the impact of the energy consumption efficiency. Calculate the quality coefficient, and consider the impact of the number of defective products. Calculate the failure rate, and consider the impact of the severity coefficient of the failure consequence. Weight and sum the results of the above parts according to the weight coefficients to obtain the comprehensive performance of the equipment during the period ​Analyze the calculated fitness function values and compare the comprehensive performance changes in different periods. Identify the key factors affecting the comprehensive performance, such as running time, output, energy consumption, quality, and failure rate, etc. According to the analysis results, formulate optimization measures, such as adjusting the operation plan, improving the process flow, strengthening equipment maintenance, etc.

[0151] The fitness function comprehensively considers multiple aspects such as the running time, output, energy consumption, quality, and failure rate of the equipment, and can comprehensively evaluate the comprehensive performance of the equipment. By regularly calculating the fitness function values, the changing trend of the equipment performance can be discovered in a timely manner, providing a basis for decision-making. By analyzing the running time ratio part in the fitness function, the running efficiency of the equipment can be evaluated, providing a basis for optimizing the equipment operation plan.

[0152] According to the difference between the actual running time and the planned running time, the equipment operation plan can be adjusted to improve the equipment utilization rate. The output part in the fitness function considers the difference between the theoretical output and the actual output, as well as the influence of the energy consumption efficiency. By analyzing this part of the data, the output bottleneck and energy consumption waste problems can be identified, and corresponding measures can be taken to increase the output and energy consumption efficiency. The quality coefficient part in the fitness function considers the influence of the number of defective products. By analyzing this part of the data, the source of quality problems can be identified, and corresponding measures can be taken to improve the product quality. The failure rate part in the fitness function considers the influence of the severity coefficient of the failure consequences. By analyzing this part of the data, the frequently failing areas and the failure types with serious consequences can be identified, and corresponding measures can be taken to reduce the failure rate and the failure consequences.

[0153] In a preferred embodiment of the present invention, according to the actual operation conditions of the liquid hydrogen filling station and the changes in the external environment, including temperature, humidity, and equipment aging, the control scheme can be automatically adjusted, which may include:

[0154] Define fuzzy sets for temperature, humidity, equipment status, and control parameter adjustment amount;

[0155] Determine the membership functions for each fuzzy set and formulate fuzzy rules;

[0156] According to the fuzzy sets, membership functions, and fuzzy rules, infer the fuzzy values of the output variables;

[0157] Convert the temperature, humidity, and equipment status data monitored in real time into fuzzy values through the membership functions, and use the fuzzy rules for inference to obtain the fuzzy values of the control parameter adjustment amount;

[0158] Convert the fuzzy values of the control parameter adjustment amount into specific numerical values through the defuzzification method, and automatically adjust the control scheme when it is detected that the temperature, humidity, or equipment status has changed.

[0159] In the embodiments of the present invention, fuzzy sets such as low temperature, medium temperature, and high temperature are defined to represent different ranges of air temperature. For example, low temperature represents an air temperature below 0°C, medium temperature represents the range between 0°C and 30°C, and high temperature represents an air temperature above 30°C. Fuzzy sets such as low humidity, medium humidity, and high humidity are defined to represent different levels of humidity. For example, low humidity may represent a humidity below 30%, medium humidity represents the range between 30% and 70%, and high humidity represents a humidity above 70%. Fuzzy sets such as good, general, and poor are defined to represent the operating states of the equipment. Good represents that the equipment has no faults and excellent performance, general represents that the equipment has slight wear or slightly decreased performance, and poor represents that the equipment has obvious faults or severely decreased performance. Fuzzy sets such as small adjustment, medium adjustment, and large adjustment are defined to represent the adjustment ranges of control parameters. Small adjustment represents an adjustment amount within 5%, medium adjustment represents the adjustment amount between 5% and 20%, and large adjustment represents an adjustment amount above 20%. Appropriate membership functions, such as triangular, trapezoidal, or Gaussian functions, are selected for each fuzzy set. For example, for the low-temperature set of air temperature, a triangular membership function with a left endpoint of 0°C and a right endpoint of 10°C (or lower) can be selected, indicating that the lower the air temperature, the higher the membership degree belonging to the low-temperature set. According to the actual operating experience and expert knowledge of the liquid hydrogen refueling station, fuzzy rules are formulated. For example, if the air temperature is high and the equipment state is poor, then the control parameters need to be adjusted greatly; if the air temperature is medium and the equipment state is good, then only small adjustments to the control parameters are required.

[0160] The real-time monitored data of air temperature, humidity, and equipment state are input into the fuzzy system. Through the membership function, these data are converted into fuzzy values. According to the fuzzy rules, fuzzy inference is carried out to obtain the fuzzy value of the control parameter adjustment amount. The air temperature, humidity, and equipment state are monitored in real time. For example, the air temperature is 25°C, the humidity is 60%, and the equipment state is general. Through the membership function, these data are converted into fuzzy values. For example, the air temperature belongs to the medium-temperature set, the humidity belongs to the medium-humidity set, and the equipment state belongs to the general set. Fuzzy rules are used for inference to obtain the fuzzy value of the control parameter adjustment amount, such as medium adjustment.

[0161] An appropriate defuzzification method, such as the centroid method, is selected. The fuzzy value of the control parameter adjustment amount is converted into a specific numerical value. For example, medium adjustment may correspond to an adjustment amount of 10%. When changes in air temperature, humidity, or equipment state are detected, the control scheme is automatically adjusted according to the converted specific numerical value.

[0162] Suppose the current temperature of the liquid hydrogen refueling station is 28°C, the humidity is 55%, and the equipment status is normal. The temperature of 28°C belongs to the medium temperature set, and the membership degree is 0.8 (assuming that the membership function of the medium temperature set changes linearly between 25°C and 30°C, and the membership degree is 0.8 at 28°C). The humidity of 55% belongs to the medium humidity set, and the membership degree is 0.6 (assuming that the membership function of the medium humidity set changes linearly between 40% and 70%, and the membership degree is 0.6 at 55%). The normal equipment status belongs to the normal set, and the membership degree is 0.9 (assuming that the membership function of the normal set changes linearly between 0.7 and 1.0, and the membership degree is 0.9 in the normal state).

[0163] If the temperature is medium and the equipment status is normal, the control parameters need to be adjusted moderately. Since the influence of humidity on moderate adjustment is small (or not clearly mentioned in the rules), the temperature and equipment status are mainly considered.

[0164] The fuzzy value of the control parameter adjustment amount obtained by reasoning is moderate adjustment. Using the weighted average method for defuzzification, the specific value corresponding to moderate adjustment is 12% (assuming the range of moderate adjustment is 5% to 20%, and the membership degree changes linearly within this range, and the center point of moderate adjustment is 12.5%, but considering the influence of the membership degree, the final result is 12%). According to the specific value of 12% corresponding to moderate adjustment, the control parameters of the liquid hydrogen refueling station are automatically adjusted, such as adjusting the operating frequency of the compressor, the flow rate of the cooling system, etc.

[0165] Through fuzzy control, the control parameters can be adjusted more precisely according to the changes in temperature, humidity, and equipment status, improving the operation efficiency and stability of the liquid hydrogen refueling station. Fuzzy control can handle uncertainty and ambiguity, adapting to factors such as changes in the external environment and equipment aging. Fuzzy control has a low dependence on the model and does not require an accurate mathematical model, so it can enhance the robustness of the system and resist external interference and internal changes.

[0166] Through fuzzy rules and membership functions, complex control logic can be simplified, making the control process more intuitive and easy to understand. Fuzzy control can automatically adjust control parameters according to real-time monitoring data, reducing manual intervention and improving the automation level of the liquid hydrogen refueling station. By precisely controlling the parameters, the utilization of energy and raw materials can be optimized, reducing operating costs and improving economic benefits.

[0167] In a preferred embodiment of the present invention, the calculation formula of the membership function:

[0168] ;

[0169] Among them, represents the degree to which the variable belongs to the fuzzy set ; Represents a variable; Represents the base value of the left endpoint; Represents the offset of the left endpoint; Represents the exponent of the left half; Represents the vertex; Represents the base value of the right endpoint; Represents the offset of the right endpoint; Represents the scaling factor of the left half; Represents the scaling factor of the right half; Represents the exponent of the right half.

[0170] In the embodiments of the present invention, its various parameters are determined: Is the base value of the left endpoint, representing the point where the fuzzy set starts to take effect. Is the offset of the left endpoint, used to adjust the position of the left endpoint to make the membership function more flexible. Is the exponent of the left half, affecting the shape of the left half curve and determining the speed of membership growth or decay. Is the vertex, representing the point where the membership reaches the maximum value. Is the base value of the right endpoint, representing the point where the fuzzy set ends to take effect. Is the offset of the right endpoint, used to adjust the position of the right endpoint. Is the scaling factor of the left half, used to adjust the size of the left half membership. Is the scaling factor of the right half, used to adjust the size of the right half membership. These parameters need to be determined according to the specific fuzzy set and the actual application scenario.

[0171] For a given variable , calculate its degree of belonging to the fuzzy set according to the formula of the membership function . When , = 0, indicating that completely does not belong to the fuzzy set . When , use the formula of the left half to calculate the membership, that is = . This formula indicates that as increases from to , the membership gradually increases, and the increasing speed is controlled by the exponent , and finally the membership reaches at the point (assuming that reaches the maximum value when it is 1, otherwise it is scaled proportionally). When When, the membership degree is calculated using the formula in the right half, that is . This formula indicates that as increases from to , the membership degree gradually decreases, and the rate of decrease is controlled by the exponent . Eventually, the membership degree becomes 0 (or close to 0, depending on the value of ) at the point. When , = 0, indicating that completely does not belong to the fuzzy set . After calculating the membership degree, it can be used in the fuzzy inference process. For example, in the control system of a liquid hydrogen refueling station, the real-time monitored temperature, humidity, or equipment status data can be converted into fuzzy values through the membership function, and then these fuzzy values are used for fuzzy inference to obtain the fuzzy value of the control parameter adjustment amount.

[0172] By using the detailed membership function formula, the degree to which a variable belongs to a fuzzy set can be described more precisely. This helps to improve the accuracy of fuzzy control and enables the control system to respond more accurately to changes in the external environment. Multiple parameters in the membership function (such as , , , , , m, , , ) provide great flexibility and can be adjusted according to specific application scenarios and requirements. This enables the membership function to adapt to different control requirements and system characteristics. The membership function can handle uncertainty and ambiguity. Even when system parameters or environmental conditions change, the membership function can still provide reasonable control outputs. The membership function is an important part of fuzzy control, and its application promotes the development of intelligent control.

[0173] An embodiment of the present invention also provides a distributed control method for a liquid hydrogen refueling station, including:

[0174] After preprocessing and feature extraction of the control instruction, it is classified by the trained classification model to obtain a classification result; according to the classification result, an execution plan is determined and the instruction is issued, and the operation data of the liquid hydrogen refueling station is collected in real time through the control node and the sensor;

[0175] Monitor the operation status and process parameters of each device in the liquid hydrogen refueling station in real time,

[0176] When an abnormal situation or potential risk is detected, including equipment failure and process parameter overrun, a warning signal is issued, and corresponding treatment measure suggestions are generated;

[0177] Control specific equipment or areas in a liquid hydrogen refueling station, including liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen dispensers, and safety monitoring equipment;

[0178] Analyze real-time sensor data, equipment operation status data, process flow parameter data, historical fault record data, and environmental monitoring data collected during the operation of the liquid hydrogen refueling station, and generate an operation efficiency report, fault prediction, and optimization suggestions based on the analysis results;

[0179] Automatically adjust the control scheme according to the actual operation conditions of the liquid hydrogen refueling station and external environmental changes, including temperature, humidity, and equipment aging.

[0180] It should be noted that this method corresponds to the above system, and all implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0181] The above is the preferred implementation mode of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A distributed control system for a liquid hydrogen refueling station, characterized in that: include: The central control unit module is used to pre-process and extract features of the control instructions, and then classify them through the trained classification model to obtain classification results; Determine the execution plan and issue instructions based on the classification results, and collect the operating data of the liquid hydrogen refueling station in real time through control nodes and sensors; Fault diagnosis and early warning module, used to monitor the operating status and process parameters of each equipment in the liquid hydrogen refueling station in real time. When abnormal conditions or potential risks are detected, including equipment failure and process parameter exceeding the limit, early warning signals are issued and corresponding treatment measures are recommended; Distributed control node modules are used to control specific equipment or areas within a liquid hydrogen refueling station, including liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen refueling machines, and safety monitoring equipment; The analysis module is used to analyze the real-time sensor data, equipment operation status data, process parameter data, historical fault record data and environmental monitoring data collected during the operation of the liquid hydrogen refueling station, and generate operation efficiency reports, fault predictions and optimization suggestions based on the data analysis results, including: continuously collecting real-time sensor data during the operation of the liquid hydrogen refueling station, including key parameters such as temperature, pressure, flow, and liquid level, and recording the operation status of the equipment, including startup, shutdown, fault status changes, operation time and maintenance records; extracting features that affect operation efficiency, fault prediction and optimization suggestion generation from real-time sensor data and equipment operation status data, including the average operation time, fault frequency and energy consumption level of the equipment; calculating the operation efficiency indicators of the liquid hydrogen refueling station based on the extracted features, including equipment utilization, energy efficiency and process stability, and generating an operation efficiency report, It includes the calculation results of efficiency indicators, trend analysis, and problem point identification content; based on the operating efficiency indicators, historical fault record data and real-time monitoring data, a fault prediction model is constructed to predict the type and probability of future faults; improving the operating efficiency of liquid hydrogen refueling stations and reducing the failure rate are taken as optimization goals, and the optimization problem is defined, and the solution space of the optimization problem is mapped to the search space of the whale group, and each individual whale represents a solution; by simulating the foraging behavior of whales, including surrounding prey, bubbling net attack and searching for prey, the position of individual whales is continuously updated, and the fitness function is used to evaluate the pros and cons of individual whales, and the process of simulating the foraging behavior of whales and updating the position of individual whales is repeated until the preset number of iterations is reached to determine the final solution; based on the final solution, optimization suggestions are generated, including suggestions on equipment adjustment, process improvement and maintenance plan optimization; the calculation formula of the fitness function is: ; in, Indicates that the device is in period Comprehensive performance; , , , represents the weight coefficient; Indicates that during the period The actual operating time of the equipment within; Indicates that during the period The planned operating time of the equipment within; Indicates unplanned downtime; Indicates the planned running time; Indicates the period Theoretical output of internal equipment; represents the theoretical output; Indicates the actual output; Indicates that during the period Theoretical energy consumption of internal equipment; represents input energy; Indicates output energy; Indicates that during the period Quality factor of internal equipment; Indicates the number of defective products; Indicates the total product quantity; Indicates the period Failure rate of internal equipment; A coefficient indicating the severity of the consequences of a failure; The adaptive control function module is used to automatically adjust the control scheme according to the actual operation of the liquid hydrogen refueling station and changes in the external environment, including temperature, humidity and equipment aging.

2. The distributed control system of the liquid hydrogen refueling station according to claim 1, characterized in that: After preprocessing and feature extraction of the control instructions, they are classified by the trained classification model to obtain classification results; Determine the execution plan and issue instructions based on the classification results, and collect the operating data of the liquid hydrogen refueling station in real time through control nodes and sensors, including: Preprocess the control instructions and extract instruction features, including instruction length, keyword frequency and instruction structure; Using the historical control instruction data set to train the preset classification model to obtain a trained classification model; The extracted instruction features are input into the trained classification model, and the types of instructions are classified according to the input feature vector, including equipment control instructions, process control instructions or safety monitoring instructions, to obtain a classified instruction set; Formulate instruction execution plans based on the priority, equipment load, process progress and safety status of the instructions in the classified instruction set; According to the instruction execution plan, the control instructions are sent to the corresponding distributed control nodes, and the operating status data of the liquid hydrogen refueling station is collected in real time through the distributed control nodes and sensors, including the current operating status of the equipment, real-time parameters of the process flow and relevant data of environmental monitoring.

3. The distributed control system of the liquid hydrogen refueling station according to claim 2 is characterized in that: The preset classification model is trained using the historical control instruction data set to obtain a trained classification model, including: Acquire a historical control instruction data set, wherein the instruction data set includes equipment control instructions, process flow control instructions, and safety monitoring instructions; Use the neural network as the preset classification model and initialize the parameters of the classification model, including setting the initial learning rate and number of iterations; During the classification model training process, the scope of parameter updates is limited by the trust region method; Determine the initial size of the trust region, and calculate the direction and size of the parameter update based on the parameters of the current classification model and the size of the trust region; Use the historical control instruction dataset to train the classification model and calculate the value of the objective function; Adjust and update the parameters of the classification model according to the value of the objective function and the limits of the trust region; After a preset number of iterative training, a trained classification model is obtained.

4. The distributed control system of the liquid hydrogen refueling station according to claim 3 is characterized in that: The objective function value is calculated as: ; in, represents the loss function; represents the total number of samples; Indicates the sample index; Indicates the total number of categories; , Represents the category index; Indicates category frequency; Representation sample For categories The weight coefficient of Representation sample Belongs to category The true label of Represents the classification model for samples Belongs to category The original output of Representation sample For categories Bias adjustment term; Represents the classification model for samples Belongs to category The original output of Representation sample For categories Bias adjustment term; represents the regularization coefficient; represents the square of the classification model parameter; Represents the base of natural logarithms.

5. The distributed control system of the liquid hydrogen refueling station according to claim 4, characterized in that: Real-time monitoring of the operating status and process parameters of each device in the liquid hydrogen refueling station. When abnormal conditions or potential risks are detected, including equipment failure and process parameter exceeding the limit, an early warning signal is issued and corresponding treatment measures are generated, including: Continuously analyze real-time monitoring equipment and process parameters to detect the real-time status of equipment and process parameters by comparing current data with the preset normal operating range or mode; Set a threshold range for each process parameter. When a process parameter exceeds the threshold range, it is considered abnormal and triggers the early warning mechanism to generate an early warning signal. The early warning signal includes the abnormal type, occurrence time, and related equipment or parameter information. According to the abnormal type in the early warning signal, the corresponding treatment measures are determined from the emergency treatment plan library, and treatment measure suggestions are generated, including operation steps, a list of required tools or equipment, and safety precautions information.

6. The distributed control system of the liquid hydrogen refueling station according to claim 5, characterized in that: According to the actual operation of the liquid hydrogen refueling station and external environmental changes, including temperature, humidity and equipment aging, the control scheme is automatically adjusted, including: Define fuzzy sets for air temperature, humidity, equipment status, and control parameter adjustments; Determine the membership function for each fuzzy set and formulate fuzzy rules; Infer the fuzzy values ​​of output variables based on fuzzy sets, membership functions and fuzzy rules; The real-time monitored temperature, humidity and equipment status data are converted into fuzzy values ​​through membership functions, and fuzzy rules are used for reasoning to obtain the fuzzy values ​​of the control parameter adjustment quantities; The fuzzy value of the control parameter adjustment amount is converted into a specific numerical value through the defuzzification method. When changes in temperature, humidity or equipment status are detected, the control plan is automatically adjusted.

7. The distributed control system of the liquid hydrogen refueling station according to claim 6, characterized in that: The calculation formula of membership function is: ; in, Representation variables Belongs to fuzzy sets the extent of; Represents a variable; Indicates the base value of the left endpoint; Indicates the offset of the left endpoint; represents the index of the left half; represents a vertex; Indicates the base value of the right endpoint; Indicates the offset of the right endpoint; Indicates the scaling factor of the left half; Indicates the scaling factor of the right half; Indicates the exponent of the right half.

8. A distributed control method for a liquid hydrogen refueling station, the method realizing the system as claimed in any one of claims 1 to 7, characterized in that: include: After preprocessing and feature extraction of the control instructions, they are classified by the trained classification model to obtain classification results; Determine the execution plan and issue instructions based on the classification results, and collect the operating data of the liquid hydrogen refueling station in real time through control nodes and sensors; Real-time monitoring of the operating status and process parameters of each equipment in the liquid hydrogen refueling station. When abnormal conditions or potential risks are detected, including equipment failure and process parameter exceeding the limit, early warning signals are issued and corresponding treatment measures are recommended; Control specific equipment or areas within a liquid hydrogen refueling station, including liquid hydrogen storage equipment, hydrogen compression equipment, hydrogen refueling machines, and safety monitoring equipment; Analyze the real-time sensor data, equipment operation status data, process parameter data, historical fault record data and environmental monitoring data collected during the operation of the liquid hydrogen refueling station, and generate operation efficiency reports, fault predictions and optimization suggestions based on the data analysis results; The control scheme is automatically adjusted according to the actual operation of the liquid hydrogen refueling station and changes in the external environment, including temperature, humidity and equipment aging.

Images