Orthohydrogen and parahydrogen conversion temperature control method

Abstract

The invention relates to the technical field of hydrogen liquefaction, and discloses an ortho-parahydrogen conversion temperature control method, which processes and judges whether temperature control is needed or not through a target neural network model, realizes an intelligent temperature control decision, avoids unnecessary control operation, and improves the system operation efficiency. Finally, when control is needed, target inflow control parameters and system advanced starting time can be accurately calculated by means of a genetic algorithm and an advanced amount calculation model, then accurate control over the ortho-parahydrogen conversion temperature can be achieved through the target inflow control parameters and the system advanced starting time, conversion from parahydrogen to ortho-hydrogen is effectively restrained, and the normal hydrogen conversion rate is improved. The consumption of liquid helium is reduced, the stability and high efficiency of the hydrogen liquefaction process are ensured, and the storage efficiency of hydrogen is improved. Meanwhile, control can be started in advance through the system in advance starting time, system response delay is compensated, temperature overshoot is prevented, and positive significance is achieved for ortho-parahydrogen conversion and hydrogen liquefaction efficiency improvement.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen liquefaction, and in particular to a method for controlling the temperature of normal-para hydrogen conversion. Background Art

[0002] Orthohydrogen and parahydrogen, two spin isomers of molecular hydrogen, play a key role in the field of hydrogen liquefaction and storage. Parahydrogen has a lower evaporation rate, so during the hydrogen liquefaction process, preferentially converting orthohydrogen into parahydrogen becomes the preferred strategy to improve storage convenience. However, the spontaneous conversion of orthohydrogen to parahydrogen releases conversion heat, which in turn increases the energy consumption required for hydrogen liquefaction. Not only that, in the actual hydrogen liquefaction process, parahydrogen also has the phenomenon of spontaneous reverse conversion into orthohydrogen, and parahydrogen also has the phenomenon of spontaneous reverse conversion into orthohydrogen. If the temperature is not properly controlled, the parahydrogen content will be significantly reduced, seriously affecting the storage efficiency of hydrogen.

[0003] However, temperature control during the conversion of ortho-parahydrogen currently faces many challenges. Traditional methods have a low level of intelligence and it is difficult to accurately control the conversion process, resulting in frequent fluctuations in temperature regulation. Summary of the Invention

[0004] In view of this, the present invention provides a temperature control method for ortho-parahydrogen conversion to solve the problem that the temperature control process of the existing ortho-parahydrogen conversion process is difficult to accurately control, resulting in frequent fluctuations in temperature regulation, which in turn leads to a significant reduction in parahydrogen content and seriously affects the storage efficiency of hydrogen.

[0005] In a first aspect, the present invention provides a method for controlling the temperature of normal-para hydrogen conversion, which is used in an intelligent control system for the temperature of normal-para hydrogen conversion; the method comprises:

[0006] A real-time temperature data set and a real-time system status data set for normal-parahydrogen conversion during the hydrogen liquefaction process are obtained; the real-time temperature data set and the real-time system status data set are processed by a target neural network model to obtain a real-time temperature change data set; whether temperature control is required is determined based on the real-time temperature change data set; when temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the real-time system status data set through genetic algorithm and advance calculation model processing; the normal-parahydrogen conversion temperature is controlled according to the system advance start time and the target inflow control parameters to obtain a temperature control result.

[0007] The para-hydrogen conversion temperature control method provided by the present invention can grasp the temperature and system status information of the para-hydrogen conversion process in real time by acquiring a real-time temperature data set and a system real-time status data set. Further, by processing and judging whether temperature control is needed through the target neural network model, intelligent temperature control decision-making is realized, unnecessary control operations are avoided, and the system operation efficiency is improved. Finally, when control is needed, with the help of genetic algorithm and advance amount calculation model, the target inflow control parameter and the system advance start time can be accurately calculated, and then by the target inflow control parameter and the system advance start time, the para-hydrogen conversion temperature can be accurately controlled, effectively suppressing the conversion of para-hydrogen to ortho-hydrogen, reducing liquid helium consumption, ensuring the stability and efficiency of the hydrogen liquefaction process, and improving the storage efficiency of hydrogen. At the same time, the control can be started in advance through the system advance start time, compensating for the system response delay, preventing temperature overshoot, and having positive significance for improving the para-hydrogen conversion and hydrogen liquefaction efficiency.

[0008] In an optional embodiment, the method further includes:

[0009] Obtain a historical temperature data set and a system historical status data set; and construct a target neural network model based on the historical temperature data set and the system historical status data set.

[0010] In an optional embodiment, a target neural network model is constructed based on a historical temperature dataset and a system historical state dataset, including:

[0011] According to the historical temperature data set and the system historical status data set, a historical temperature change data set is obtained through a preset calculation method; the target neural network model is constructed with the historical temperature data set and the system historical status data as input and the historical temperature change data set as output.

[0012] The temperature control method for ortho-parahydrogen conversion provided by the present invention provides a rich set of training samples for neural network model construction by acquiring a historical temperature dataset and a historical system state dataset. Furthermore, a historical temperature change dataset is calculated based on the historical temperature dataset and the historical system state dataset. A model is constructed using the historical temperature dataset and the historical system state data as input and the historical temperature change dataset as output. This allows the constructed target neural network model to learn the inherent relationship between temperature changes and system states during ortho-parahydrogen conversion, thereby accurately processing and predicting real-time data and improving the output accuracy of the temperature change dataset.

[0013] In an optional embodiment, the method further includes: verifying the target neural network model using a dual-supervision mechanism.

[0014] The temperature control method for ortho-parahydrogen conversion provided by the present invention verifies the target neural network model through a dual-supervision mechanism, effectively avoiding large deviations in the model, thereby ensuring that the model's prediction results are true and reliable, and enhancing the stability and credibility of the temperature control method.

[0015] In an optional embodiment, a dual-supervision mechanism is used to verify the target neural network model, including:

[0016] The physical temperature change value is determined using a dual-supervision mechanism; the predicted temperature change value output by the target neural network model is obtained; a loss function is constructed based on the physical temperature change value and the predicted temperature change value; and the accuracy of the target neural network model is verified based on the loss function.

[0017] In an optional embodiment, when temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the system real-time status data set through genetic algorithm and advance calculation model processing, including:

[0018] When temperature control is required, the real-time temperature change data set is processed by genetic algorithm to obtain the target inflow control parameters; the temperature change information set and the system real-time status data set are input into the advance calculation model to obtain the system advance start time.

[0019] The para-hydrogen conversion temperature control method provided by the present invention uses a genetic algorithm to iteratively optimize a multitude of potential parameters, obtaining the optimal target inflow control parameter. This allows the liquid helium or cold helium flow rate to precisely match the temperature regulation requirements, improving temperature control accuracy while reducing flow consumption and energy waste. Furthermore, by integrating temperature changes and system status data and processing them through a lead calculation model, a reasonable lead start time can be calculated, providing data support for subsequent precise control of the para-hydrogen conversion temperature.

[0020] In an optional embodiment, when temperature control is required, the real-time temperature change data set is processed by a genetic algorithm to obtain target inflow control parameters, including:

[0021] Obtain a preset fitness function, which is used to characterize temperature stability and the degree of proximity between the actual temperature and the target temperature. When temperature control is required, input the real-time temperature change data set into the preset fitness function, and optimize the inflow control parameters until the target inflow control parameters are obtained.

[0022] In an optional embodiment, when temperature control is required, the real-time temperature change data set is input into a preset fitness function, and the inflow control parameters are optimized until the target inflow control parameters are obtained, including:

[0023] When temperature control is required, the inflow control parameters are used to encode individuals and generate an initial population; the real-time temperature change data set is input into a preset fitness function, and each individual in the initial population is evaluated to obtain multiple individual fitness values. The preset fitness function is used to characterize temperature stability and the degree of proximity between the actual temperature and the target temperature; the initial population is genetically operated based on the multiple individual fitness values and the initial population is updated; based on the updated initial population, individual evaluation and genetic operations are repeatedly performed until the termination condition is met and the iteration is stopped and the target inflow control parameters are obtained.

[0024] The temperature control method for ortho-parahydrogen conversion provided by the present invention uses inflow control parameters as individuals to encode an initial population and utilizes genetic algorithms to perform a global search within a large parameter space. This significantly increases the probability of finding the optimal solution, avoids being trapped in a local optimum, and ultimately achieves target inflow control parameters that are more suitable for current temperature control requirements. Furthermore, by evaluating individuals based on their temperature stability and proximity to the target temperature, the algorithm can adaptively select and evolve individuals based on the actual temperature control conditions. This, in turn, enhances the adaptability and effectiveness of temperature control by continuously optimizing parameters.

[0025] In an optional embodiment, the method further includes:

[0026] Obtain a historical temperature change information set and a system historical state data set; based on the historical temperature change information set and the system historical state data set, after processing using the law of conservation of energy and the principle of heat transfer, construct an advance calculation model.

[0027] The para-hydrogen conversion temperature control method provided by the present invention combines the law of conservation of energy and the principle of heat transfer to construct a model, so that the model has a solid physical theoretical basis and can accurately reflect the relationship between heat transfer and temperature changes during the para-hydrogen conversion process, thereby providing support for ensuring that the subsequent calculated advance start time is scientific and reasonable.

[0028] In an optional embodiment, based on the historical temperature change information set and the system historical state data set, after processing according to the law of conservation of energy and the principle of heat transfer, a lead calculation model is constructed, including:

[0029] Based on the historical temperature change information set and the system historical state data set, an energy balance model is constructed after processing according to the law of conservation of energy; based on the historical temperature change information set and the system historical state data set, a heat transfer model is constructed after processing according to the principle of heat transfer; based on the energy balance model and the heat transfer model, the advance calculation model is determined. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 1 is a flow chart of a method for controlling the temperature of normal-parahydrogen conversion according to an embodiment of the present invention;

[0032] Figure 2 1 is a schematic flow chart of another method for controlling temperature of normal-para hydrogen conversion according to an embodiment of the present invention;

[0033] Figure 3 1 is a flow chart of another method for controlling the temperature of normal-para hydrogen conversion according to an embodiment of the present invention;

[0034] Figure 4 2. It is a control flow diagram of an intelligent control system for normal-parahydrogen conversion temperature according to an embodiment of the present invention;

[0035] Figure 5 4 is a flow chart of a method for predicting temperature changes using a neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0037] The embodiment of the present invention provides a method for controlling the temperature of ortho-parahydrogen conversion. By combining a target neural network model, a genetic algorithm, and an advance calculation model, the method achieves precise control of the ortho-parahydrogen conversion temperature, effectively inhibits the conversion of parahydrogen to orthohydrogen, reduces liquid helium consumption, ensures the stability and efficiency of the hydrogen liquefaction process, and improves the storage efficiency of hydrogen.

[0038] According to an embodiment of the present invention, an embodiment of a method for controlling the temperature of normal-parahydrogen conversion is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] In this embodiment, a method for controlling the temperature of normal-parahydrogen conversion is provided, which can be used in an intelligent control system for the temperature of normal-parahydrogen conversion. The intelligent control system for the temperature of normal-parahydrogen conversion can include a liquid helium pump, a converter, and multiple temperature measurement devices with different functions. Figure 1 : is a flow chart of a method for controlling the temperature of normal-parahydrogen conversion according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0040] Step S101, obtaining a real-time temperature data set and a real-time system status data set for normal-parahydrogen conversion during hydrogen liquefaction.

[0041] Among them, the real-time temperature data represents the temperature data set collected in real time at each key position during the normal-parahydrogen conversion process, which is used to reflect the current temperature conditions.

[0042] The system real-time status data set may include multiple real-time status information related to normal and para-hydrogen conversion, such as helium flow rate and equipment operating status.

[0043] Specifically, temperature measuring devices with different functions can be deployed in the converter and related flow paths within the intelligent control system for normal and para-hydrogen conversion temperature to collect temperature data in real time.

[0044] In step S102 , the real-time temperature data set and the system real-time status data set are processed by the target neural network model to obtain a real-time temperature change data set.

[0045] The target neural network model represents a trained network model for processing temperature and system status data and predicting temperature changes.

[0046] The real-time temperature change data set represents a data set that presents the future change trend and characteristics of the normal-parahydrogen conversion temperature after the target neural network model processes the input data, and may include temperature prediction values and temperature change rates, etc.

[0047] Specifically, the acquired real-time temperature data set and system real-time status data set are input into the trained target neural network model, which can be used to predict the temperature change trend in the future period of time and then output the corresponding real-time temperature change data set.

[0048] Step S103: determining whether temperature control is required based on the real-time temperature change data set.

[0049] Specifically, when the neural network model predicts that the temperature has a trend of continuing to rise, it further determines whether this trend will cause the temperature to exceed the set threshold.

[0050] The threshold value can be determined based on the ideal temperature range of the normal-parahydrogen conversion process and the safe operation requirements of the equipment.

[0051] Assuming that the ideal temperature range for the conversion of normal and secondary hydrogen is (T min ,T max ), considering the stability of the system and the tolerance of the equipment, the temperature rise threshold is set to T thresh-up , when the predicted temperature is likely to exceed T max -T thtesh-up When the temperature drops, it is necessary to take measures to control the temperature in advance. Similarly, the temperature drop threshold is set to T thresh-down , when the predicted temperature is likely to be lower than T min +T thresh-down When the controller is in operation, corresponding control operations are also required.

[0052] Furthermore, the temperature prediction value in the real-time temperature change data set is compared with the set threshold. If the temperature prediction is greater than or equal to the set threshold, temperature control is required, otherwise it is not required.

[0053] Step S104: When temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the system real-time status data set through genetic algorithm and advance calculation model processing.

[0054] The Genetic Algorithm (GA) represents a random search algorithm that simulates biological evolution and is used to find the optimal inflow control parameters in the parameter space.

[0055] The advance calculation model represents a mathematical model built based on physical principles and system characteristics, and is used to calculate the advance start time of temperature control.

[0056] The target inflow control parameter represents a liquid helium or cold helium gas inflow parameter that can effectively adjust the normal-para-hydrogen conversion temperature.

[0057] The system advance start time indicates the time when the control device starts in advance, which is used to ensure effective temperature control.

[0058] Specifically, when it is predicted that the temperature has a tendency to exceed a threshold value, that is, temperature control is required, the optimal inflow control parameters can be calculated using a genetic algorithm by combining the real-time temperature change data set and the system real-time status data set.

[0059] Furthermore, the advance calculation model can be used to calculate the time for starting the liquid helium pump in advance or controlling the flow of cold helium gas for heat exchange, that is, the system advance start time.

[0060] Step S105 , controlling the normal-para hydrogen conversion temperature according to the system advance start time and the target inflow control parameters to obtain a temperature control result.

[0061] Specifically, the liquid helium pump can be started in advance or the cold helium flow control device can be adjusted based on the system's pre-start time. The inflow of liquid helium or cold helium can then be precisely controlled according to the target inflow control parameters. Furthermore, after the cold helium and liquid helium enter the normal-para-hydrogen converter, they undergo heat exchange with the hydrogen undergoing the conversion reaction, achieving temperature control.

[0062] Among them, due to its low temperature, cold helium absorbs the heat generated by the conversion of hydrogen to para-hydrogen, thereby lowering the temperature of the hydrogen; after absorbing heat, the state of liquid helium itself may change (such as partial vaporization), further taking away a large amount of heat and achieving effective regulation of the temperature of the conversion area.

[0063] Furthermore, during the control process, the temperature is continuously monitored and new temperature data is fed back, and adjustments and optimizations are made in real time to ultimately obtain the temperature control result and stabilize the temperature within the target range.

[0064] The para-hydrogen conversion temperature control method provided in this embodiment can grasp the temperature and system status information of the para-hydrogen conversion process in real time by acquiring a real-time temperature data set and a real-time system status data set. Further, by processing and judging whether temperature control is required through the target neural network model, intelligent temperature control decision-making is realized, unnecessary control operations are avoided, and the system operation efficiency is improved. Finally, when control is required, with the help of genetic algorithm and advance amount calculation model, the target inflow control parameters and the system advance start time can be accurately calculated, and then the target inflow control parameters and the system advance start time can be used to achieve precise control of the para-hydrogen conversion temperature, effectively inhibit the conversion of para-hydrogen to ortho-hydrogen, reduce liquid helium consumption, ensure the stability and efficiency of the hydrogen liquefaction process, and improve the storage efficiency of hydrogen. At the same time, the control can be started in advance through the system advance start time, which compensates for the system response delay and prevents temperature overshoot, which has positive significance for improving the para-hydrogen conversion and hydrogen liquefaction efficiency.

[0065] In this embodiment, a method for controlling the temperature of normal-parahydrogen conversion is provided, which can be used in an intelligent control system for the temperature of normal-parahydrogen conversion. The intelligent control system for the temperature of normal-parahydrogen conversion can include a liquid helium pump, a converter, and multiple temperature measurement devices with different functions. Figure 2 : is a flow chart of a method for controlling the temperature of normal-parahydrogen conversion according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0066] Step S201: Obtain the real-time temperature data set and system real-time status data set of the normal-parahydrogen conversion during the hydrogen liquefaction process. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0067] Step S202: Acquire a historical temperature data set and a system historical status data set.

[0068] The specific process can refer to the description and acquisition of relevant data in the above step S101, which will not be repeated here.

[0069] Step S203: constructing a target neural network model based on the historical temperature data set and the system historical status data set.

[0070] Specifically, the above step S203 includes:

[0071] Step S2031 : obtaining a historical temperature change data set based on the historical temperature data set and the system historical status data set through a preset calculation method.

[0072] Specifically, the historical temperature dataset and the historical system status dataset can be associated and integrated. For each time point, the temperature, temperature change rate, hydrogen flow rate, cold flow flow rate, and other data at that moment are combined with information such as the normal and para-hydrogen conversion heat and conversion rate (using temperature as a query parameter to obtain the corresponding conversion rate and conversion heat from the database).

[0073] Furthermore, the para-hydrogen conversion rate can be calculated based on the historical temperature data and the obtained para-hydrogen conversion rate, combined with the hydrogen flow data in the system historical status data set.

[0074] Furthermore, the total heat released during the conversion of para-hydrogen can be calculated based on the conversion rate of para-hydrogen and the heat of conversion of para-hydrogen.

[0075] Furthermore, the cold flow rate is obtained from the system historical status data set, and combined with the helium heat transfer coefficient (which can be measured experimentally or determined based on equipment parameters) and the temperature difference between liquid helium and the system (which can be calculated from the historical temperature data set), the total heat carried away by the cold flow during the normal-parahydrogen conversion process is calculated.

[0076] Furthermore, the temperature change can be calculated based on the total heat released during the conversion of normal and para-hydrogen, the total heat removed by the cold flow, and the heat capacity of the converter. The heat capacity of the converter is related to factors such as the amount of hydrogen in it and the material of the equipment, and can be determined through preliminary experimental measurements or based on system design parameters.

[0077] Finally, the corresponding historical temperature change data set is formed by combining the conversion heat of para-hydrogen, conversion rate, total heat released during the conversion of para-hydrogen, total heat carried away by the cold flow, and temperature change values.

[0078] Step S2032: construct a target neural network model with the historical temperature data set and the system historical status data as input and the historical temperature change data set as output.

[0079] First, you can choose a recurrent neural network (RNN) or a long short-term memory network (LSTM) as the basic structure of the neural network, which can process data with time series characteristics and better learn the changes in temperature and system status over time.

[0080] Next, an input layer is constructed to receive historical temperature data sets (including temperature, temperature change rate, etc.) and historical system status data sets (such as hydrogen flow rate, cold flow flow rate, etc.). Simultaneously, multiple hidden layers are set up, and feature extraction and analysis are performed on the input data in these hidden layers to explore potential relationships between the data. Furthermore, an output layer is constructed to output prediction results corresponding to the historical temperature change data sets, such as the heat of conversion to para-hydrogen, the conversion rate, the total heat released during the conversion to para-hydrogen, the total heat removed by the cold flow, and the temperature change value.

[0081] Furthermore, the parameters such as weights and biases in the neural network are initialized.

[0082] Further, the constructed neural network f NN The mapping relationship is shown in the following equation (1):

[0083]

[0084] Where: T t Indicates the temperature value at the current moment; Indicates the rate of temperature change; Indicates the hydrogen flow rate inside the system; Q flow represents the cold flow rate; α represents the conversion rate of normal and para hydrogen; q represents the heat of normal and para hydrogen conversion; ΔT represents the system temperature change value predicted by the neural network.

[0085] Finally, the historical temperature data set and the system historical status data are input into the neural network for training until a trained target neural network model is obtained.

[0086] In an optional implementation, after the above step S204, the method further includes: verifying the target neural network model using a dual-supervision mechanism.

[0087] Among them, the dual supervision mechanism represents a method for verifying the reliability and accuracy of the target neural network model. It ensures the validity of the model output results by constraining and supervising the model output from two different angles.

[0088] Specifically, the target neural network model is verified using a dual-supervision mechanism, including: determining the physical temperature change value using the dual-supervision mechanism; obtaining the predicted temperature change value output by the target neural network model; constructing a loss function based on the physical temperature change value and the predicted temperature change value; and verifying the accuracy of the target neural network model based on the loss function.

[0089] Specifically, for each sample in the historical data, a physically calculated temperature change value ΔT can be calculated based on the physical model and relevant physical laws, such as the law of conservation of energy, heat transfer formula, etc., combined with known system parameters (such as system heat capacity, helium heat transfer coefficient, etc.) and the temperature, flow rate and other data corresponding to the sample. phys .

[0090] Furthermore, by calculating the loss function L = ||ΔT-ΔT phys ||Measure the temperature change value ΔT obtained by physical calculation phys The difference between the temperature change value ΔT output by the target neural network model.

[0091] Finally, the model's prediction accuracy can be judged based on the value of the loss function. If the loss value is large, it means that the model's prediction results deviate significantly from the physical calculated values. In this case, it is necessary to adjust the model's structure, parameters, or training methods, such as adjusting the number of neurons in the hidden layer, changing the learning rate, etc., and then retrain and verify until the difference between the model's prediction results and the physical calculated values is within an acceptable range. The final target neural network model is output. Through continuous adjustment, the effectiveness and reliability of the model can be ensured.

[0092] In step S204, the real-time temperature data set and the system real-time status data set are processed by the target neural network model to obtain a real-time temperature change data set.

[0093] Specifically, according to the description of step S202 to step S203, the obtained real-time temperature data set and system real-time status data set are input into the input layer of the target neural network model.

[0094] Furthermore, the data is extracted and analyzed in the model's hidden layer. By connecting neurons with weights and performing activation functions, the model explores potential features and relationships within the real-time data. Calculations are then performed in the output layer, outputting results related to temperature changes, such as the heat of conversion to para-hydrogen, the conversion rate, the total heat released during the conversion to para-hydrogen, the total heat removed by the cold flow, and the temperature change, forming a corresponding real-time temperature change dataset.

[0095] Step S205: Determine whether temperature control is required based on the real-time temperature change data set. Figure 1Step S103 of the illustrated embodiment will not be described in detail here.

[0096] Step S206: When temperature control is required, the target inflow control parameters and system advance start time are obtained based on the real-time temperature change data set and the system real-time status data set through genetic algorithm and advance calculation model processing. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0097] Step S207: Control the normal-para hydrogen conversion temperature according to the system advance start time and target inflow control parameters to obtain the temperature control result. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0098] The temperature control method for ortho-parahydrogen conversion provided in this embodiment provides a rich set of training samples for neural network model construction by acquiring historical temperature datasets and system historical state datasets. Furthermore, a historical temperature change dataset is calculated based on the historical temperature datasets and system historical state datasets. Model construction is performed using the historical temperature datasets and system historical state data as input and the historical temperature change dataset as output. This allows the constructed target neural network model to learn the inherent laws between temperature changes and system states during ortho-parahydrogen conversion, thereby accurately processing and predicting real-time data and improving the output accuracy of the temperature change dataset. Furthermore, the target neural network model is verified through a dual-supervision mechanism, effectively avoiding large model deviations, thereby ensuring the authenticity and reliability of the model's prediction results and enhancing the stability and credibility of the temperature control method.

[0099] In this embodiment, a method for controlling the temperature of normal-parahydrogen conversion is provided, which can be used in an intelligent control system for the temperature of normal-parahydrogen conversion. The intelligent control system for the temperature of normal-parahydrogen conversion can include a liquid helium pump, a converter, and multiple temperature measurement devices with different functions. Figure 3 : is a flow chart of a method for controlling the temperature of normal-parahydrogen conversion according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0100] Step S301: Obtain the real-time temperature data set and system real-time status data set of normal and para hydrogen conversion in the hydrogen liquefaction process. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0101] Step S302: Process the real-time temperature data set and the system real-time status data set through the target neural network model to obtain a real-time temperature change data set. Figure 2 Step S205 of the illustrated embodiment will not be described in detail here.

[0102] Step S303: Determine whether temperature control is required based on the real-time temperature change data set. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0103] Step S304: When temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the system real-time status data set through genetic algorithm and advance calculation model processing.

[0104] Specifically, the above step S304 includes:

[0105] Step S3041: When temperature control is required, the real-time temperature change data set is processed by a genetic algorithm to obtain a target inflow control parameter.

[0106] Specifically, when temperature control is required, genetic algorithms can iteratively search for the best among numerous potential parameters to obtain the optimal target inflow control parameters, thereby enabling the liquid helium or cold helium gas flow rate to accurately match the temperature regulation requirements, thereby improving temperature control accuracy, reducing flow consumption, and reducing energy waste.

[0107] In an optional implementation, the above step S3041 includes:

[0108] Step b1, obtaining a preset fitness function.

[0109] The preset fitness function Fitness is used to characterize the temperature stability and the degree of closeness between the actual temperature and the target temperature, as shown in the following equation (2):

[0110]

[0111] Where: T actual Indicates the actual temperature value; T target represents the target temperature value; β represents the weight coefficient, which is used to adjust the impact of temperature changes on fitness.

[0112] Step b2: When temperature control is required, the real-time temperature change data set is input into a preset fitness function, and the inflow control parameters are optimized until the target inflow control parameters are obtained.

[0113] In an optional embodiment, the above step b2 includes:

[0114] Step b21: When temperature control is required, the inflow control parameters are used to encode individuals and generate an initial population.

[0115] Step b22: input the real-time temperature change data set into a preset fitness function, and evaluate each individual in the initial population to obtain multiple individual fitness values.

[0116] Step b23: performing genetic operations on the initial population according to the fitness values of the multiple individuals and updating the initial population.

[0117] In step b24, individual evaluation and genetic operation are repeatedly performed according to the updated initial population until the termination condition is met, the iteration is stopped, and the target inflow control parameter is obtained.

[0118] Specifically, when it is determined that temperature control is required, the goal is to find appropriate liquid helium or cold helium inflow control parameters through genetic algorithms so that the normal-parahydrogen conversion temperature reaches and stabilizes within a preset target temperature range.

[0119] First, the liquid helium or cold helium inflow control parameter is encoded to form an individual in the genetic algorithm. For example, binary encoding or real number encoding can be used. Assume that the value range of the inflow control parameter is (0-Q maxl ,0-Q maxg ), if real number coding is used, each individual is a real number within the range, representing a specific inflow value.

[0120] Next, a group of individuals is randomly generated, forming the initial population. Each individual in the population represents a potential solution to the temperature control problem. The size of the initial population is determined by the specific problem and computing resources, and can generally be set to dozens to hundreds of individuals.

[0121] Secondly, the preset fitness function shown in the above relational expression (2) is determined using the temperature stability and the degree of proximity to the target temperature as measurement criteria.

[0122] Furthermore, the real-time temperature change dataset is fed into a pre-set fitness function, which evaluates each individual in the initial population and calculates their fitness. A higher fitness value indicates a better temperature control effect achieved by the inflow control parameters corresponding to that individual, meaning the predicted temperature is closer to the target temperature and the temperature changes are more stable.

[0123] Then, genetic operations are performed on the initial population according to multiple individual fitness values, including:

[0124] (1) Selection operation: Based on the fitness value of the individual, an appropriate selection method (such as roulette selection, tournament selection, etc.) can be used to select individuals with higher fitness from the current population, so that they have a greater probability of entering the next generation population. For example, in roulette selection, the probability of each individual being selected is proportional to its fitness value.

[0125] (2) Crossover operation: Perform a crossover operation on the selected individuals to generate new individuals. Crossover operation is an important way to generate new solutions in genetic algorithms, and it can combine the excellent characteristics of different individuals. For example, for real-number coded individuals, the arithmetic crossover method can be used to randomly select two individuals and linearly combine their gene values according to a certain ratio to generate a new individual.

[0126] (3) Mutation: Mutation is performed on newly generated individuals with a certain probability, randomly changing the values of certain genes in the individual's code. Furthermore, mutation can increase the diversity of the population and prevent the algorithm from falling into a local optimal solution. For example, for individuals with real number codes, a small random perturbation can be added to their gene values.

[0127] Furthermore, individual evaluation and genetic manipulation are repeated, continuously iterating and updating the population. Individuals in each generation evolve continuously under the evaluation of the fitness function, gradually approaching the optimal solution. Furthermore, during the iteration process, termination conditions can be set, such as reaching a preset number of iterations or meeting specific convergence conditions (e.g., minimal change in the optimal solution across several generations).

[0128] Finally, when the termination condition is met, the iteration stops. At this time, the code corresponding to the individual with the highest fitness in the population, after decoding, is the target inflow control parameter calculated by the genetic algorithm.

[0129] Step S3042: Input the temperature change information set and the system real-time status data set into the advance calculation model to obtain the system advance start time.

[0130] Specifically, when it is predicted that the temperature has a tendency to exceed a threshold, the time required to start the liquid helium pump or cold helium flow control for heat exchange in advance can be calculated based on factors such as the current temperature change rate, the thermal response characteristics of the system, and the response time of the control actuator (such as the liquid helium pump), that is, the system advance start time.

[0131] In one embodiment, if the current temperature change rate is The response time from system startup to temperature drop is Δt response , in order to effectively control the temperature before it reaches the threshold, the advance amount Δ tadvance A preliminary estimate can be made using the following relationship (3):

[0132]

[0133] Furthermore, the above equation (3) indicates that the time required for the temperature to rise to the threshold is calculated based on the current temperature rise rate. Subtracting the system response time, we can get the advance startup time required, that is, the system advance startup time.

[0134] The advance calculation model can be constructed through the following steps:

[0135] Step b1: Obtain a historical temperature change information set and a system historical status data set.

[0136] Specifically, the historical temperature change information set may include the temperature value and temperature change rate at the current moment, which can be obtained through the target neural network model.

[0137] Step b2: Based on the historical temperature change information set and the system historical state data set, an advance calculation model is constructed through processing based on the law of conservation of energy and the principle of heat transfer.

[0138] In an optional embodiment, the above-mentioned step b2 includes: based on the historical temperature change information set and the system historical status data set, after processing the law of conservation of energy, constructing an energy balance model; based on the historical temperature change information set and the system historical status data set, after processing the heat transfer principle, constructing a heat transfer model; based on the energy balance model and the heat transfer model, determining the advance calculation model.

[0139] Specifically, the system's heat capacity C, which reflects the system's ability to store heat, the helium heat transfer coefficient h, which represents the ability of helium to exchange heat with the system, and the response time Δt required from issuing a control signal to actually starting to change the flow rate can be determined based on the system's historical status data set. pump , set temperature threshold T thresh wait.

[0140] Among them, the heat capacity C is related to factors such as the amount of hydrogen in the system historical status dataset and the material of the equipment; the helium heat transfer coefficient h is related to factors such as the helium flow rate, flow velocity, and heat exchange area in the system historical status dataset.

[0141] Furthermore, taking into account the above-mentioned temperature change trend, system thermal characteristics, liquid helium heat exchange characteristics, and response characteristics of the control actuator, an equation for the advance amount calculation model is established based on the law of conservation of mass and the principle of heat transfer.

[0142] Specifically, assuming that in time t, the heat generated by the system due to the conversion of para-hydrogen is Q reaction , the heat taken away by the cold helium is Q cooling , the change of system temperature is ΔT, then the following relationship (4) can be obtained:

[0143] Q reaction -Q cooling =CΔT(4)

[0144] Among them, Q reactionIt can be calculated based on the conversion rate and conversion heat of normal and para hydrogen. Assuming the conversion rate of normal and para hydrogen is The heat of conversion of normal and para hydrogen per unit amount of substance is ΔH, then

[0145] Further, Q cooling With the helium flow rate Q flow , helium heat transfer coefficient h and the temperature difference ΔT between liquid helium and the system he-sys Relevant, namely:

[0146] Q cooling =hQ flow ΔT he-sys Δt(5)

[0147] Furthermore, when it is predicted that the temperature has a tendency to exceed the threshold, in order to control the temperature before reaching the threshold, let ΔT = T thresh -T t At this time, it is necessary to solve the Δt that satisfies the energy balance equation. This Δt is the advance amount.

[0148] Furthermore, the above heat expression Relationship (5) and ΔT = T thresh -T t Substituting into the energy balance equation shown in the above relation (4), the following relation (6) is obtained:

[0149]

[0150] Finally, the corresponding advance calculation model can be constructed through the above process.

[0151] Furthermore, the above equations can be solved by numerical methods. Since the equations involve multiple variables, and some variables (such as the conversion rate of parahydrogen, the flow rate of cold helium, etc.) may change with time, it is more appropriate to use an iterative method to solve them. First, based on the current system state and measurement data, each variable is initialized and estimated. Then, these initial values are substituted into the equation to calculate a preliminary Δt value. Then, according to the calculated Δt value, the values of each variable are updated (for example, according to the control characteristics of the flow control, the change in the liquid helium flow rate within time t is calculated, and then the heat taken away by the liquid helium is updated), and then the updated variables are substituted into the equation to recalculate Δt. This process is iterated continuously until the difference between the Δt values calculated twice is less than the set accuracy requirement. Through this iterative solution method, a temperature control advance that meets actual needs can be obtained.

[0152] Step S305: Control the normal-para hydrogen conversion temperature according to the system advance start time and target inflow control parameters to obtain the temperature control result. Figure 1 Step S105 of the illustrated embodiment will not be described in detail here.

[0153] The para-hydrogen conversion temperature control method provided in this embodiment can iteratively optimize from a number of potential parameters through a genetic algorithm, and can obtain the optimal target inflow control parameter, thereby enabling the liquid helium or cold helium flow rate to accurately match the temperature regulation requirements, thereby improving the temperature control accuracy, while reducing flow consumption and energy waste. Furthermore, the model is constructed by combining the law of conservation of energy and the principle of heat transfer, so that the model has a solid physical theoretical basis, and can accurately reflect the relationship between heat transfer and temperature change in the para-hydrogen conversion process, thereby providing support for ensuring that the subsequent calculated advance start time is scientific and reasonable. Finally, the temperature change and system status data are combined and processed through the advance calculation model to calculate a reasonable advance start time.

[0154] In an optional embodiment, in combination with the above Figures 1 to 3 The control method of the normal-para hydrogen conversion temperature is shown in FIG. 1 , and the control process of the normal-para hydrogen conversion temperature intelligent control system is as follows: Figure 4 shown.

[0155] In one example, an intelligent temperature control method for ortho-parahydrogen conversion is provided. By combining a neural network algorithm and a genetic algorithm, the process of parahydrogen to ortho-hydrogen conversion is predicted, and flow is introduced into the converter in advance to cool it down. This can not only avoid the conversion of a large amount of parahydrogen to ortho-hydrogen, but also reduce flow consumption.

[0156] Specifically, a group of initial individuals are randomly generated, which constitute the initial population. Each individual represents a potential solution to the problem, and its encoding method can be binary encoding, real number encoding, etc. For example, in the temperature control problem, the control parameter of the inflow of liquid helium or cold helium gas can be encoded as an individual. Assume that the value range of the inflow of liquid helium or cold helium gas is 0-0-. The design of the fitness function is the key to the application of genetic algorithms. In temperature control, temperature stability and the degree of proximity to the target temperature are used as measurement criteria. Consider the impact of temperature changes on fitness:

[0157]

[0158] Where β is a weight coefficient used to adjust the degree of influence of temperature changes on fitness. In this way, not only the actual temperature is required to be close to the target temperature, but also the temperature changes are required to be smooth to avoid drastic temperature fluctuations.

[0159] In the actual temperature control process, a trained neural network model is first used to predict future temperature trends based on current temperature monitoring data and system status. This predicted temperature information is then used as input to a genetic algorithm, which optimizes and calculates the optimal inflow control parameters. Based on these calculated inflow control parameters, the liquid helium pump or cold helium gas flow control is controlled to adjust the inflow and precisely control the temperature of the normal-parahydrogen conversion process in the converter. Throughout the control process, temperature changes are continuously monitored, and new temperature data is fed back to the neural network model and genetic algorithm for real-time adjustments and optimization to ensure that the temperature remains stable within the target range.

[0160] The implemented algorithm was simulated and analyzed using test data. During the simulation, the temperature changes during the normal-parahydrogen conversion process were simulated. The liquid helium inflow control parameters calculated by the algorithm were applied to the temperature control system, and the temperature changes were observed. The performance advantages of the genetic algorithm in temperature control were evaluated by comparing it with traditional temperature control methods (such as PID control). Comparison metrics included steady-state temperature error, overshoot, and settling time.

[0161] In addition, since the temperature change during the conversion of ortho-parahydrogen is not completely random and is affected by factors such as the conversion rate and the heat exchange efficiency of the system, the potential trend of temperature change can be captured through in-depth analysis of these factors and the use of neural network models to learn and predict temperature data. After calculating the abnormal change, lower-temperature cold helium is introduced into the converter to cool it down and inhibit the conversion of parahydrogen to orthohydrogen. By introducing it in advance, it can not only avoid the conversion of a large amount of parahydrogen to orthohydrogen, but also reduce the use of liquid helium and save costs. The calculation scheme is as follows:

[0162] When the neural network model predicts that the temperature has a trend of continuous increase, it is necessary to further determine whether this trend will cause the temperature to exceed the set threshold. The setting of the threshold is determined based on the ideal temperature range of the normal-parahydrogen conversion process and the safe operation requirements of the equipment. Assuming that the ideal temperature range of normal-parahydrogen conversion is (T min ,T max ), considering the stability of the system and the tolerance of the equipment, the temperature rise threshold is set to T thresh-up , when the predicted temperature is likely to exceed T max -T thresh-up When the temperature drops, it is necessary to take measures to control the temperature in advance. Similarly, the temperature drop threshold is set to thresh-down .

[0163] The calculation of the advance amount is to calculate the time point or advance inflow amount when the temperature is predicted to exceed the threshold, based on the current temperature change rate, the thermal response characteristics of the system and the response time of the control actuator (such as the liquid helium pump). The response time from system startup to temperature drop is Δt response , in order to effectively control the temperature before it reaches the threshold, the advance Δt advance A preliminary estimate can be made using the formula:

[0164]

[0165] The above formula shows that the time required for the temperature to rise to the threshold is calculated based on the current temperature rise rate. Subtracting the system response time yields the required start-up lead time. In actual calculations, additional factors, such as the system's thermal inertia and the efficiency of the liquid helium heat exchanger, must be considered to revise and optimize this initial estimate.

[0166] In order to accurately calculate the lead time of temperature control, a multi-factor lead time calculation model is established. The model comprehensively considers factors such as temperature change trend, system thermal characteristics, liquid helium heat transfer characteristics, and response characteristics of the control actuator.

[0167] 1. Determine the model input parameters: The model input parameters include the current temperature value T t , temperature change rate These two parameters can be obtained by analyzing real-time temperature data using a neural network model; the system's heat capacity C, which reflects the system's ability to store heat and is related to factors such as the amount of hydrogen in the system and the material of the equipment; the helium heat transfer coefficient h, which indicates the ability of helium to exchange heat with the system and is related to factors such as the helium flow rate, flow velocity, and heat exchange area; and the response time Δt. pump , that is, the time required from the issuance of the control signal to the actual start of flow change; and the set temperature threshold T thresh .

[0168] 2. Establish model equation: According to the law of conservation of energy and the principle of heat transfer, establish the equation of the advance calculation model. Assume that in time t, the heat generated by the system due to the conversion of normal and para hydrogen is Q reaction , the heat taken away by the cold helium is Q cooling , the change in system temperature is ΔT, then:

[0169] Q reaction -Q cooling =CΔT

[0170] Among them, Qreaction It can be calculated based on the conversion rate and conversion heat of normal and para hydrogen. Assuming the conversion rate of normal and para hydrogen is The heat of conversion of normal and para hydrogen per unit amount of substance is ΔH, then

[0171] Further, Q cooling With the helium flow rate Q flow , helium heat transfer coefficient h and the temperature difference ΔT between liquid helium and the system he-sys Relevant, namely:

[0172] Q cooling =hQ flow ΔT he-sys Δt

[0173] Furthermore, when it is predicted that the temperature has a tendency to exceed the threshold, in order to control the temperature before reaching the threshold, let ΔT = T thresh -T t At this time, it is necessary to solve the Δt that satisfies the energy balance equation. This Δt is the advance amount.

[0174] Furthermore, the above heat expression Q cooling =hQ flow ΔT he-sys Δt and ΔT=T thresh -T t Substitute into the energy balance equation Q reaction -Q cooling =CΔT, we get:

[0175]

[0176] The specific way to predict temperature is to use neural network methods. The neural network can obtain the current temperature value and temperature change rate, and convert different temperatures, pressures and para-hydrogen into heat. First, the neural network obtains the current temperature and temperature change rate. Then, based on the temperature, it can combine the existing database information to obtain the para-hydrogen conversion heat and the para-hydrogen conversion rate. Then, based on the hydrogen flow rate inside the system, the total heat released during the conversion process can be obtained. Then, a cold flow that changes with time is provided. In this process, the total heat taken away by the cold flow can be obtained. Then, based on the heat values released and taken away and the heat capacity of the system, the system temperature change value can be obtained to achieve temperature prediction. The analysis process is as follows: Figure 5 As shown in the figure, a recurrent neural network (RNN) or long short-term memory network (LSTM) is used to process time series data and analyze cold flow changes. Since is a physically calculated value, a dual-supervision mechanism can be used to constrain the neural network prediction and physical calculation to ensure validity.

[0177]

[0178] L=||ΔT-ΔT phys ||

[0179] Mapping is achieved based on the current temperature combined with existing temperature change data. Physical models and large amounts of data can provide feedback on temperature changes. Therefore, the input layer of the neural network includes temperature, temperature change rate, hydrogen flow rate, cold flow flow rate, etc. The hidden layer extracts features, and the output layer includes conversion rate, conversion heat, heat release, and temperature change value. Time series processing can take the time step into account. Using the RNN structure, the time series of the cold flow is input into the RNN, and the current cold flow parameters are input at each time step. The database integration involved requires temperature as a query parameter, and the corresponding conversion rate and conversion heat can be obtained through the information in the database. Finally, the total heat release and the total heat absorbed by the cold flow are calculated, and the temperature change value can be obtained by dividing it by the heat capacity of the system.

[0180] 4. Model solution method: The above equations can be solved by numerical methods. Since the equations involve multiple variables, and some variables (such as the conversion rate of parahydrogen, cold helium flow, etc.) may change with time, it is more appropriate to use an iterative method to solve them. First, based on the current system state and measurement data, each variable is initialized and estimated. Then, these initial values are substituted into the equation to calculate a preliminary Δt value. Next, based on the calculated Δt value, the values of each variable are updated (for example, based on the control characteristics of the flow control, the change in the liquid helium flow rate within time t is calculated, and then the heat taken away by the liquid helium is updated), and then the updated variables are substituted into the equation to recalculate Δt. This process is iterated continuously until the difference between the Δt values calculated before and after is less than the set accuracy requirement.

[0181] This example also provides an implementation example to verify the feasibility of the above solution:

[0182] Set the target temperature for normal and para hydrogen conversion to T target =25K, temperature rise threshold T thresh-up , that is, when the predicted temperature exceeds T thresh-up =1K, control measures need to be initiated.

[0183] Over a period of time, the temperature data in the device, parameters related to the conversion of normal and para hydrogen (such as hydrogen flow rate, pressure, etc., used to calculate the conversion rate of normal and para hydrogen) and parameters of the helium supply system (such as liquid helium flow rate, cold helium flow rate, pressure, etc.) are collected in real time. The collected temperature data are analyzed using the neural network model constructed previously to obtain the temperature change trend and the current temperature change rate. At the same time, based on the design parameters and actual operation of the device, the system's heat capacity C = 1000 J / K and the heat transfer coefficient of liquid helium h = 500 W / (m 2K), corresponding to the response time Δt of flow control pump =10s.

[0184] For example, at a certain moment, the neural network model predicts that the temperature has a trend of continuing to rise. The current temperature T t =24.5K, temperature change rate Calculate the model Δt based on the advance amount advance = 30s, substitute the relevant parameters for calculation. After iterative solution, the lead time is obtained. That is, it is necessary to start the liquid helium pump after 30 seconds to increase the inflow of liquid helium to control the temperature rise. According to the calculation results, the control system starts the liquid helium pump after 30 seconds, and the liquid helium pump control flow is Q flow =5L / min of liquid helium.

[0185] Subsequent temperature monitoring revealed that the temperature began to decrease after reaching 25.8K, ultimately stabilizing at around 25.2K, successfully preventing the temperature from exceeding the threshold of 26K. This compares to conventional control methods that don't employ lead calculations. These methods activate other cooling agents only after the temperature reaches 26.5K, resulting in a significant temperature overshoot, reaching a maximum of 27.2K. The temperature also takes a long time to stabilize, taking approximately 15 minutes to stabilize at around 25.5K. Even using liquid helium for cooling, this method requires a flow rate of 15L / min.

[0186] The above examples show that the temperature control method based on lead calculation can more promptly and effectively control the temperature during the normal-parahydrogen conversion process, reduce temperature fluctuations and overshoot, and improve system stability and control accuracy. At the same time, the effectiveness and rationality of the lead calculation model are verified, providing a reliable strategy for temperature control in the hydrogen liquefaction process. However, in practical applications, it is necessary to consider the impact of various uncertainties on the lead calculation, such as sensor measurement errors and changes in system parameters, and further optimize and improve the lead calculation method to adapt to more complex operating conditions and higher control requirements.

[0187] Finally, multiple sets of temperature sensors can be placed within the converter. Each sensor in the same area measures the temperature of that area with a delay of approximately 0.1 to 1 second. The system then records and compares the temperature. Using a comparative test method, the system determines whether the temperature difference recorded by multiple sensors in the same area can be within 0.5K under short-term delay recording. If the difference in temperature measured by a sensor exceeds 0.5K, a measurement error may have occurred, and the data from other sensors can be used for calculations. If the error exceeds 0.5K across multiple measurements, a fault has occurred and the device should be replaced.

[0188] The intelligent control method for normal-parahydrogen conversion temperature provided in this example has the following effects:

[0189] 1. During the temperature control process, a trained neural network model is used to predict future temperature trends based on current temperature monitoring data and system status. This predicted temperature information is then used as input for a genetic algorithm, which optimizes and calculates the optimal liquid helium inflow control parameters. Cooler liquid helium or cold helium gas is used to absorb the generated heat, promoting conversion.

[0190] 2. Based on the calculated helium inflow control parameters, the cold helium flow control or the liquid helium pump is controlled to adjust the liquid helium inflow or control the cold helium flow control, thereby achieving precise control of the temperature during the normal-parahydrogen conversion process. During the control process, temperature changes are continuously monitored and the new temperature data is fed back to the neural network model and genetic algorithm for real-time adjustment and optimization to ensure that the temperature is always stable within the target range.

[0191] 3. Simulation results show that, under identical initial conditions and disturbances, the traditional PID control achieves a steady-state temperature error of ±2K, an overshoot of 15%, and a settling time of 10 minutes. In contrast, the genetic algorithm-based temperature control achieves a steady-state temperature error of ±0.5K, an overshoot of 5%, and a settling time of 5 minutes. These results clearly demonstrate that the genetic algorithm employed in this example can more effectively reduce the steady-state temperature error and overshoot, shorten the settling time, and achieve more precise temperature control. Simulation analysis under various operating conditions also verifies the algorithm's robustness and adaptability, ensuring its stable and reliable operation in a variety of complex situations.

[0192] 4. The algorithm can control the inflow of liquid helium or cold helium in advance, reduce the consumption of liquid helium, and at the same time judge the detection accuracy of the sensor, achieving multiple goals at one stroke.

[0193] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for controlling the temperature of normal-parahydrogen conversion, characterized in that: For an intelligent control system for normal and para-hydrogen conversion temperature; the method comprises: Obtain real-time temperature data sets and system status data sets for normal and para hydrogen conversion during hydrogen liquefaction; Processing the real-time temperature data set and the real-time system status data set through a target neural network model to obtain a real-time temperature change data set; Determining whether temperature control is required based on the real-time temperature change data set; When temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the system real-time status data set through genetic algorithm and advance calculation model processing; The normal-para hydrogen conversion temperature is controlled according to the system advance start-up time and the target inflow control parameter to obtain a temperature control result.

2. The method according to claim 1, characterized in that The method further comprises: Obtain historical temperature data sets and system historical status data sets; The target neural network model is constructed based on the historical temperature data set and the system historical status data set.

3. The method according to claim 2, characterized in that Constructing the target neural network model based on the historical temperature data set and the system historical state data set includes: According to the historical temperature data set and the system historical state data set, a historical temperature change data set is obtained through a preset calculation method; The target neural network model is constructed with the historical temperature data set and the system historical state data set as input and the historical temperature change data set as output.

4. The method according to claim 2, characterized in that The method further comprises: The target neural network model is verified using a dual supervision mechanism.

5. The method according to claim 4, characterized in that The target neural network model is verified using a dual-supervision mechanism, including: Determining a physical temperature change value using the dual supervision mechanism; Obtaining a predicted temperature change value output by the target neural network model; Constructing a loss function based on the physical temperature change value and the predicted temperature change value; The accuracy of the target neural network model is verified according to the loss function.

6. The method according to claim 1, characterized in that When temperature control is required, the target inflow control parameters and the system advance start time are obtained based on the real-time temperature change data set and the real-time system status data set through genetic algorithm and advance calculation model processing, including: When temperature control is required, the real-time temperature change data set is processed by a genetic algorithm to obtain the target inflow control parameter; The temperature change information set and the system real-time status data set are input into an advance calculation model to obtain the advance start time of the system.

7. The method according to claim 6, characterized in that When temperature control is required, the real-time temperature change data set is processed by a genetic algorithm to obtain target inflow control parameters, including: Obtaining a preset fitness function, wherein the preset fitness function is used to characterize temperature stability and the degree of proximity between the actual temperature and the target temperature; When temperature control is required, the real-time temperature change data set is input into the preset fitness function, and the inflow control parameters are optimized until the target inflow control parameters are obtained.

8. The method according to claim 7, characterized in that When temperature control is required, the real-time temperature change data set is input into the preset fitness function, and the inflow control parameters are optimized until the target inflow control parameters are obtained, including: When temperature control is required, the inflow control parameters are used to encode individuals and generate the initial population; Inputting the real-time temperature change data set into a preset fitness function, and evaluating each individual in the initial population to obtain multiple individual fitness values; performing genetic operations on the initial population according to the plurality of individual fitness values and updating the initial population; According to the updated initial population, individual evaluation and genetic operation are repeatedly performed until the termination condition is met, and the iteration is stopped and the target inflow control parameter is obtained.

9. The method according to claim 6, characterized in that The method further comprises: Obtain historical temperature change information set and system historical status data set; Based on the historical temperature change information set and the system historical state data set, the lead time calculation model is constructed through processing according to the law of conservation of energy and the principle of heat transfer.

10. The method according to claim 9, characterized in that Based on the historical temperature change information set and the system historical state data set, and after processing according to the law of conservation of energy and the principle of heat transfer, the lead time calculation model is constructed, including: Based on the historical temperature change information set and the system historical state data set, and after processing according to the law of conservation of energy, an energy balance model is constructed; Based on the historical temperature change information set and the system historical state data set, a heat transfer model is constructed through processing according to the heat transfer principle; The advance amount calculation model is determined according to the energy balance model and the heat transfer model.

Images