Regulation and control method of hydrogen liquefaction process based on AI

Abstract

The invention relates to the technical field of hydrogen liquefaction, and discloses an AI-based hydrogen liquefaction process regulation and control method, which comprises the following steps: generating a preset number of parameter sets according to equipment parameter intervals of multiple sets of hydrogen liquefaction systems; by taking the energy loss as an optimization target, performing optimization iteration on the preset number of parameter sets by utilizing a preset fitness function until a target parameter set is obtained; the target parameter set is used for executing the hydrogen liquefaction process, and operation parameters of the hydrogen liquefaction process in the execution process are obtained; according to the method, the energy loss is used as an optimization target, the parameter set is optimized, the target parameter set is corrected when the abnormality occurs, and the hydrogen liquefaction process is regulated and controlled. And the energy waste in the hydrogen liquefaction process is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrogen liquefaction, and particularly to a control method for a hydrogen liquefaction process based on AI. Background Art

[0002] The hydrogen liquefaction process is a process of converting gaseous hydrogen into liquid hydrogen, which usually needs to go through multiple stages such as precooling and refrigeration to convert hydrogen from normal temperature and pressure state into low-temperature liquid state. Currently, common hydrogen liquefaction processes include the hydrogen liquefaction process based on a liquefied natural gas (LNG) receiving terminal, the mixed refrigerant hydrogen liquefaction process, etc.

[0003] The hydrogen liquefaction process includes multiple links. The traditional hydrogen liquefaction process generally uses LNG cold energy precooling, mixed refrigerant refrigeration and other methods to convert hydrogen into liquid hydrogen. In the process of converting hydrogen into liquid hydrogen, hydrogen needs to be cooled to an extremely low temperature, resulting in a large amount of energy consumption in the refrigeration process. Therefore, the traditional hydrogen liquefaction process will cause a large amount of energy waste. Summary of the Invention

[0004] In view of this, the present invention provides a control method for a hydrogen liquefaction process based on AI to solve the problem of a large amount of energy waste caused by the traditional hydrogen liquefaction process.

[0005] In a first aspect, the present invention provides a control method for a hydrogen liquefaction process based on AI, including: generating a preset number of parameter sets according to the equipment parameter ranges of multiple hydrogen liquefaction systems; the parameter sets are parameter sets obtained by randomly selecting multiple parameters in the equipment parameter ranges; taking the energy loss of each hydrogen liquefaction system as an optimization target, and using a preset fitness function to perform optimization iteration on the preset number of parameter sets until a target parameter set is obtained; the preset fitness function is used to characterize the optimization degree of the energy loss of the hydrogen liquefaction system; the fitness value corresponding to the target parameter set is greater than a first preset value; executing the hydrogen liquefaction process using the target parameter set, and obtaining the operation parameters during the execution of the hydrogen liquefaction process; determining whether the operation parameters are abnormal, and if the operation parameters are abnormal, correcting the target parameter set to control the hydrogen liquefaction process.

[0006] According to the equipment parameter ranges of multiple hydrogen liquefaction systems, the present invention generates a preset number of parameter sets randomly selected from the equipment parameter ranges, which can ensure that the preset number of parameter sets fully cover the equipment parameter ranges of each key equipment belonging to each hydrogen liquefaction system, providing a large number of samples for subsequent optimization iterations. The present invention takes energy loss as the optimization goal, and uses a preset fitness function to optimize and iterate the preset number of parameter sets until a target parameter set is obtained. The present invention quantifies the optimization degree of energy loss, uses a genetic algorithm to drive AI to globally optimize the parameter set, and continuously screens out target parameter sets with lower energy consumption, realizing the minimization of the energy loss corresponding to the target parameter set. During the execution of the hydrogen liquefaction process, the present invention acquires operating parameters and determines whether there are abnormalities in the operating parameters. If there are abnormalities in the operating parameters, the target parameter set is corrected to achieve dynamic adjustment of the target parameter set, prevent failures and abnormalities in the hydrogen liquefaction process, and ensure the stable execution of the hydrogen liquefaction process. Compared with related technologies, the present invention reduces the energy loss of the hydrogen liquefaction process and realizes precise control of the hydrogen liquefaction process.

[0007] In an alternative embodiment, generating a preset number of parameter sets according to the equipment parameter ranges of multiple hydrogen liquefaction systems includes: obtaining the equipment parameter ranges corresponding to multiple key equipments belonging to each hydrogen liquefaction system; randomly selecting the multiple parameters in each equipment parameter range for a preset number of times, and adding the parameters corresponding to each hydrogen liquefaction system obtained by each random selection to a set to obtain a preset number of parameter sets.

[0008] In an alternative embodiment, taking the energy loss of each hydrogen liquefaction system as the optimization goal and using a preset fitness function to optimize and iterate the preset number of parameter sets until a target parameter set is obtained includes: encoding the preset number of parameter sets to obtain a preset number of first individuals; constructing a preset fitness function according to the energy loss of each hydrogen liquefaction system, and using the preset fitness function and the total energy loss data corresponding to each first individual to generate the fitness value corresponding to each first individual; updating the preset number of first individuals according to the fitness values to obtain a plurality of second individuals; using the second individuals to update the first individuals, and returning to the step of generating the fitness value corresponding to each second individual using the preset fitness function until the iteration termination condition is reached to obtain a plurality of target individuals; among the plurality of target individuals, selecting the optimal target individual with a fitness value greater than the first preset value; decoding the optimal target individual to obtain the target parameter set.

[0009] The present invention generates individuals by encoding a parameter set, updates the individuals in combination with evolutionary mechanisms such as genetic algorithms, constructs a fitness function directly related to energy loss, converts the degree of energy loss optimization into a computable numerical index, ensures that the iteration always progresses around the goal of reducing energy loss, and enables the finally obtained target parameter set to reach a better level in energy consumption optimization.

[0010] In an alternative embodiment, a preset fitness function is constructed according to the energy loss of each hydrogen liquefaction system, including: constructing a preset fitness function according to the reciprocal of the energy loss of each hydrogen liquefaction system.

[0011] In an alternative embodiment, the preset number of first individuals are updated according to the fitness value to obtain a plurality of second individuals, including: performing probability analysis on each first individual according to the fitness value to obtain the target probability corresponding to each first individual; selecting a plurality of first target individuals from the plurality of first individuals according to the target probability; the fitness value of the first target individual is greater than the first preset value; the second preset value is less than the first preset value; performing crossover operations on each first target individual to obtain a plurality of second target individuals; and performing mutation operations on the plurality of second target individuals to obtain a plurality of second individuals.

[0012] In an alternative embodiment, the hydrogen liquefaction process is executed using the target parameter set, including: obtaining a plurality of target parameters in the target parameter set, configuring the plurality of target parameters on corresponding key devices, and controlling the operation of the key devices to execute the hydrogen liquefaction process.

[0013] In an alternative embodiment, it is determined whether the operating parameters are abnormal, including: inputting the historical operating parameters into a trained operating parameter prediction model to obtain the predicted value of the operating parameters at the current moment; the input of the operating parameter prediction model is the historical operating parameters, and the output of the operating parameter prediction model is the predicted value of the operating parameters; comparing the operating parameters with the predicted value of the operating parameters; if the difference between the operating parameters and the predicted value of the operating parameters is greater than a preset threshold, it is determined that the operating parameters are abnormal; if the difference between the operating parameters and the predicted value of the operating parameters is less than or equal to the preset threshold, it is determined that the operating parameters are normal.

[0014] In an alternative embodiment, the target parameter set is corrected, including: updating the parameter set using the target parameter set, and returning the steps of optimizing and iterating the target parameter set using the preset fitness function with the energy loss of each hydrogen liquefaction system as the optimization goal until the target corrected parameter set is obtained, so as to correct the target parameter set.

[0015] When the operating parameters are abnormal, the present invention automatically triggers a new round of optimization, continuously adjusts the parameters, ensures that the hydrogen liquefaction process always approaches the optimal energy consumption state, adapts to the dynamic changes during the operation of key equipment, maintains the stability of the hydrogen liquefaction process, and improves the energy consumption optimization effect.

[0016] In an alternative embodiment, correcting the target parameter set includes: obtaining the outlier and removing the target parameter in the target parameter set corresponding to the outlier to correct the target parameter set.

[0017] In an alternative embodiment, if the operating parameters are normal, a prompt message is generated; the prompt message is used to prompt that the operating parameters are normal. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related art, the following will briefly introduce the drawings required for the description of the specific embodiments or the related art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a flowchart of a method for regulating a hydrogen liquefaction process based on AI according to an embodiment of the present invention.

[0020] Figure 2 It is a flowchart of another method for regulating a hydrogen liquefaction process based on AI according to an embodiment of the present invention.

[0021] Figure 3 It is a flowchart of yet another method for regulating a hydrogen liquefaction process based on AI according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0023] The hydrogen liquefaction process is a process of converting gaseous hydrogen into liquid hydrogen. The hydrogen liquefaction process usually needs to go through multiple stages such as precooling and refrigeration to convert hydrogen from normal temperature and pressure state into low-temperature liquid state. At present, common hydrogen liquefaction processes include hydrogen liquefaction processes based on liquefied natural gas receiving stations, mixed refrigerant hydrogen liquefaction processes, etc. Common hydrogen liquefaction processes generally use liquefied natural gas cold energy for precooling, mixed refrigerant refrigeration, etc. to convert hydrogen into liquid hydrogen, and multiple factors such as energy consumption, efficiency, equipment manufacturing, and safe operation need to be considered.

[0024] Existing genetic algorithms have been applied in the design of hydrogen liquefaction processes. Usually, the hydrogen liquefaction rate, para-hydrogen content, and conversion efficiency are used as optimization objectives to design the hydrogen liquefaction process. However, in the actual operation process, the hydrogen liquefaction process involves many decision variables. For example, hydrogen liquefaction needs to cool hydrogen to extremely low temperatures, and the efficiency of traditional refrigeration cycles needs to be improved. The hydrogen liquefaction process contains multiple links, and the energy matching and coordinated operation between each link are not optimized enough. A large amount of energy is consumed in the refrigeration process, resulting in a large amount of energy waste.

[0025] An embodiment of the present invention provides a method for regulating a hydrogen liquefaction process based on AI. By taking energy loss as the optimization objective, the parameter set is optimized, and the target parameter set is corrected when an anomaly occurs, so as to achieve the effect of reducing energy waste in the hydrogen liquefaction process.

[0026] According to an embodiment of the present invention, an embodiment of a method for regulating a hydrogen liquefaction process based on AI 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 the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0027] In this embodiment, a method for regulating a hydrogen liquefaction process based on AI is provided, which can be used in computer devices. Figure 1 is a flowchart of a method for regulating a hydrogen liquefaction process based on AI according to an embodiment of the present invention, as Figure 1 shown, the process includes the following steps:

[0028] Step S101, generate a preset number of parameter sets according to the equipment parameter intervals of multiple hydrogen liquefaction systems; the parameter set is a parameter set obtained by randomly selecting multiple parameters in the equipment parameter interval.

[0029] Among them, the hydrogen liquefaction system is a system that executes the hydrogen liquefaction process. Each hydrogen liquefaction system includes multiple key devices. The key devices are the devices used to execute the hydrogen liquefaction process. Exemplarily, the key devices include compressors, precoolers, main coolers, throttle valves, etc. The equipment parameter range is the value range of the equipment parameters of the key devices. One key device corresponds to at least one equipment parameter range. Exemplarily, for the key device compressor, the corresponding equipment parameters include the intake pressure and the discharge pressure. Among them, the equipment parameter range corresponding to the intake pressure is 0.8 MPa to 1.2 MPa.

[0030] In some alternative embodiments, the preset quantity can be set according to the actual situation. Exemplarily, the preset quantity can be 100.

[0031] In some alternative embodiments, according to the equipment parameter ranges of multiple hydrogen liquefaction systems, a preset number of parameter sets are generated, including: obtaining the equipment parameter ranges corresponding to multiple key devices belonging to each hydrogen liquefaction system; randomly selecting the multiple parameters in each equipment parameter range for a preset number of times, and adding the parameters corresponding to each hydrogen liquefaction system obtained by each random selection to a set to obtain a preset number of parameter sets.

[0032] Exemplarily, among the multiple parameters in each equipment parameter range corresponding to each key device, randomly select one parameter and add it to a set, and perform 100 random selections to obtain 100 parameter sets. Each parameter set includes the parameters corresponding to all key devices used in the hydrogen liquefaction process.

[0033] Step S102: Using the energy loss of each hydrogen liquefaction system as the optimization objective, optimize and iterate the preset number of parameter sets by using a preset fitness function until a target parameter set is obtained; the preset fitness function is used to characterize the optimization degree of the energy loss of the hydrogen liquefaction system; the fitness value corresponding to the target parameter set is greater than a first preset value.

[0034] Among them, the energy loss of each hydrogen liquefaction system is the total energy loss caused by each hydrogen liquefaction system executing the hydrogen liquefaction process. According to the sum of the equipment energy losses of each key device belonging to each hydrogen liquefaction system, the energy loss of each hydrogen liquefaction system is determined. The first preset value can be set according to the actual situation. The setting of the first preset value is as large as possible so that the fitness value corresponding to the target parameter set is as large as possible. Since the preset fitness function can be defined as the reciprocal of the energy loss, when the fitness value is large, the energy loss is as small as possible.

[0035] In some alternative embodiments, the preset fitness function can be customized according to requirements. Exemplarily, the preset fitness function can be defined as the reciprocal of the energy loss. The expression of the preset fitness function is:

[0036]

[0037] Among them, Fitness is a preset fitness function, and E total is the energy loss of each hydrogen liquefaction system.

[0038] In some alternative embodiments, with the energy loss as the optimization objective, a genetic algorithm is used to drive AI (Artificial Intelligence), and a preset fitness function is used to perform optimization iteration on a preset number of parameter sets, and a target parameter set is output.

[0039] Specifically, encode a preset number of parameter sets to obtain a preset number of first individuals; construct a preset fitness function according to the energy loss of each hydrogen liquefaction system, and use the preset fitness function and the total energy loss data corresponding to each first individual to generate the fitness value corresponding to each first individual; update the preset number of first individuals according to the fitness value to obtain a plurality of second individuals; use the second individuals to update the first individuals, and return to the step of generating the fitness value corresponding to each second individual using the preset fitness function until the iteration termination condition is reached to obtain a plurality of target individuals; among the plurality of target individuals, select the optimal target individual whose fitness value is greater than the first preset value; decode the optimal target individual to obtain the target parameter set.

[0040] Step S103, execute the hydrogen liquefaction process using the target parameter set, and obtain the operating parameters during the execution of the hydrogen liquefaction process.

[0041] Among them, the hydrogen liquefaction process is a process of converting gaseous hydrogen into liquid hydrogen with the help of key equipment belonging to each hydrogen liquefaction system. Exemplarily, the hydrogen liquefaction process includes: pressurizing hydrogen with a compressor; using a precooler to preliminarily cool and lower the temperature of the compressed hydrogen with liquid nitrogen or other cooling media (such as mixed working fluids); using a main cooler to further lower the temperature of hydrogen by heat exchange with a low-temperature refrigeration medium (such as hydrogen or helium that has been cooled by expansion in an expander); using a throttle valve to cause isenthalpic expansion of high-pressure hydrogen, etc.

[0042] In some alternative embodiments, executing the hydrogen liquefaction process using the target parameter set includes: obtaining a plurality of target parameters in the target parameter set, configuring the plurality of target parameters on the corresponding key equipment, and controlling the operation of the key equipment to execute the hydrogen liquefaction process.

[0043] In some alternative embodiments, the operating parameters may include liquid level data and pressure data. Obtaining the operating parameters during the execution of the hydrogen liquefaction process includes: obtaining in real time the liquid level data and pressure data collected by the liquid level sensors and pressure sensors on multiple liquid hydrogen storage tanks.

[0044] Step S104, determine whether the operating parameters are abnormal. If the operating parameters are abnormal, correct the target parameter set to regulate the hydrogen liquefaction process.

[0045] In some alternative embodiments, if the operating parameters are normal, generate a prompt message; the prompt message is used to prompt that the operating parameters are normal.

[0046] In some alternative embodiments, determining whether the operating parameters are abnormal includes: inputting the historical operating parameters into the trained operating parameter prediction model to obtain the predicted value of the operating parameters at the current moment; the input of the operating parameter prediction model is the historical operating parameters, and the output of the operating parameter prediction model is the predicted value of the operating parameters; comparing the operating parameters with the predicted value of the operating parameters; if the difference between the operating parameters and the predicted value of the operating parameters is greater than the preset threshold, it is determined that the operating parameters are abnormal; if the difference between the operating parameters and the predicted value of the operating parameters is less than or equal to the preset threshold, it is determined that the operating parameters are normal.

[0047] Wherein, the preset threshold is a critical value of the difference change set in advance. When the difference between the operating parameters and the predicted value of the operating parameters is greater than the preset threshold, it indicates that the operating parameters are abnormal.

[0048] In some alternative embodiments, determining whether the operating parameters are abnormal includes: determining whether the change trend of the operating parameters is abnormal. If the change trend of the operating parameters is abnormal, it is determined that the operating parameters are abnormal. If the change trend of the operating parameters is not abnormal, it is determined that the operating parameters are normal.

[0049] Exemplarily, if the liquid level data in the operating parameters rises rapidly and the pressure data in the operating parameters fluctuates abnormally, it is determined that the change trend of the operating parameters is abnormal.

[0050] In some alternative embodiments, a method for regulating a hydrogen liquefaction process based on AI further includes the training process of the operating parameter prediction model. The training process of the operating parameter prediction model includes: obtaining historical operating parameter time series data, processing the outliers in the historical operating parameter time series data to obtain a first processing result, normalizing the processing result to obtain a second processing result, dividing the second processing result by a preset time step to form an input data set, and inputting the input data set into the operating parameter prediction model to train the operating parameter prediction model.

[0051] Exemplarily, the historical operating parameter time series data includes liquid level data and pressure data at different times; processing the outliers in the historical operating parameter time series data includes: processing the outliers using the moving average method. If the difference between the liquid level data L at a certain time t and the liquid level data at the adjacent times before and after exceeds a certain threshold, update L t to the average value of the liquid level data at the adjacent times before and after; for missing values, use linear interpolation to fill them.

[0052] Performing normalization on the processing results includes: for the liquid level data L t , the normalization formula is:

[0053]

[0054] where, L norm is the processed result of the liquid level in the second processing result, L t is the liquid level data at a certain time, L min is the minimum value of the liquid level data in the historical operating parameter time series data, L max is the maximum value of the liquid level data in the historical operating parameter time series data.

[0055] For the pressure data P, the normalization formula is:

[0056]

[0057] where, P norm is the processed result of the pressure in the second processing result, P is the pressure data at a certain time, P min is the minimum value of the pressure data in the historical operating parameter time series data, P max is the maximum value of the pressure data in the historical operating parameter time series data.

[0058] In some alternative embodiments, the preset time step can be set according to the actual situation. Exemplarily, the preset time step can be 5 minutes. The liquid level data and pressure data of the previous n preset time steps are used as input features, and the liquid level data and pressure data of the next preset time step are used as labels.

[0059] In some alternative embodiments, the operating parameter prediction model can be an LSTM (Long Short-Term Memory) model. The LSTM model includes an input layer, multiple LSTM hidden layers, and an output layer. The input layer receives the normalized second processing result. The LSTM hidden layers process the long-term and short-term dependencies in the historical operating parameter time series data through a gating mechanism, and the output layer outputs the predicted value of the operating parameter at the next predicted time step.

[0060] Among them, the operating parameter prediction model uses the mean squared error as the loss function, and the loss function is:

[0061]

[0062] where MSE is the mean squared error, N is the total amount of historical operating parameter time series data, y is the actual value of the operating parameter, and y i is the predicted value of the operating parameter.

[0063] In some alternative embodiments, the gradient descent algorithm is used to train the operating parameter prediction model, and the parameters of the operating parameter prediction model are continuously adjusted to minimize the loss function.

[0064] In some alternative embodiments, the target parameter set is corrected, including: updating the parameter set using the target parameter set, returning the step of optimizing and iterating the target parameter set using a preset fitness function with the energy loss of each hydrogen liquefaction system as the optimization target until the target corrected parameter set is obtained, so as to correct the target parameter set.

[0065] In some alternative embodiments, the target parameter set is corrected, including: obtaining outliers and removing the target parameters in the target parameter set corresponding to the outliers, so as to correct the target parameter set.

[0066] A control method for a hydrogen liquefaction process based on AI provided in this embodiment generates a preset number of parameter sets randomly selected from the equipment parameter intervals according to the equipment parameter intervals of multiple hydrogen liquefaction systems, which can ensure that the preset number of parameter sets fully cover the equipment parameter ranges of each key equipment belonging to each hydrogen liquefaction process, providing a large number of samples for subsequent optimization and iteration. The embodiment of the present invention takes the energy loss as the optimization target, optimizes and iterates the preset number of parameter sets using a preset fitness function until the target parameter set is obtained. The present invention quantifies the optimization degree of the energy loss, uses the genetic algorithm to drive the AI to globally optimize the parameter set, and continuously screens out the target parameter sets with lower energy consumption, realizing the minimization of the energy loss corresponding to the target parameter set. In the execution process of the hydrogen liquefaction process in the embodiment of the present invention, the operating parameters are obtained, and it is judged whether the operating parameters are abnormal. If the operating parameters are abnormal, the target parameter set is corrected to realize the dynamic adjustment of the target parameter set, prevent the hydrogen liquefaction process from malfunctioning and abnormal conditions, and ensure the stable execution of the hydrogen liquefaction process. Compared with the related technologies, the embodiment of the present invention reduces the energy loss of the hydrogen liquefaction process and realizes the precise control of the hydrogen liquefaction process.

[0067] In this embodiment, a control method for a hydrogen liquefaction process based on AI is provided, which can be used in computer equipment, Figure 2It is a flowchart of another regulation method for a hydrogen liquefaction process based on AI according to an embodiment of the present invention. As Figure 2 shown, this process includes the following steps:

[0068] Step S201, generate a preset number of parameter sets according to the equipment parameter ranges of multiple hydrogen liquefaction systems; the parameter sets are parameter sets obtained by randomly selecting multiple parameters in the equipment parameter ranges. For details, please refer to Figure 1 step S101 of the embodiment shown here, which will not be elaborated herein.

[0069] Step S202, taking the energy loss of each hydrogen liquefaction system as the optimization objective, use a preset fitness function to optimize and iterate the preset number of parameter sets until a target parameter set is obtained; the preset fitness function is used to characterize the optimization degree of the energy loss of the hydrogen liquefaction system; the fitness value corresponding to the target parameter set is greater than the first preset value.

[0070] Specifically, the above step S202 includes:

[0071] Step S2021, encode the preset number of parameter sets to obtain the first individuals of the preset number.

[0072] Among them, encoding the preset number of parameter sets to obtain the first individuals of the preset number includes: encoding multiple parameters in each parameter set respectively to obtain the encoding string of each parameter set, and taking each encoding string as an individual to obtain the first individuals of the preset number.

[0073] Exemplarily, for the inlet pressure P in of the compressor, its equipment parameter range is 0.8 MPa to 1.2 MPa, divided into 2 8 intervals, then the length of each interval is Represent each interval with an 8-bit binary number to complete the encoding of the inlet pressure. For other key equipment and other parameters, they are encoded in a similar way to obtain the encoding string of each parameter set.

[0074] Step S2022, construct a preset fitness function according to the energy loss of each hydrogen liquefaction system, and use the preset fitness function and the total energy loss data corresponding to each first individual to generate the fitness value corresponding to each first individual.

[0075] Among them, construct a preset fitness function according to the reciprocal of the energy loss of each hydrogen liquefaction system. The total energy loss data can be obtained by the sum of the equipment energy losses of multiple key equipment of each hydrogen liquefaction system, and the equipment energy loss is calculated or analyzed through the parameters of each key equipment.

[0076] Exemplarily, the equipment energy loss of the compressor is:

[0077]

[0078] where P comp is the equipment energy loss of the compressor, P out is the exhaust pressure of the compressor, P in is the intake pressure of the compressor, m is the mass flow rate of the compressor, T is the inlet temperature of the compressor, R is the gas parameter, and γ is the adiabatic index of the gas.

[0079] In some alternative embodiments, the corresponding relationships between the energy losses of the refrigerant circulation pump and other components of the precooler and factors such as temperature difference and flow rate are obtained. The temperature difference parameter and flow rate parameter and other equipment parameters of the precooler are substituted into the corresponding relationships to obtain the equipment energy loss P pre .

[0080] In some alternative embodiments, the corresponding relationship between the energy loss of the main condenser and parameter factors is obtained. The equipment parameters of the main condenser are substituted into the corresponding relationship to obtain the equipment energy loss P main .

[0081] In some alternative embodiments, the corresponding relationship between the energy loss of the throttle valve and parameter factors such as pressure difference is obtained. The equipment parameters of the throttle valve are substituted into the corresponding relationship to obtain the equipment energy loss P throttle .

[0082] In some alternative embodiments, a thermodynamic model can also be constructed based on the parameter set. The thermodynamic model is used to simulate the hydrogen liquefaction process to collect the energy losses of each key equipment.

[0083] In some alternative embodiments, the overall energy loss data can be obtained by summing the equipment energy losses of each key equipment in each hydrogen liquefaction system. Then the overall energy loss data is:

[0084] E total = P comp + P pre + P main + P throttle

[0085] where E total is the overall energy loss data, P comp is the equipment energy loss of the compressor, P pre is the equipment energy loss of the precooler, P main is the equipment energy loss of the main condenser, P throttle is the equipment energy loss of the throttle valve.

[0086] In some alternative embodiments, using a preset fitness function and the overall energy loss data corresponding to each first individual to generate a fitness value corresponding to each first individual, including: substituting the overall energy loss data corresponding to each first individual into the preset fitness function to obtain the fitness value corresponding to each first individual.

[0087] Step S2023, update a preset number of first individuals according to the fitness value to obtain a plurality of second individuals.

[0088] In some alternative embodiments, updating a preset number of first individuals according to the fitness value to obtain a plurality of second individuals includes: performing probability analysis on each first individual according to the fitness value to obtain a target probability corresponding to each first individual; selecting a plurality of first target individuals from the plurality of first individuals according to the target probability; the first target individuals are the first individuals whose fitness values are greater than a second preset value; the second preset value is less than the first preset value; performing a crossover operation on each first target individual to obtain a plurality of second target individuals; performing a mutation operation on the plurality of second target individuals to obtain a plurality of second individuals.

[0089] Among them, the calculation formula for performing probability analysis on each first individual according to the fitness value to obtain the target probability corresponding to each first individual is:

[0090]

[0091] where P i is the target probability, Fitness j is the fitness function value of the j-th first individual, and M is the total number of first individuals.

[0092] In some alternative embodiments, selecting a plurality of first target individuals from the plurality of first individuals according to the target probability includes: selecting, according to the target probability, the first individuals whose fitness values are greater than the second preset value by simulating a roulette wheel, so as to obtain a plurality of first target individuals.

[0093] In some alternative embodiments, performing a crossover operation on each first target individual to obtain a plurality of second target individuals includes: performing a crossover operation on the first target individuals with binary coding, using a single-point crossover or a multi-point crossover method. For example, for two first target individuals A and B, randomly select a crossover point (assuming in the middle position of the coding string), and exchange the coding parts of the two first target individuals after the crossover point to obtain two new second target individuals A' and B'. The crossover operation can combine the excellent genes of different first target individuals and may produce better second target individuals.

[0094] In some alternative embodiments, mutation operations are performed on multiple second target individuals to obtain multiple second individuals, including: for the binary encoding of each second target individual, randomly changing the values of certain bits with a certain mutation probability to obtain multiple second individuals. Exemplarily, changing a certain bit in the encoding string of each second target individual from 0 to 1 or from 1 to 0. The mutation operation can increase the diversity of the population and prevent the genetic algorithm from converging to a local optimal solution prematurely.

[0095] Step S2024: Update the first individual using the second individuals, and return to the step of generating the fitness value corresponding to each second individual using a preset fitness function until the iteration termination condition is reached, obtaining multiple target individuals.

[0096] Among them, the iteration termination condition can be set according to actual requirements. Exemplarily, the iteration termination condition can be whether the number of iterations reaches 100 times.

[0097] Step S2025: Among the multiple target individuals, select the optimal target individual whose fitness value is greater than the first preset value.

[0098] Step S2026: Decode the optimal target individual to obtain a set of target parameters.

[0099] In the embodiments of the present invention, the total energy loss data caused by executing the hydrogen liquefaction process using the set of target parameters is less than the total energy loss data caused by executing the hydrogen liquefaction process using the original parameter set that is not processed by the genetic algorithm.

[0100] Exemplarily, assume that in the original parameter set, the inlet pressure of the compressor is 1.0 megapascal (MPa), the discharge pressure is 5.0 MPa, the mass flow rate is 10 kilograms per hour, assume that the inlet temperature of the compressor is 300 Kelvin (K), and γ = 1.4. Then, according to the formula:

[0101]

[0102] Calculate the original equipment energy loss P of the compressor comp1 is approximately 139.8 kilowatts.

[0103] Assume that in the original parameter set, the hydrogen inlet temperature of the pre-cooler is 300K, the refrigerant inlet temperature is 200K, the overall heat transfer coefficient is 200 watts per square meter Kelvin (W / (m 2 ·K)), and the heat transfer area is 5 square meters. According to the preset correspondence, obtain the original equipment energy loss P of the pre-cooler pre1 is 5 kilowatts (kW).

[0104] Assume that in the original parameter set, the hydrogen inlet temperature of the main cooler is 240K, the refrigerant inlet temperature is 80K, the overall heat transfer coefficient is 150 W / (m2 ·K), with a heat transfer area of 8 square meters, the original equipment energy loss P of the main condenser is obtained according to the preset corresponding relationship main1 is 5 kW.

[0105] The energy loss of the throttle valve itself is relatively small. Assume that the original equipment energy loss P of the throttle valve throttle1 is 1 kW.

[0106] Then, the total energy loss data caused by executing the hydrogen liquefaction process using the original parameter set without genetic algorithm processing is:

[0107] E total1 = P comp1 + P pre1 + P main1 + P throttle1 = 153.8 kW

[0108] After optimization by the genetic algorithm, the intake pressure of the compressor in the target parameter set is 0.9 MPa, the discharge pressure is 4.0 MPa, and other parameters remain unchanged. Calculate the optimized equipment energy loss P of the compressor comp2 is approximately 105.2 kW.

[0109] The optimized equipment energy loss P of the precooler pre2 is 3 kW, and the optimized equipment energy loss P of the main condenser main2 is approximately 6 kW. The optimized equipment energy loss P of the throttle valve throttle2 is approximately 0.8 kW.

[0110] Then, the total energy loss data caused by executing the hydrogen liquefaction process using the target parameter set processed by the genetic algorithm is:

[0111] E total2 = P comp2 + P pre2 + P main2 + P throttle2 = 115 kW

[0112] Therefore, the energy loss is reduced by:

[0113]

[0114] Among them, ΔE is the energy loss reduction ratio, and E total1 is the total energy loss data caused by executing the hydrogen liquefaction process using the original parameter set without genetic algorithm processing, and E total2 is the total energy loss data caused by executing the hydrogen liquefaction process using the target parameter set processed by the genetic algorithm.

[0115] Step S203: Execute the hydrogen liquefaction process using the target parameter set and obtain the operating parameters during the execution of the hydrogen liquefaction process. For details, please refer to Figure 1 Step S103 of the embodiment shown, which will not be elaborated here.

[0116] Step S204: Determine whether the operating parameters are abnormal. If the operating parameters are abnormal, correct the target parameter set to regulate the hydrogen liquefaction process. For details, please refer to Figure 1 Step S104 of the embodiment shown, which will not be elaborated here.

[0117] In some optional embodiments, a trained LSTM model is used to process the measured pressure data and liquid level data, complete result prediction and data comparison, correct the target parameter set, and regulate the hydrogen liquefaction process. It is found that when 1000 kg of hydrogen (calculated by a gas flow meter) is introduced, the mass of liquid hydrogen converted before regulating the hydrogen liquefaction process using the embodiment of the present invention is 708 kg, and the liquid hydrogen liquefaction rate is 70.8%. After regulating the hydrogen liquefaction process using the embodiment of the present invention, the mass of liquid hydrogen is 885 kg, and the liquefaction rate can still reach 88.5%. Therefore, the embodiment of the present invention can make up for problems such as equipment performance differences, can not only reduce energy loss, but also improve the liquefaction rate.

[0118] A regulation method for a hydrogen liquefaction process based on AI provided in this embodiment generates individuals by encoding a parameter set, updates the individuals in combination with evolutionary mechanisms such as genetic algorithms, constructs a fitness function directly related to energy loss, and converts the degree of energy loss optimization into a computable numerical index to ensure that the iteration always progresses around the goal of reducing energy loss, so that the finally obtained target parameter set reaches a better level in energy consumption optimization.

[0119] In this embodiment, a regulation method for a hydrogen liquefaction process based on AI is provided, which can be used in computer equipment. Figure 3 It is a flowchart of another regulation method for a hydrogen liquefaction process based on the embodiment of the present invention, as Figure 3 shown. This process includes the following steps:

[0120] Step S301: Encode a preset number of parameter sets. For details, please refer to Figure 2 Step S202 of the embodiment shown, which will not be elaborated here.

[0121] Step S302: Set the range intervals of multiple parameters in each parameter set. For details, please refer to Figure 1 Step S101 of the embodiment shown, which will not be elaborated here.

[0122] Step S303: Establish an energy loss calculation model according to the parameter set. For details, please refer to Figure 2 Step S202 of the embodiment shown, which will not be elaborated here.

[0123] Step S304: Define a preset fitness function. For details, please refer to Figure 1 Step S102 of the embodiment shown, which will not be elaborated here.

[0124] Step S305: Use the genetic algorithm to optimize and iterate the parameter set to obtain the target parameter set. For details, please refer to Figure 2 Step S202 of the embodiment shown, which will not be elaborated here.

[0125] Step S306: Modify the target parameter set to regulate the hydrogen liquefaction process. For details, please refer to Figure 1 Step S104 of the embodiment shown, which will not be elaborated here.

[0126] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A control method for a hydrogen liquefaction process based on AI, characterized in that, The method includes: Generating a preset number of parameter sets according to the equipment parameter intervals of multiple hydrogen liquefaction systems; the parameter sets are parameter sets obtained by randomly selecting multiple parameters in the equipment parameter intervals; Taking the energy loss of each hydrogen liquefaction system as the optimization objective, and using a preset fitness function to perform optimization iteration on the preset number of the parameter sets until a target parameter set is obtained; the preset fitness function is used to characterize the optimization degree of the energy loss of the hydrogen liquefaction system; the fitness value corresponding to the target parameter set is greater than a first preset value; Executing the hydrogen liquefaction process using the target parameter set, and obtaining the operating parameters during the execution of the hydrogen liquefaction process; Judging whether the operating parameters are abnormal. If the operating parameters are abnormal, correcting the target parameter set to regulate the hydrogen liquefaction process.

2. The method according to claim 1, wherein The generating a preset number of parameter sets according to the equipment parameter intervals of multiple hydrogen liquefaction systems includes: Obtaining the equipment parameter intervals corresponding to multiple key equipment belonging to each hydrogen liquefaction system; Performing random selection on multiple parameters in each equipment parameter interval for a preset number of times, and adding the parameters corresponding to each hydrogen liquefaction system obtained by each random selection to a set to obtain the preset number of the parameter sets.

3. The method according to claim 1 or 2, characterized in that, The taking the energy loss of each hydrogen liquefaction system as the optimization objective, and using a preset fitness function to perform optimization iteration on the preset number of the parameter sets until a target parameter set is obtained includes: Encoding the preset number of the parameter sets to obtain the first individuals of the preset number; Constructing the preset fitness function according to the energy loss of each hydrogen liquefaction system, and using the preset fitness function and the total energy loss data corresponding to each first individual to generate the fitness value corresponding to each first individual; Updating the first individuals of the preset number according to the fitness value to obtain a plurality of second individuals; Updating the first individuals using the second individuals, and returning to the step of generating the fitness value corresponding to each second individual using the preset fitness function until the iteration termination condition is reached to obtain a plurality of target individuals; Among the plurality of target individuals, selecting the optimal target individual whose fitness value is greater than the first preset value; Decoding the optimal target individual to obtain the target parameter set.

4. The method according to claim 3, wherein The constructing the preset fitness function according to the energy loss of each hydrogen liquefaction system includes: Constructing the preset fitness function according to the reciprocal of the energy loss of each hydrogen liquefaction system.

5. The method according to claim 3, characterized in that The updating the first individuals of the preset number according to the fitness value to obtain a plurality of second individuals includes: Performing probability analysis on each first individual according to the fitness value to obtain the target probability corresponding to each first individual; Select a plurality of first target individuals from the plurality of first individuals according to the target probability; the first individuals whose fitness value of the first target individuals is greater than a second preset value; the second preset value is less than the first preset value. Perform a crossover operation on each of the first target individuals to obtain a plurality of second target individuals. Perform a mutation operation on the plurality of second target individuals to obtain the plurality of second individuals.

6. The method according to claim 2, wherein The execution of the hydrogen liquefaction process using the target parameter set includes: Obtain a plurality of target parameters in the target parameter set, configure the plurality of target parameters to the corresponding key devices, and control the operation of the key devices to execute the hydrogen liquefaction process.

7. The method according to claim 1 or 2, characterized in that, The judgment of whether the operating parameters are abnormal includes: Input the historical operating parameters into the trained operating parameter prediction model to obtain the predicted value of the operating parameters at the current moment; the input of the operating parameter prediction model is the historical operating parameters, and the output of the operating parameter prediction model is the predicted value of the operating parameters. Compare the operating parameters with the predicted value of the operating parameters. If the difference between the operating parameters and the predicted value of the operating parameters is greater than a preset threshold, it is judged that the operating parameters are abnormal. If the difference between the operating parameters and the predicted value of the operating parameters is less than or equal to the preset threshold, it is judged that the operating parameters are normal.

8. The method according to claim 1 or 2, characterized in that, The correction of the target parameter set includes: Use the target parameter set to update the parameter set, and return the step of optimizing and iterating the target parameter set using a preset fitness function with the energy loss of each hydrogen liquefaction system as the optimization target until a target corrected parameter set is obtained, so as to correct the target parameter set.

9. The method according to claim 1 or 2, characterized in that, The correction of the target parameter set includes: Obtain the outlier, and remove the target parameter in the target parameter set corresponding to the outlier to correct the target parameter set.

10. The method according to claim 1 or 2, characterized in that, The method further includes: If the operating parameters are normal, generate a prompt message; the prompt message is used to prompt that the operating parameters are normal.

Images