Online proxy model assisted multi-objective optimization method for green hydrogen coupling hydrogenation process

Abstract

The invention provides a green hydrogen coupling hydrogenation process on-line agent model auxiliary multi-objective optimization method, which comprises the following steps: aiming at green hydrogen-grey hydrogen mixing fluctuation in a green hydrogen coupling hydrogenation process, constructing an economic and carbon emission reduction multi-objective operation parameter optimization problem, carrying out sensitivity analysis on operation parameters, and obtaining to-be-optimized operation parameters and a value range of the to-be-optimized operation parameters; the method comprises the following steps: establishing a first principle model for a hydrogenation process, performing Latin hypercube sampling, and establishing an offline proxy model by adopting Gaussian process regression; under the condition of mixed hydrogen fluctuation, an initial online agent model is generated based on the Gaussian process, the economic and carbon emission reduction multi-objective optimization problem is solved by adopting a rapid non-dominated sorting optimization method, and the online agent model is updated periodically. According to the method, the investment cost and the carbon emission reduction benefit are both considered, and multi-target balance optimization of the economic benefit and the carbon emission reduction under frequent fluctuation of green hydrogen-grey hydrogen mixing is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical hydrogenation, and specifically but not limited to a method for online surrogate model-assisted multi-objective optimization of a green hydrogen-coupled hydrogenation process. Background Art

[0002] With the low-carbon transformation of the global energy structure, the renewable energy power generation and hydrogen production technology is accelerating the reshaping of the hydrogen source structure in the chemical industry. Hydrogen is a key raw material for chemical products such as synthetic ammonia, ethylene glycol, and methanol. Coupling renewable energy power generation with electrolytic water hydrogen production technology and replacing traditional grey hydrogen with green hydrogen to supply the chemical hydrogenation process can simultaneously solve the dual problems of renewable energy consumption and chemical production decarbonization.

[0003] However, the most important sources of green hydrogen, namely wind energy and solar energy, have significant volatility, resulting in significant spatio-temporal imbalance characteristics in green hydrogen supply. The strict requirements of the hydrogenation process for hydrogen purity, pressure, and flow rate form a significant conflict with the fluctuating characteristics of green hydrogen. Traditional solutions rely on large-scale hydrogen storage facilities for peak shaving, which will significantly increase the system cost. For example, a large amount of CO2 is generated during the production of grey hydrogen in the traditional chemical hydrogenation production process, and the high energy consumption in the rectification process will also increase indirect carbon emissions. Therefore, some studies have utilized the new process of combining hydrogen generated by wind energy / solar energy combined power generation with subsequent chemical production. However, to maintain the hydrogen volatility below 10%, a large number of energy storage and hydrogen storage devices need to be built to suppress the fluctuations of green hydrogen. Such solutions increase the production cost of green hydrogen by more than 35%. Therefore, some scholars have tried to construct a hybrid hydrogen source system, which can effectively reduce the volatility of hydrogen used in chemical production without significantly increasing the investment cost. However, existing studies generally do not deeply explore the impact of the fluctuation of the green hydrogen-grey hydrogen ratio on the optimization of chemical operation parameters, resulting in the difficulty of the system to simultaneously achieve the optimal economic benefits and carbon emission reduction goals when dealing with frequent fluctuations of the green hydrogen-grey hydrogen ratio.

[0004] In view of this, a new method is needed to solve at least some of the above problems. Summary of the Invention

[0005] Aiming at one or more problems in the prior art, the present invention proposes a method for online surrogate model-assisted multi-objective optimization of a green hydrogen-coupled hydrogenation process, realizing the balanced optimization of multi-objectives of economic benefits and carbon emission reduction under the mixed fluctuation of green hydrogen and grey hydrogen.

[0006] The technical solution to achieve the object of the present invention is as follows:

[0007] An online surrogate model-assisted multi-objective optimization method for a green hydrogen-coupled hydrogenation process, including: by improving the hydrogenation process technology, converting the green hydrogen flow fluctuation into the fluctuation of the ratio of green hydrogen to gray hydrogen flow, constructing an economic and carbon emission reduction multi-objective operation parameter optimization problem, and establishing an offline surrogate model using Gaussian process regression; under the condition of mixed hydrogen fluctuation, generating an initial online surrogate model based on Gaussian process, using the fast non-dominated sorting optimization method to solve the economic-carbon emission reduction multi-objective optimization problem, and periodically updating the online surrogate model to improve the solution efficiency and achieve the balanced optimization of economic benefits and carbon emission reduction multi-objectives under frequent fluctuations of green hydrogen-gray hydrogen mixture. Specifically, the present invention includes the following steps:

[0008] (1) In the green hydrogen-gray hydrogen mixed hydrogenation process, green hydrogen is used as the main hydrogen source. When green hydrogen is insufficient, gray hydrogen is used for dynamic compensation, and when there is excess, it is supplied to the chemical process through power-to-heat, thereby converting the green hydrogen flow fluctuation into the dynamic change of the green hydrogen-gray hydrogen ratio.

[0009] (2) For the hydrogenation process under the fluctuation of green hydrogen-gray hydrogen mixture, construct an economic and carbon emission reduction multi-objective operation parameter optimization problem, conduct a sensitivity analysis on the operation parameters, and obtain the optimized decision variables and their value ranges.

[0010] The objective functions of maximizing economic benefits and minimizing carbon emissions are established as follows:

[0011]

[0012] Where x represents the set of operation variables, that is, x = [x1, x2,..., x n . The economic benefit ∑E i is calculated according to the profit per unit time, that is, the sum of the value of the main product and the value of the by-product produced per hour, and then subtracting the total cost. The total cost specifically includes various raw material costs ∑C j , the electricity cost C pow consumed during the process, and the total heat energy cost C heat consumed during the process. The carbon emission optimization target includes the direct carbon emission CE d converted from the organic matter that has not entered the product or has not been recycled, the indirect carbon emission CE i and the carbon emission brought by the H2 raw material

[0013] Take the purity of the produced product and the upper and lower limits of each operation parameter as constraints:

[0014]

[0015] Where ω i is the purity of the produced product, ω i,limited is the minimum purity limit of product i; xm,lowerbound and x m,upperbound represent the lower and upper limits of the m-th operating parameter respectively, and their magnitudes need to be determined according to the regulations in the process manual and sensitivity analysis.

[0016] (3) Conduct Latin hypercube sampling on the first-principles model (FPM), and use Gaussian process regression to establish an offline surrogate model.

[0017] FPM is a theoretical model established based on the basic laws of physics, chemistry, and mathematics. It starts from the most fundamental physical and chemical laws, and describes the properties and behaviors of the system by deriving formulas, without relying on empirical data or experimental fitting. FPM has significant advantages, including strong universality, strong prediction ability, and clear mechanism, etc. It can be applied to a wide range of conditions and scenarios, and can reveal the internal working mechanism and essential laws of the system. However, FPM also has some disadvantages, such as high computational complexity, especially in multi-phase or multi-component systems, which require a large amount of computing resources. In the chemical engineering field, FPM usually involves the basic equations of thermodynamics, kinetics, and transport phenomena, such as the first law of thermodynamics, mass conservation equation, and kinetic equation, so as to describe processes such as fluid flow, heat and mass transfer, and chemical reactions. Commonly used commercial simulation software such as Aspen Plus is used to construct FPM to solve complex chemical engineering process problems.

[0018] Generate N groups of data containing the operating parameter x and the perturbation variable d by Latin hypercube sampling, input them into the FPM, and use the optimized objective values output by the FPM to train the offline surrogate model. The Gaussian process surrogate model is defined as shown in the following formula:

[0019] y(x) = μ + ∈(x), (x) ~ N(0, σ 2 )

[0020] where μ represents the mean value estimated by the GP for y(x); ∈(x) is the error term, which follows a normal distribution with a mean of 0 and a variance of σ 2 . For two input variables x1 and x2 randomly selected from the samples, the correlation between their errors is calculated by the Gaussian exponential function, as shown in the following formula:

[0021]

[0022] where R(∈(x1), ∈(x2)) represents the correlation between the two error terms, and its value range is [0, 1]. The larger the value, the stronger the correlation; x d represents the dimension of the input variable x, θ k represents the scale parameter used to control the sensitivity of the GP to the input in the k-th dimension, l kIt represents the length scale parameter used to control the smoothness of the data points by the GP in the k-th dimension. To optimize the GP surrogate model, the hyperparameters θ k and l k need to be determined. Therefore, the method of maximizing the log marginal likelihood function is adopted, as shown in the following formula:

[0023]

[0024] where X = {x 1 , x 2 , …, x p} represents the inputs of all training data; f = {y(x 1 ), y(x 2 ), …, y(x p )} represents the outputs of all training data; p represents the number of training data; CM represents the P×P correlation matrix.

[0025] (4) Under the condition of fluctuating mixed hydrogen, an initial online surrogate model is generated based on the Gaussian process. The fast non-dominated sorting optimization method is used to solve the economic-carbon emission reduction multi-objective optimization problem, and the solution efficiency is improved by periodically updating the online surrogate model.

[0026] When significant changes occur in disturbance variables such as the green hydrogen-to-gray hydrogen ratio, feed flow rate, or product price, an initial population is constructed with the current disturbance variables and operating parameters of the Gaussian distribution. The NSGA-II algorithm is used in the offline surrogate model and optimized to the N off th generation through crossover, mutation, and selection;

[0027] Among them, the NSGA-II algorithm is an improved algorithm of the first-generation non-dominated sorting genetic algorithm NSGA based on non-dominated sorting and sharing. Its advantages lie in the fast non-dominated sorting method, the concept of crowding degree, and the elitist retention strategy. It has good accuracy and rapidity in solving multi-objective problems and is not easily trapped in local optimization. It generates offspring by using simulated binary crossover and polynomial mutation, then uses fast non-dominated sorting for individual selection, and retains elite solutions during this process. In fast non-dominated sorting, an individual has two key parameters, one is the non-dominated rank F, and the other is the crowding degree i d . When comparing different individuals, the non-dominated rank is first determined, and the higher the non-dominated rank, the better. In the case of the same non-dominated rank, the individual with a smaller crowding degree operator is better. If an individual z1 is superior to another individual z2 in all optimization objectives, then z1 dominates z2. All individuals that are not dominated by any other individual are assigned the non-dominated level F1. Then, the individuals with the assigned non-dominated level are removed, and similarly, the non-dominated levels F2, F3… are assigned. Finally, the crowding degree comparison operator of the individual is calculated, as shown in the following formula:

[0028]

[0029] All individuals can be sorted according to the non-dominated rank and crowding degree.

[0030] Take the operation parameters of the N off th generation and the current perturbation variables as inputs, perform real evaluation in the FPM to obtain the objective values f1 and f2, and store the evaluated individuals in the dataset S. If the maximum running time t max is not reached, the system enters the online optimization loop. When the number of individuals in the dataset S exceeds the online model capacity, by deleting the individual with the largest non-dominated rank, use the updated dataset to train the online surrogate model, regenerate the initial population for N on th generation NSGA-II optimization, and screen the non-dominated high-quality solutions through the convergence (CI) and diversity (DI) metrics for real evaluation in the FPM, continuously enrich the dataset S. The online surrogate model-assisted multi-objective optimization problem is shown as follows:

[0031]

[0032] subject to X * ∈ Trial Solutions

[0033] where, Z * is the ideal point composed of the optimal values of each objective in the current population (such as the minimum value of each objective in the minimization problem), F(X * ) is the predicted objective value of the solution by the surrogate model; X p is the individual in the parental population, ‖·‖ is the Euclidean distance; CI(X * ) is used to measure the closeness between the predicted objective value F(X * ) of the solution and the ideal point Z * . The smaller the CI value, the closer the predicted objective value of the solution is to the ideal point, and the better the convergence; DI(X * ) is used to measure the difference between the new solution X * and the nearest individual in the parental population. The larger the DI value, the greater the difference between the new solution and the parental population, and the better the diversity. Finally, after reaching the time threshold, the system extracts the non-dominated optimal solutions from the dataset S and outputs the actually available optimization parameters through inverse normalization processing.

[0034] Compared with the prior art, the present invention adopts the above technical solutions and has the following technical effects:

[0035] 1. The online surrogate model-assisted multi-objective optimization method for the green hydrogen-coupled hydrogenation process of the present invention uses green hydrogen as the main hydrogen source in the green hydrogen-coupled hydrogenation process. When green hydrogen is insufficient, gray hydrogen is used for dynamic compensation. When green hydrogen is in excess, it is converted into heat energy through power-to-heat for use in the hydrogenation process. In contrast, the traditional methods are either to use gray hydrogen with a huge carbon footprint for the hydrogenation process or to adopt the green hydrogen + hydrogen storage solution with excessively high investment costs. The solution of the present invention takes into account both the investment cost and the carbon emission reduction benefit, and on this basis, a first-principles model is established for operation optimization to achieve the online balance of economy and carbon emissions.

[0036] 2. The online surrogate model-assisted multi-objective optimization method for the green hydrogen-coupled hydrogenation process of the present invention constructs an economic and carbon emission reduction multi-objective operation parameter optimization problem based on the dynamic change of the green hydrogen and gray hydrogen flow ratio. In contrast, the existing technologies mainly focus on the operation optimization under fixed working conditions. After large working condition changes, it usually requires time-consuming re-optimization and is difficult to adapt to the dynamic change of the green hydrogen and gray hydrogen ratio caused by the fluctuation of renewable energy.

[0037] 3. The online surrogate model-assisted multi-objective optimization method for the green hydrogen-coupled hydrogenation process of the present invention constructs a surrogate-assisted multi-objective optimization algorithm APB-NSGAII integrating adaptive parameters, Pareto double-index sampling, and NSGA-II, effectively utilizes the offline surrogate model established in the early stage and updates the online surrogate model for optimization, reduces the number of FPM evaluations with a long time consumption, and realizes the re-acquisition of the optimal solution with less operation time of the optimization algorithm after the change of the hydrogen ratio. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings are used to provide a further understanding of the present invention and, together with the description, are used to explain the embodiments of the present invention and do not constitute a limitation to the present invention. In the drawings:

[0039] Figure 1 The process flow diagram of the hydrogenation of dimethyl oxalate to ethylene glycol under the green hydrogen-gray hydrogen mixed hydrogen source according to an embodiment of the present invention is shown.

[0040] Figure 2 The flow chart of the online surrogate model-assisted multi-objective optimization method for the green hydrogen-coupled hydrogenation process of the present invention is shown.

[0041] Figure 3 The comparison diagram of the optimization results of different algorithms when green hydrogen is sufficient according to an embodiment of the present invention is shown.

[0042] Figure 4 The comparison diagram of the optimization results of different algorithms when green hydrogen and gray hydrogen are mixed according to an embodiment of the present invention is shown.

[0043] Figure 5 The comparison diagram of the optimization results of a 3-month test cycle according to an embodiment of the present invention is shown. Detailed implementation manners

[0044] To further understand the present invention, the preferred implementation manners of the present invention will be described below in conjunction with embodiments. However, it should be understood that these descriptions are only for further explaining the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0045] The description of this part is only for typical embodiments, and the present invention is not limited to the scope described in the embodiments. Combinations of different embodiments, mutual replacement of some technical features in different embodiments, and mutual replacement of the same or similar prior art means and some technical features in the embodiments are also within the scope described and protected by the present invention.

[0046] Taking the process of hydrogenation of dimethyl oxalate (DMO) to synthesize ethylene glycol (EG) driven by mixed green hydrogen as an example, an online surrogate model-assisted multi-objective optimization method for the green hydrogen coupling hydrogenation process will be introduced.

[0047] First, a process for hydrogenation of DMO to synthesize EG with a mixed hydrogen source of green hydrogen and gray hydrogen is proposed. In the hydrogenation process, green hydrogen is used as the main hydrogen source. When the green hydrogen is insufficient, it is supplemented by gray hydrogen. When the green hydrogen is in excess, it is supplied to the chemical process through power-to-heat, reducing the energy storage and hydrogen storage costs by about 80% and converting the fluctuation of the green hydrogen flow rate into the dynamic change of the ratio of green hydrogen to gray hydrogen. The process flow diagram of the hydrogenation of dimethyl oxalate to synthesize ethylene glycol with a green hydrogen-gray hydrogen mixed hydrogen source is as Figure 1 shown.

[0048] The hydrogenation of DMO to synthesize EG is a typical reaction-separation-circulation process. First, the DMO feed is mixed with the liquid-phase methanol stream, and then mixed with the hydrogen feed and another three gas-phase recycle streams to form a mixed feed stream. These mixed raw materials are heated up by a preheater at the reactor inlet and then enter the reactor to react. The heat source of the preheater is the heat exchange with the reactor outlet stream, so as to improve the energy efficiency and save energy. In the reactor, the DMO hydrogenation reaction occurs to produce ethylene glycol, methanol and ethanol. The main function of the methanol recovery tower is to gather the reusable hydrogen and methanol vapor at the top of the tower. After flashing, most of the gas remains in the gas phase and is recycled back to the mixed feed stream, and part of the methanol becomes the liquid phase, which is used as the liquid-phase product of the methanol recovery tower. The by-products such as methanol and ethanol remaining at the bottom of the tower are separated out in the gas-phase stream of the dehydration tower. The liquid-phase stream at the bottom of the dehydration tower is mainly EG and methyl glycolate (MG), which enter the EG product tower for further rectification. DMO and MG are drawn out from the top of the tower and returned to the feed stream, and polyester-grade ethylene glycol is obtained at the bottom of the tower.

[0049] Aiming at the requirements of maximizing the economic benefits and minimizing the total carbon emissions in the process of synthesizing EG from mixed green hydrogen and DMO, an economic - carbon emission multi - objective optimization model is established, and the purities of EG and CH4O as well as the upper and lower limits of each operating parameter are used as constraints:

[0050]

[0051] Among them, x represents a set of 8 operating variables, that is, x = [x1, x2, …, x8]. The economic benefit is calculated according to the profit per unit time, that is, the value E EG of the main product EG produced per hour, the value E ET of the by - product ET, and the value E CH4O of the by - product CH4O, and then subtract the total cost. The total cost specifically includes the DMO raw material cost C DMO , the H2 raw material cost , the electric energy cost C pow consumed in the process, and the total heat energy cost C heat consumed in the process. The carbon emission optimization objectives include the direct carbon emission CE d converted from the organic matter that does not enter the product or is not recycled, the indirect carbon emission CE i , and the carbon emission brought by the H2 raw material

[0052]

[0053] Among them, ω EG and ω CH4O are the purities of the products EG and CH4O respectively; x m、lowerbound and x m、upperbound represent the lower and upper limits of the m - th operating parameter respectively, and their magnitudes need to be determined according to the regulations of the process manual and sensitivity analysis.

[0054] The calculations of each part of the optimization objective are shown as follows:

[0055]

[0056] Among them, F wEG , F wET , F wCH4O , F wDMO , F wH2green , F wH2gray represent the mass flow rates of the product EG, the by - product ET, the by - product CH4O, the consumed DMO, the consumed green hydrogen, and the consumed grey hydrogen respectively. P EG , P ET , P CH4O , P DMO , P H2green , P H2grayRepresent the unit prices of product EG, by-product ET, by-product CH4O, DMO raw material, green hydrogen raw material, and gray hydrogen raw material, respectively. E pow 、E powg 、E heat 、E heatg Represent the total power consumed in the production process, the green electricity power consumed, the total heat energy consumed, and the heat energy converted from green electricity, respectively. P pow 、P powg 、P heat 、P heatg Represent the grid power supply price, the green electricity cost price, the average heat energy price of the steam pipe network, and the electricity-to-heat conversion cost price, respectively. F nc Represents the molar flow rate of various organic compounds that do not enter the product or by-product, and η CE Represents the proportion of unit molar flow rate of organic compounds converted into CO2. CE pow and CE heat Represent the indirect carbon emissions corresponding to electricity consumption and the indirect carbon emissions corresponding to heat energy consumption, respectively. η CEpow 、η CEpowg 、η CEheat 、η CEheatg Represent the unit carbon emissions of grid power supply, green electricity, steam pipe network, and green electricity-to-heat conversion, respectively. η P 、η COM 、η C 、η h Represent the working efficiencies of the pump, compressor, distillation column, and heat exchanger, respectively. E P 、E P1 、E P2 、E P3 Represent the total power of the pump, P1 power, P2 power, and P3 power, respectively. E COM 、E COM1 、E COM2 、E COM3 Represent the total power of the compressor, COM1 power, COM2 power, and COM3 power, respectively. E C 、E C1 、E C2 、E C3 Represent the total heat consumption of the distillation column, C1 heat consumption, C2 heat consumption, and C3 heat consumption, respectively. C H2green and C H2gray Represent the unit carbon emissions of green hydrogen and gray hydrogen, respectively. The default reference prices are: ethylene glycol 4200 yuan / ton, methanol 2500 yuan / ton, ethanol 6900 yuan / ton, DMO reference cost 3000 yuan / ton, gray hydrogen cost 20448 yuan / ton, grid power supply cost 0.725 yuan / kwh, and heat energy cost 26.28 yuan / GJ.

[0057] For safety reasons, the ranges of operating parameters set in the process design manual are often too conservative. However, if the operating parameter ranges are too large, it often leads to too few feasible solutions in the initial population during the optimization process. Therefore, the sensitivity analysis method is used to select operating parameters and determine their ranges. In this case, 8 main operating variables involved in the process flow are used, including reactor temperature, pressure, pressure drop, molar ratio of hydrogen to dimethyl oxalate (HDMR), molar recovery rate of methanol at the top of the methanol recovery column, and reflux ratios of three towers. Subsequently, according to the sensitivity analysis, the value ranges of the 8 operating variables are selected, and the sequential method is used to determine their typical values. The default values and value ranges of the 8 operating variables x and 5 disturbance variables d = [d1, d2, d3, d4, d5] are shown in the following table:

[0058]

[0059] Normalize the FPM and use Latin hypercube sampling to generate 2000 groups of data containing 8 operating parameters x and 5 disturbance variables d, input them into the FPM, and establish an offline surrogate model using the optimized objective values output by the FPM.

[0060] When significant changes occur in disturbance variables such as the ratio of green hydrogen to grey hydrogen, feed flow rate, or product price in d, construct an initial population with the current disturbance variable d c and the operating parameter x with a Gaussian distribution, use the NSGA-II algorithm in the offline surrogate model, and optimize to the N off th generation through crossover, mutation, and selection. Use the operating parameters of the N off th generation and the current disturbance variable as inputs for real evaluation in the FPM to obtain the objective values f1 and f2, and store the evaluated individuals (numbering S n ) in the dataset S. If the maximum running time t max is not reached, the system enters the online optimization loop. When the number of individuals S n in the dataset S exceeds the online model capacity S a , by deleting the individuals with the largest non-dominated rank, use the updated dataset to train the online surrogate model, regenerate the initial population for N on generations of NSGA-II optimization, screen the non-dominated high-quality solutions through convergence (CI) and diversity (DI) indicators for real evaluation in the FPM, continuously enrich the dataset S, and finally, after reaching the time threshold, the system extracts the non-dominated optimal solutions from the dataset S and outputs the actually available optimized parameters through denormalization processing.

[0061] In the optimization comparison experiment, two typical cases of excessive green hydrogen and insufficient green hydrogen were selected respectively. When there is excessive green hydrogen, the system converts the excess hydrogen energy into heat energy to reduce the carbon emissions caused by heat energy consumption. When the hydrogen is insufficient, the system introduces gray hydrogen and adjusts the operating parameters to achieve the balance between economic benefits and carbon emissions. The optimization results of three comparison algorithms, namely APB-NSGAII, NSGA-II, Pareto global efficient optimization algorithm (ParEGO), and surrogate model-assisted multi-objective particle swarm optimization (SAPSO), were compared under the same operating condition of 1 hour. The experimental results are as Figure 3 , Figure 4 shown.

[0062] In the simulation of online optimization experiment, continuous three-month data were used, and the carbon tax was set at 450 yuan / ton CO2. For the convenience of comparison, the change time of the five interference variables was set at the same time every day. The optimization results of APB-NSGAII optimization, NSGA-II optimization, and the fixed operating parameter strategy were compared respectively. The experiment counted the daily economic benefits including carbon tax. To better compare different operating strategies, we took the economic benefits including carbon tax under the fixed operating parameters as the benchmark, and the optimization results of the two algorithms, APB-NSGAII and NSGA-II, are as Figure 5 shown.

[0063] The optimization results show that within the cumulative three-month period, compared with the fixed parameter strategy, the economic benefits including carbon tax increased by 2,291,829 yuan, and the carbon emissions decreased by 20,418 tons of CO2. Compared with the optimization results of NSGA-II, the economic benefits including carbon tax increased by 9,005,20 yuan, and the carbon emissions decreased by 11,129 tons of CO2. From the above experimental results, it can be seen that compared with NSGA-II and the fixed operating parameter strategy, APB-NSGAII can more effectively balance the goals of improving economic benefits and reducing carbon emissions during long-term operation, which proves its superiority in the operation optimization of the complex hydrogenation process coupled with green hydrogen.

[0064] The description and application of the present invention here are illustrative, and it is not intended to limit the scope of the present invention to the above embodiments. The related descriptions of effects or advantages in the specification may not be reflected in the actual experimental examples due to the uncertainty of specific condition parameters or other factors. The related descriptions of effects or advantages are not used to limit the scope of the invention. The deformations and changes of the disclosed embodiments here are possible, and the substitutions and equivalent various components of the embodiments are well-known to those of ordinary skill in the art. Those skilled in the art should clearly understand that without departing from the spirit or essential characteristics of the present invention, the present invention can be implemented in other forms, structures, arrangements, proportions, and with other components, materials, and parts. Without departing from the scope and spirit of the present invention, other deformations and changes can be made to the disclosed embodiments here.

Claims

1. An online agent model-assisted multi-objective optimization method for green hydrogen coupled hydrogenation process, characterized in that: The following steps are involved: S1. Obtain the dynamic change of the ratio of green hydrogen to grey hydrogen flow rate during green hydrogen coupled hydrogenation; S2. Based on the dynamic changes in the ratio of green hydrogen to grey hydrogen flow, a multi-objective operation parameter optimization problem of economy and carbon emission reduction is constructed, and a sensitivity analysis is performed on the operation parameters to obtain the operation parameters to be optimized and their value ranges; S3. Establish a first principle model FPM for the green hydrogen coupled hydrogenation process, perform Latin hypercube sampling on the FPM, and establish an offline proxy model based on the sampled data using Gaussian process regression; S4. Construct an agent-assisted multi-objective optimization algorithm APB-NSGAII that integrates adaptive parameters, Pareto dual-index sampling and NSGA-II. The algorithm first generates an initial online agent model based on the Gaussian process GP of adaptive parameters, and then uses a fast non-dominated sorting optimization method to solve the multi-objective operation parameter optimization problem of economy and carbon emission reduction, obtain the optimal operation parameters, and periodically update the online agent model. Multi-objective optimization and model update are performed alternately until the optimization running time is exhausted.

2. The online agent model-assisted multi-objective optimization method for green hydrogen coupled hydrogenation process according to claim 1 is characterized in that: The green hydrogen coupled hydrogenation process described in S1 uses green hydrogen as the main hydrogen source, and uses grey hydrogen for dynamic compensation when green hydrogen is insufficient. When green hydrogen is in excess, it is converted into thermal energy through electric-to-heat conversion for use in the hydrogenation process.

3. The online agent model-assisted multi-objective optimization method for green hydrogen coupled hydrogenation process according to claim 1 is characterized in that: The objective function of the multi-objective operation parameter optimization problem of economy and carbon emission reduction described in S2 is: Among them, maxf1(x) represents the maximization of economic benefits, minf2(x) represents the minimization of carbon emissions, x represents the set of operating variables, E i represents the economic benefit of the i-th product, C j represents the cost of the jth raw material, C pow Represents the cost of consumed electricity, C heat Represents the total heat energy cost consumed, CE d Indicates direct carbon emissions from organic matter that does not enter the product or is not recycled. i represents indirect carbon emissions, Represents the carbon emissions from hydrogen raw materials; The constraints of the objective function are: Among them, ω i For the purity of the product, ω i、limited is the minimum purity limit of product i; x m、lowerbound and x m、upperbound Represent the lower limit and upper limit of the mth operating parameter respectively.

4. The online agent model-assisted multi-objective optimization method for green hydrogen coupled hydrogenation process according to claim 1 is characterized in that: S3 specifically includes: S3-1. Establish the FPM of green hydrogen coupled hydrogenation process based on the kinetic and thermodynamic mechanism of hydrogenation process, and use Latin hypercube sampling to generate N sets of data including operating parameters x and disturbance variables d from the model; S3-2, input the N groups of data into FPM, and use the optimization objective function value output by FPM to train the GP-based offline proxy model, the proxy model is as follows: y(x)=μ+∈(x)、(x)~N(0、σ 2 ) Among them, μ represents the mean of GP's estimate of y(x); ∈(x) is the error term, which has a mean of 0 and a variance of σ 2 Normal distribution of S3-3, two input variables x1 and x2 are randomly selected from the sample, the correlation between the error terms of x1 and x2 represents the dependency between the function values ​​corresponding to different input points, and the Gaussian exponential function is selected as the kernel function. The calculation formula of the Gaussian exponential function is as follows: Among them, R(∈(x1),∈(x2)) represents the correlation between the two error terms, and its value range is [0, 1]; x d represents the dimension of the input variable x, θ k represents the scale parameter used to control the sensitivity of GP to the input in the kth dimension, l k represents the length scale parameter used to control the degree of smoothing of data points by GP in the kth dimension; S3-4. Determine the hyperparameter θ by maximizing the log-marginal likelihood function k and l k , the maximized log-marginal likelihood function is as follows: Where X = {x 1 、x 2 , …, x p } represents the input of all training data; f = {y(x 1 )、y(x 2 ),…,y(x p )} represents the output of all training data; θ is the set of all hyperparameters; p(f|X, θ) represents the conditional probability density function of the output f given the input data X and hyperparameter θ; n represents the number of training data; CM represents the P×P correlation matrix; T represents transpose.

5. The online agent model-assisted multi-objective optimization method for green hydrogen coupled hydrogenation process according to claim 1 is characterized in that: S4 specific steps: S4-1. When a significant change is detected in disturbance variables such as the ratio of green hydrogen to gray hydrogen, feed flow rate or product price, an initial population is constructed using the current disturbance variables and operating parameters of the Gaussian distribution; S4-2. Execute the NSGA-II algorithm in the offline agent model to perform optimization iterations to the Nth off Generation, get the optimized operating parameters and input them into FPM for evaluation, get the target values ​​f1, f2, and store the evaluated individuals into the data set S; Then the APB-NSGAII algorithm enters the online optimization loop; S4-3, in the online optimization cycle, when the number of individuals in the data set S exceeds the individual capacity of the online agent model, the individuals with the lowest non-dominated level are deleted to update the data set, and then the online agent model is trained and updated, and the initial population is generated by the online agent model for NSGA-II optimization to the Nth on generation; S4-4, from the Nth on The optimal solutions of the generation are selected, and the high-quality solutions with convergence index and diversity index that are not dominated are input into FPM for evaluation. The evaluated solutions are put into the data set S. If the running time of the optimization algorithm does not reach the maximum running time t max , then continue the online optimization cycle. If t max , then the optimal solution in the data set S is output; the multi-objective optimization problem for screening high-quality solutions where the convergence index and diversity index are not dominated is as follows: subject to X * ∈Trial Solutions Among them, Z * is the ideal point composed of the optimal values ​​of each target in the current population, F(X * ) is the predicted target value of the surrogate model of the solution, X p is an individual in the parent population, ∥·∥ is the Euclidean distance, CI(X * ) is used to measure the predicted target value of the solution F(X * ) and the ideal point Z * The degree of closeness, DI(X * ) is used to measure the new solution X * The degree of difference from the nearest individual in the parent population; after finally reaching the time threshold, the system extracts the non-dominated optimal solution from the data set S and outputs the actually available optimization parameters through denormalization.

Images