Security domain mapping and optimization method, system and equipment for electro-hydrogen ammonia alcohol comprehensive energy system integrated with deep learning, and storage medium

Abstract

The invention discloses an electro-hydrogen-ammonia-alcohol comprehensive energy system security domain mapping and optimization method, system and device integrated with deep learning, and a storage medium, and relates to the field of guaranteeing safe and economical operation of a power system, and the method comprises the steps: building a park system security domain model based on the production and consumption data of the electro-hydrogen-ammonia-alcohol industry and considering the multi-energy flow balance constraint; solving the model by using an improved vertex search algorithm; constructing a feature set based on a solving result, and designing and training an LSTM-BiLSTM-BiGRU integrated learning model; calculating a security domain space index after the evaluation model is optimized; establishing a collaborative optimization operation model in combination with a distribution network safety and economy operation target setting function, solving a production and consumption scheduling strategy, and realizing system optimization; the method can optimize the electric hydrogen ammonia alcohol production consumption scheduling, reduce the operation cost, improve the energy utilization rate, guarantee the safety and stability of the park energy system, and promote the economic operation of the comprehensive energy system.

Description

Technical Field

[0001] This invention relates to the field of ensuring the safe and economical operation of power systems, and in particular to a method, system, device, and storage medium for security domain mapping and optimization of an integrated energy system combining electricity, hydrogen, ammonia, and alcohol using deep learning. Background Technology

[0002] Integrated energy systems that couple multiple energy carriers such as electricity, hydrogen, ammonia, and alcohol can effectively solve the problem of energy absorption caused by the randomness and volatility of renewable energy, and achieve cross-seasonal and cross-regional energy storage and efficient utilization, thus contributing to the deep decarbonization of the energy sector. As a key link in the convergence, conversion, and distribution of multiple energy flows, the safe, stable, and economically efficient operation of the park-level integrated energy system for electricity, hydrogen, ammonia, and alcohol is of paramount importance.

[0003] Existing technologies still have significant shortcomings when dealing with such highly coupled, multi-energy-flow interaction complex systems: traditional security domain analysis methods are mainly geared towards power systems, making it difficult to accurately characterize the comprehensive security boundary of dynamic conversion processes involving multiple energy sources such as hydrogen production, storage, transportation, ammonia, and methanol synthesis, and also making it difficult to achieve rapid solutions, with a trade-off between model accuracy and computational efficiency; the high-dimensional and non-convex characteristics of the system security domain boundary make it difficult to solve using traditional optimization methods that rely on physical models, and it is also difficult to establish a rapid mapping relationship between the system's operating state and the security domain, thus failing to support real-time scheduling decisions; most existing operation optimization models consider safety and economic objectives separately, lacking a collaborative mechanism to achieve both security domain improvement and economic optimization within a unified framework; these problems severely restrict the maximization of efficiency and large-scale application of integrated energy systems of electricity, hydrogen, ammonia, and methanol. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention provides a method, system, device and storage medium for security domain mapping and optimization of an integrated energy system for hydrogen, ammonia and alcohol using deep learning.

[0005] Therefore, the technical problem solved by this invention is: To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, comprising: Based on the production and consumption data, topology parameters, and historical operation data of the electro-hydrogen, ammonia, and alcohol industry, and considering the constraints of multi-energy flow balance related safety equations and inequalities, multiple park system safety domain models considering electro-hydrogen, ammonia, and alcohol energy are established. Based on the campus system security domain model, multi-energy constraint generation is coupled, and the final vertex set is calculated based on the improved vertex search algorithm to solve the campus system security domain model; Based on the obtained security domain of the park system, a feature set is constructed with the green park system operation status characteristics and security domain boundary information as the mapping relationship. Based on the feature set, an LSTM-BiLSTM-BiGRU ensemble learning model is designed and trained. The trained LSTM-BiLSTM-BiGRU ensemble learning model is evaluated, and the optimal model is obtained. The spatial index of the campus security domain is then calculated. Based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, an objective function is set, and a collaborative optimization operation model of the integrated energy distribution network system is established. The production and consumption scheduling strategy of hydrogen, ammonia, and methanol is obtained by solving the model, thereby achieving optimization of the safe and economical operation of the system.

[0006] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: Based on the production and consumption data, topology parameters, and historical operating data of the electro-hydrogen, ammonia, and methanol industry, and considering the constraints of multi-energy flow balance-related safety equations and inequalities, multiple safety domain models for the industrial park system considering electro-hydrogen, ammonia, and methanol energy are established, including: Construct a safe operation domain model for a park system that integrates the conversion and flow of multiple energy sources, including electricity, hydrogen, ammonia, and methanol, including the physical characteristics and operational constraints of the coupled power subsystem, hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem; The hydrogen energy subsystem model integrates the electrolytic hydrogen production, hydrogen storage, and dynamic transportation processes, and uses a position state matrix to describe the scheduling logic of hydrogen transportation vehicles between hydrogen production stations and chemical synthesis facilities in order to determine the spatiotemporal allocation of hydrogen. The ammonia synthesis subsystem model, based on hydrogen provided by the hydrogen energy subsystem and combined with nitrogen input, determines ammonia production through a chemical reaction process with dynamic adjustment characteristics.

[0007] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: The establishment of multiple park system security domain models considering the energy of electric hydrogen, ammonia, and methanol, based on the industrial production and consumption data, topology parameters, and historical operating data of the electric hydrogen, ammonia, and methanol industry, and considering the constraints of multi-energy flow balance correlation security equations and inequalities, also includes: The methanol synthesis subsystem model is based on hydrogen provided by the hydrogen energy subsystem and carbon dioxide captured and compressed from a coal-fired power plant, and determines the methanol yield through a catalytic synthesis reaction. The power subsystem model provides power to the power consumption units in the hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem, and meets grid operation constraints including unit output, ramping, carbon emissions, and power flow safety.

[0008] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: The process of solving the campus system security domain model by coupling multi-energy constraint generation with an improved vertex search algorithm to calculate the final vertex set includes: Define high-dimensional decision variables that include flexible resource and power decision variables of hydrogen-ammonia-alcohol; calculate the initial vertices of the safety domain of the multidimensional integrated park, define the boundary normal vector, take random vertices to form a closed region and determine the initial normal, and calculate the initial vertices.

[0009] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: The method of solving the campus system security domain model by coupling multi-energy constraint generation and calculating the final vertex set based on an improved vertex search algorithm also includes: The initial vertex coordinates are transformed, and new vertex coordinates are calculated based on the transformed vertices. The calculation is repeated to obtain the safety boundary. A solution termination condition is set. When the second norm of the normal vector of a certain line is less than the set value, the search for boundary vertices is stopped. The obtained vertices determine the boundary of the multidimensional constrained safety domain, and the safety domain of the park system is obtained.

[0010] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: Based on the solved security domain of the park system, a feature set is constructed with the green park system operation status characteristics and security domain boundary information as the mapping relationship. The LSTM-BiLSTM-BiGRU ensemble learning model is designed and trained based on this feature set, including: Using the park system security domain as the output label and high-dimensional space decision variables containing hydrogen-ammonia-alcohol flexible resource and power decision variables as input features, the input feature data is preprocessed to construct a feature set; the constructed feature set is divided into a training set and a test set; an LSTM-BiLSTM-BiGRU ensemble learning model is designed, which is composed of LSTM, BiLSTM and BiGRU modules in sequence; the designed ensemble learning model is trained using the training set.

[0011] The beneficial effects of this preferred technical solution are as follows: by constructing a feature set with the security domain of the park system as the output label, designing and training an LSTM-BiLSTM-BiGRU ensemble learning model, the advantages of each module can be fully utilized, effectively mining the complex relationship between the system operating state and the security domain boundary, and improving the accuracy and reliability of prediction.

[0012] As a preferred scheme for a safety domain mapping and optimization method integrating deep learning into an electro-hydrogen-ammonia-ethanol integrated energy system, the following is proposed: The process includes evaluating the trained LSTM-BiLSTM-BiGRU ensemble learning model, obtaining the optimal model, and then calculating the spatial indices of the campus security domain, including: The regression data of the ensemble learning model is compared with the actual data. The mean absolute error, mean relative error, and mean squared error are selected as model evaluation indicators. The difference between the regression data and the actual data is measured by the model evaluation indicators. The trained ensemble learning model is evaluated, and the optimal model is selected from the evaluation results. The spatial index of the park security domain is calculated based on the obtained optimal model.

[0013] The beneficial effects of this preferred technical solution are as follows: by using multi-index evaluation of the trained ensemble learning model, the optimal model is selected to calculate the spatial index of the park's safety domain, which can ensure the scientificity and accuracy of the index calculation and provide a reliable basis for subsequent production and consumption scheduling strategy optimization.

[0014] Secondly, this invention provides a safety domain mapping and optimization system for an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, comprising: The park security domain modeling module is used to establish multiple park system security domain models that consider the energy of electric hydrogen, ammonia, and alcohol based on the production and consumption data, topology parameters, and historical operation data of the electric hydrogen, ammonia, and alcohol industry, and considering the constraints of multi-energy flow balance related security equations and inequalities. The model solving module is used to solve the campus system security domain model by coupling multi-energy constraint generation based on the campus system security domain model and calculating the final vertex set based on the improved vertex search algorithm. The feature learning model training module is used to construct a feature set based on the obtained security domain of the park system, with the green park system operation status features and security domain boundary information as the mapping relationship, and to design and train an LSTM-BiLSTM-BiGRU ensemble learning model based on the feature set. The index calculation module is used to evaluate the trained LSTM-BiLSTM-BiGRU ensemble learning model, and calculate the spatial index of the park security domain after obtaining the optimal model. The production and consumption scheduling optimization module is used to set an objective function based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, establish a collaborative optimization operation model of the integrated energy distribution network system, solve for the production and consumption scheduling strategy of hydrogen, ammonia and alcohol, and realize the optimization of safe and economical operation of the system.

[0015] Thirdly, the present invention provides a computer device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the integrated deep learning method for safety domain mapping and optimization of the electro-hydrogen-ammonia-ethanol integrated energy system are implemented.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of a method for security domain mapping and optimization of an integrated deep learning-based electrohydrogen-ammonia-ethanol energy system.

[0017] The beneficial effects of this invention are as follows: By integrating an improved vertex search algorithm, multiple deep learning networks, and a two-layer optimization framework, this invention forms a complete methodological system from accurate characterization of the safety domain to system-wide collaborative optimization. The integrated energy system of hydrogen, ammonia, and methanol faces multiple challenges, including difficulties in modeling the safety domain under strong coupling of multiple energy flows, complex solutions to high-dimensional boundaries, and difficulties in balancing safety and economic objectives. This invention effectively addresses these challenges by fully utilizing system operating data and physical models, through a progressive step-by-step approach and advanced hybrid model design. The application of the improved vertex search algorithm makes the calculation of high-dimensional non-convex safety domain boundaries more efficient and accurate. The introduction of the LSTM-BiLSTM-BiGRU integrated deep learning model enables intelligent and rapid mapping from system operating characteristics to safety domain boundaries, overcoming the bottleneck of slow calculation in traditional physical models. Finally, by constructing a two-layer optimization model with safety domain indicators and economic efficiency as objectives, the system's economic efficiency within the safety operating boundary is maximized. This invention not only provides a solid theoretical tool for the panoramic perception and optimized operation of complex integrated energy systems but also opens up new technical paths for improving renewable energy consumption and ensuring energy security, demonstrating broad engineering application prospects and value. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an overall flowchart of a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system provided by the present invention.

[0020] Figure 2 This is a simulation example of the park-distribution network structure diagram in the "Safety Domain Mapping and Optimization Method for Integrated Deep Learning-based Electro-hydrogen-ammonia-ethanol Integrated Energy System" provided by this invention.

[0021] Figure 3This is a simulation example of the LSTM-BiLSTM-BiGRU structure diagram in the "Safety Domain Mapping and Optimization Method for Integrated Deep Learning in an Electro-hydrogen-ammonia-ethanol Integrated Energy System" provided by this invention.

[0022] Figure 4 The simulation example shows the wind power and photovoltaic output curves in a method for safety domain mapping and optimization of an integrated deep learning-based energy system for hydrogen, ammonia, and alcohol provided by this invention.

[0023] Figure 5 This is a load demand curve in a simulation example of a safety domain mapping and optimization method for an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system provided by this invention.

[0024] Figure 6 This is a simulation example of the integrated park safety domain map in a deep learning-integrated deep learning-based safety domain mapping and optimization method for an integrated energy system of hydrogen, ammonia, and alcohol provided by this invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, comprising: S1: Based on the production and consumption data, topology parameters and historical operation data of the electro-hydrogen, ammonia and alcohol industry, and considering the constraints of multi-energy flow balance related safety equations and inequalities, establish multiple park system safety domain models that consider electro-hydrogen, ammonia and alcohol energy. S2: Based on the campus system security domain model, coupled multi-energy constraint generation, and based on the improved vertex search algorithm, the final vertex set is calculated to solve the campus system security domain model; S3: Based on the obtained security domain of the park system, construct a feature set with the green park system operation status characteristics and security domain boundary information as the mapping relationship, and design and train an LSTM-BiLSTM-BiGRU ensemble learning model based on the feature set. S4: Evaluate the trained LSTM-BiLSTM-BiGRU ensemble learning model, and calculate the park security domain spatial index after obtaining the optimal model. S5: Based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, an objective function is set to establish a collaborative optimization operation model for the integrated energy distribution network system. The production and consumption scheduling strategy of hydrogen, ammonia, and methanol is obtained by solving the model, thereby achieving optimization of the safe and economical operation of the system.

[0027] It should be noted that, through steps S1-S5, this invention achieves the mapping and optimization of the safety domain of the integrated energy system of hydrogen, ammonia, and methanol using integrated deep learning. From constructing the safety domain model of the park system, to solving the model, training the integrated learning model, evaluating the model, and calculating the safety domain spatial index, the invention ultimately optimizes the hydrogen, ammonia, and methanol production and consumption scheduling strategy, effectively ensuring the safe and stable operation of the integrated energy system, reducing operating costs, improving energy utilization efficiency, and promoting the development of the park's energy system towards a safe and economical direction.

[0028] Example 2, refer to Figure 1 As one embodiment of the present invention, based on the previous embodiment, a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system is provided, comprising: In this embodiment, the above step S1, based on the production and consumption data, topology parameters, and historical operating data of the electro-hydrogen, ammonia, and methanol industry, and considering the constraints of multi-energy flow balance-related security equations and inequalities, establishes multiple park system security domain models that consider electro-hydrogen, ammonia, and methanol energy, including: The park system includes multiple energy exchange systems such as electro-hydrogen, alcohol, and ammonia. The space that satisfies all relevant energy conversion and safety constraints is called the electro-hydrogen, alcohol, and ammonia park safety domain. The compact model of the safety domain is represented as follows: In the formula: x It represents the key decision vector corresponding to the system operating point within the operating cycle, covering parameters such as unit output and load demand at each time section; z These are discrete variables of the system, including the unit operating status at each time point; y It is a vector composed of the other remaining variables of the system; f ( x , y ) represents the power flow equation constraint; g ( x , y , z )≤0 is the mathematical expression of the relevant inequality constraints; h ( x , p If )≤0, then the corresponding set of security constraints is; R represents the set of real numbers, and Z represents the set of integers.

[0029] The various constraints in the aforementioned compact model are specifically constituted by the physical models and coupling relationships of the various energy subsystems within the park, including: 1) Hydrogen production and storage Hydrogen energy systems are a key link connecting electricity and chemical synthesis. The hydrogen energy system model mainly includes three parts: hydrogen production by electrolysis, hydrogen storage, and transportation.

[0030] First, the power balance and hydrogen production constraints of the water electrolysis hydrogen production process are as follows: In the formula, represent t Always on the way to the hydrogen production station g Input electrical power, refer to t Hydrogen production station g The power consumed by auxiliary equipment; It is the conversion coefficient between power consumption and hydrogen production flow rate; Indicates hydrogen production flow rate; and These correspond to the lower and upper limits of the input power of the electrolytic cell, respectively. This represents the amount of hydrogen stored in the buffer tank. This refers to the hydrogen flow rate at the inlet of the buffer tank. Δ represents the hydrogen flow rate at the outlet of the buffer tank. t This represents the time interval during the hydrogen storage process. and These are the lower and upper limits of the buffer tank's storage capacity, respectively. This represents the power consumption of the compressor. The molar mass of hydrogen gas. R It is the ideal gas constant. T This represents the operating temperature of the compressor. p 1 and p 2 corresponds to the pressure of hydrogen gas before compression and the pressure after compression, respectively. or compress This indicates the compressor's operating efficiency.

[0031] Hydrogen produced by water electrolysis, after compression, mainly flows in three directions: one part is directly transported to the chemical synthesis stage, one part is directly stored in hydrogen storage tanks, and the other part is directly loaded for transportation. The mass balance relationship is as follows: In the formula, This represents the outlet hydrogen flow rate of the compressor. This refers to the flow rate of hydrogen directly loaded for transport. This indicates the hydrogen flow rate used for hydrogen storage. This is the hydrogen flow rate used for chemical synthesis.

[0032] Secondly, for hydrogen storage tanks used for medium- and long-term hydrogen storage, the dynamic processes and capacity constraints are described as follows: In the formula, C HST The capacity parameter representing the hydrogen storage tank; This refers to the energy value stored in the hydrogen storage tank; and These represent the hydrogen storage flow rate and hydrogen release flow rate of the hydrogen storage tank, respectively; This is the maximum capacity coefficient of the hydrogen storage tank; and These represent the upper limits of the hydrogen storage flow rate and the hydrogen release flow rate of the hydrogen storage tank, respectively.

[0033] Finally, the inter-park transportation of hydrogen is accomplished through conveyor vehicles, and its dynamic transportation status is described using a position-state matrix: The corresponding transportation logic and hydrogen charging / discharging constraints are as follows: In the formula, U c Representative conveyor c Position state matrix; u c,g,t To characterize the transport tool c The 0-1 variable of position status, when the conveyor... c Located at the hydrogen production station g When the hydrogen production station is in a certain state, its value is 1; otherwise, it is 0. This embodiment only includes hydrogen production stations in three park systems. g The value range is 1, 2, 3; TD refers to the operating cycle; G This refers to the total number of hydrogen production stations; Oh HT Represents a collection of transport tools. Oh HGS ={HGS P2A HGS P2MeOH} is a collection of hydrogen production stations, including HGS P2A HGS P2MeOH These correspond to the hydrogen production stations for the ammonia synthesis system and the methanol synthesis system, respectively. Indicates conveying tools c Hydrogen charging flow rate.

[0034] The change in the hydrogen storage capacity of the transport vehicle itself must satisfy the following dynamic and inequality constraints: In the formula, Q c,t This represents the hydrogen storage capacity of the transportation vehicle. This refers to the hydrogen emission rate of the transport vehicle; and These represent the lower and upper limits of the hydrogen storage capacity of the transportation tool c, respectively. This is a state variable for the transport vehicle. A value of 1 indicates that hydrogen is being charged, while a value of 0 indicates the opposite.

[0035] 2) Ammonia synthesis manufacturing model The ammonia synthesis subsystem utilizes a catalytic reaction between hydrogen and nitrogen under high temperature and pressure. The ammonia synthesis manufacturing model first includes calculations of the power consumption of nitrogen compression: Its core chemical reaction is: Considering the large inertia of the ammonia synthesis reactor, its ammonia production flow dynamic process is described using a first-order inertial element: Meanwhile, the ammonia synthesis process also needs to meet the following operational constraints: Furthermore, the relationship between ammonia production and reactant input follows the laws of conservation of mass and stoichiometry: In the formula, This represents the power consumption of the compressor. The molar mass of nitrogen gas. R Let be the ideal gas constant. T This indicates the operating temperature of the compressor; p 1 and p 3 represents the pressure of nitrogen before and after compression, respectively. This represents the operating efficiency of the compressor; the scheduling cycle of the ammonia synthesis reactor is set to ΔT. AS =8h; where h This is used as an indicator for the adjustment cycle of the ammonia synthesis reactor, with a value range of 1, 2, and 3. The ammonia synthesis reactor is represented in the first stage. h The initial ammonia production flow rate for each adjustment cycle, Refers to the first h Flow adjustment amount within one adjustment cycle express ammonia production flow rate at any given time. This is the upper limit of the adjustment amount; Represents the rate of ammonia synthesis. and These correspond to the lower and upper limits of the climbing rate, respectively. and These represent the lower and upper limits of ammonia yield, respectively.

[0036] 3) Manufacturing model for methanol synthesis Alongside ammonia synthesis, the methanol synthesis system forms a crucial link in the "electro-hydrogen-carbon-alcohol" value chain. This system captures carbon dioxide from coal-fired power plants and catalytically synthesizes it with hydrogen. The manufacturing model for methanol synthesis first describes the relationship between energy consumption and capture rate in the carbon capture stage: The carbon emissions from coal-fired power plants used for methanol synthesis are determined by their power generation capacity. The captured carbon dioxide needs to be compressed to the synthesis pressure, and the electrical power consumed in this process is: Its core chemical reaction is: The yield of the methanol synthesis reactor must meet upper and lower limits: Ultimately, methanol production is determined by the amount of hydrogen input and the amount of carbon dioxide captured, following a stoichiometric and mass balance relationship: In the formula, This represents the energy consumption of the carbon capture device. This refers to the amount of carbon dioxide emitted by coal-fired power plants used for methanol synthesis. λ represents the carbon capture rate. CCS The energy consumption coefficient for capturing a unit mass of carbon dioxide. Carbon capture amount; N M Represents the total number of thermal power units. c n Carbon emission factor; The power generation capacity of the nth thermal power unit used for methanol synthesis; represent P 2 MeOH The electrical power consumed by the system in compressing carbon dioxide. p 1 and p 4 represents the pressure of carbon dioxide before and after compression, respectively. The efficiency of the carbon dioxide compression process; Represents methanol production. and These correspond to the lower and upper limits of methanol yield, respectively. This represents the conversion coefficient during the methanol synthesis process.

[0037] 4) Power constraints All the energy conversion processes mentioned above, including hydrogen production from electricity, hydrogen storage and transportation, ammonia synthesis, and methanol synthesis, are deeply coupled with the power subsystem. Their operational status directly depends on the power grid's supply capacity and security level. Therefore, the power subsystem must satisfy the following conventional constraints and balance relationships, which together constitute the power boundary of the security domain model: ①Constraints on thermal power output: In the formula, and thermal power units k The lower and upper limits of output.

[0038] ②Ramp-up constraints for thermal power plants: In the formula, and These represent the maximum values ​​for the thermal power unit's upward and downward climbing speeds, respectively.

[0039] ③ Power balance constraints: ④ Constraints on new energy output: In the formula, / and / Wind and solar turbines w / s The lower and upper limits of output.

[0040] ⑤ Carbon emission constraints: In the formula, This is the daily carbon emission limit.

[0041] ⑥ Current constraints: To ensure the safe and stable operation of the system, the power flow must meet the power flow safety constraints, as shown below: In the formula, , They are nodes i , j The square of the voltage; Pij , Q ij Branch roads ij Active and reactive power; P i , Q i They are nodes i , j The net inflow of active and reactive power; r ij , x ij Branch roads ij Resistance and reactance; For the line ij The square of the current; , These are the squares of the minimum and maximum node voltages, respectively. V 0 represents the square of the voltage at the balancing node; S ij,max branch road ij The maximum transmission capacity of the line.

[0042] The park system that satisfies all the above equations and inequalities is the park's safe operation domain. Oh , Oh i Represented as the first i This embodiment considers three campus systems as a security domain for a comprehensive campus system.

[0043] In another possible implementation, when establishing multiple safety domain models for the park system that consider the energy of electro-hydrogen, ammonia, and alcohol, it is also possible to dynamically update the model parameters by introducing real-time monitoring data. Sensors are used to collect data on the production, consumption, and storage of electro-hydrogen, ammonia, and alcohol in real time, and the constraints and parameters in the model are adjusted in a timely manner based on this data to make the model more in line with the actual operating conditions.

[0044] In another possible implementation, when establishing multiple park system security domain models that consider the energy of hydrogen, ammonia, and methanol, it is also possible to combine the energy characteristics of different seasons and time periods, analyze the demand and supply patterns of hydrogen, ammonia, and methanol in different seasons and time periods, and establish corresponding sub-models for these patterns to improve the relevance and accuracy of the models.

[0045] In another possible implementation, the physical characteristics and operational constraints of the coupled power subsystem, hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem can also be implemented by establishing a unified physical model, integrating the physical characteristics and operational constraints of the four subsystems into one model, and achieving the coupling of each subsystem through solving the model.

[0046] In another possible implementation, the physical characteristics and operational constraints of the coupled power subsystem, hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem can also be implemented through a data interaction interface. A data interaction interface is developed for each subsystem, and each subsystem exchanges data in real time through the interface and adjusts its own operating parameters according to the operating status of other subsystems, thereby realizing the coupling between subsystems.

[0047] In this embodiment, step S2 above, based on the campus system security domain model, coupled with multi-energy constraint generation, and calculating the final vertex set based on the improved vertex search algorithm, solves the campus system security domain model, including: Define decision variables X(X1, X2, ..., X...) in a high-dimensional space containing flexible resource and electricity decision variables related to hydrogen, ammonia, and alcohols. q Multi-dimensional integrated park security domain Oh The initial vertex calculation includes: defining Oh Boundary normal vector b=[ n 1, n 2... n q ], where 1, 2, ... q for Oh The number of vertices is an unknown quantity, but q At least one exists on the 2D plane q A closed region can only be formed by adding one non-overlapping point, therefore the initial value is taken as... q+ 1 vertex, q+ The initial normal to the boundary of a closed region consisting of one vertex (randomly selected) is b = {[1,0]1,[-1,0]2,...,[0,1]} q+1 Substituting b into the following formula and calculating, we can obtain the result. Oh The initial vertices (i.e., the final values ​​of the coordinates of the three randomly selected vertices): The obtained initial vertex coordinates are denoted as v 0. Transform the coordinates of the initial vertex: In the formula, v 1 represents the transformed vertex coordinates; N 0 represents the number of vertices; For the first i The coordinates of each vertex.

[0048] Taking a two-dimensional safety surface as an example, let the coordinates of any two initial vertices be ( m 1, n 1), ( m 2, n 2), then the straight line formed by these two verticesj The equation of the normal is as follows: The coordinates of the fourth vertex can then be obtained, expressed as: The other two vertices are calculated. After one iteration, a polygon is obtained, which is the safe boundary in the two-dimensional plane. The termination condition is as follows: In the formula, Indicates the first j The second norm of the normal vector of a straight line; ε is a set value, the smaller the value, the more vertices are obtained, i.e., the multidimensional constraint safety region. Oh The more complex the boundary, the higher the accuracy. Specifically, if the second norm value of the normal vector is less than a predetermined value in any given calculation, the search for the multidimensional constrained safe region is terminated. Oh Boundary vertices are used as the obtained vertices for multidimensional constraint safety regions. Oh The boundary is ultimately obtained as the security domain of the park system. Oh .

[0049] In another possible implementation, defining high-dimensional decision variables that include hydrogen-ammonia-alcohol flexible resource and power decision variables can also be implemented through cluster analysis. Cluster analysis is performed on hydrogen-ammonia-alcohol flexible resource and power decision variables to group similar variables into one class. Then, each class of variables is defined and processed to reduce the complexity of the high-dimensional space.

[0050] In another possible implementation, defining high-dimensional decision variables that include hydrogen-ammonia-alcohol flexible resource and power decision variables can also be done through principal component analysis. Principal component analysis is used to extract the main components in the hydrogen-ammonia-alcohol flexible resource and power decision variables, and the high-dimensional variables are transformed into a few principal component variables for definition and analysis.

[0051] In this embodiment, step S3 above, based on the solved security domain of the park system, constructs a feature set with a mapping relationship between the green park system operating status characteristics and security domain boundary information, and designs and trains an LSTM-BiLSTM-BiGRU ensemble learning model based on the feature set, including: Security Domain Oh As output labels, the decision variables X(X1, X2, ..., X...) in the high-dimensional space containing hydrogen-ammonia-alcohol flexible resource and electricity decision variables are... q () as input features. Preprocess the input feature data: in, Xstd This represents the normalized data values, i.e., the constructed feature set; x This represents unnormalized data values. x max and x min These represent the maximum and minimum values ​​in the unnormalized data, respectively.

[0052] The constructed feature set is divided into 7:3 parts. Xtrain std and Xtest std .

[0053] Design an LSTM-BiLSTM-BiGRU ensemble learning model, including: ①LSTM model The structure contains three gates: an input gate, a forget gate, and an output gate. These three gates control the flow of information between the tuples and the network. In the following formula... i t , o t , f t These represent the state values ​​of the input gate, output gate, and forget gate, respectively. The sigmoid layer of the forget gate determines the state to be retrieved from the old cell state. C t-1 Forgot the information, input is the input of the current layer. x t and the output of the previous layer h t-1 The cell state output at this moment is: Generating and storing the information that needs to be updated in the cell state involves two steps: i) The result of the input gate passing through the sigmoid layer. i t to update information; ii) new candidate values ​​generated by the tanh layer. C i It will be added to the cell state. C i The old cell state is multiplied by the forgotten information and new candidate information. i t * C i The result is obtained by adding them together.

[0054] The output information is determined by the output gate. First, the initial output is obtained through a sigmoid layer. Then, a tanh layer is used to scale the cell state values ​​to the range [-1, 1]. These values ​​are then multiplied pairwise with the output obtained from the sigmoid layer to obtain the final output. .

[0055] In the formula: W 1. For communication x t The weight matrix of the input gate, forget gate, output gate, and tuple input; W h For connection h t-1 The weight matrix of the input gate, forget gate, output gate, and tuple input; b σ represents the bias vectors for the input gate, forget gate, output gate, and tuple input; σ denotes the sigmoid activation function.

[0056] Using the feature data obtained in the previous step as input to the LSTM network model, the output is: In the formula, h t+1 For predicted values, h t These are historical data. x The input parameters are the result data obtained from the optimization of each integrated electrohydrogen alcohol ammonia system park.

[0057] ② Bidirectional Long Short-Term Memory Network (BiLSTM) Composed of a forward LSTM network and a backward LSTM network, this system enables bidirectional synchronous processing of sequence data, fusing the results from both directions before outputting the final product. The network employs a two-layer LSTM architecture: specifically, data is first fed into the forward LSTM layer for processing, and then the same data sequence is input into the backward LSTM layer in reverse chronological order. This design allows the model to consider the contextual information of the data simultaneously, thus capturing the features of the time series more comprehensively. The LSTM update mechanism is as follows: ③ Bidirectional Gating Unit (BiGRU) The bidirectional gated recurrent unit accepts the output data from the BiLSTM neural network as input, learns the complex correlation between vehicle state data before and after time series, and deeply mines the hidden feature information in the time series. Hidden layer state output: In the formula: GRU fwd , GRU bwd These represent the forward and reverse network structures, respectively. α t , β t They represent t time Bi GRU The corresponding forward hidden state h t and reverse hidden state h t . x t This is the input data. h t and h t They are t The state of the forward and backward GRU at any given moment.

[0058] In this embodiment, step S4 includes evaluating the trained LSTM-BiLSTM-BiGRU ensemble learning model, obtaining the optimal model, and then calculating the campus security domain spatial index, including: The regression data Ω and the actual data Ω* are compared, and the mean absolute error, mean relative error, and mean squared error are used as model evaluation metrics to evaluate the trained LSTM-BiLSTM-BiGRU ensemble learning model: ① Mean Absolute Error: ② Average relative error: ③ Root mean square error: The trained model is evaluated using these metrics to obtain the optimal model.

[0059] Calculate the spatial index of the park's security domain based on the optimal model: Volume index V Ω This represents the range of low-carbon operating space; the larger the value, the greater the flexible, safe, and optimizable space, providing a higher level of assurance for safe operation. Specifically, it represents a volume density of 1. p Multiple integrals: Sensitivity S ( C ), representing V Ω For different analysis objects C The rate of change reflects the efficiency of the object under analysis in changing the benefits of the safe-economic operating environment: In this embodiment, step S5 above, based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, sets an objective function, establishes a collaborative optimization operation model for the integrated energy distribution network system, and solves the production and consumption scheduling strategy for hydrogen, ammonia, and methanol to achieve system safety and economical operation optimization, including: Based on the security domain space indicators and with the goal of safe and economical operation of the distribution network system, the objective function of the collaborative optimization operation model of the integrated energy distribution network system is set as follows: In the formula: C λ , C buy_P , C buy_H , C buy_G , C buy_up V Ω These include costs for handling uncertainties related to new energy sources, electricity purchase costs, hydrogen purchase costs, methanol and ammonia purchase costs, and compensation costs for park system support (C). up And volumetric cost; k The cost coefficient for the security domain space index.

[0060] The specific calculation is expressed as follows: In the formula, N G and N M They are respectively a combination of ammonia and methanol and gas turbines. m For the corresponding node, and These are the upper and lower control volumes of the gas source, respectively. aU G and aD G To regulate costs for the unit, w For device indexing, PM,U w,t and PM.D wt These are the up and down control parameters for the gas turbine. a UM and aD M To regulate costs; N Re and N slack They are respectively new energy power stations and root node sets. Contribute to new energy power stations and These are the root nodes of the distribution network. i The power purchased and sold by the superior power grid, a Re , a buy Re and a Sell ​​and Re represent the on-grid tariff for renewable energy power plants, the purchase and sale tariffs for the distribution network and the upstream power grid, respectively; N H For the collection of hydrogen refueling mother and daughter stations, p For hydrogen refueling mother and daughter station index, For the amount of hydrogen purchased by the hydrogen refueling stations, a H represents the unit hydrogen valence; and abuy G represents the planned purchase quantity and price of ammonia; N Park Represents a collection of systems within the park. a up G、 a up H and a up M represents the compensation price for providing electricity, hydrogen, and ammonia / methanol energy support to the park system. 、 and These represent the electricity, hydrogen, and gas support quantities for the park's system, respectively.

[0061] The constraints of the integrated energy distribution network system collaborative optimization operation model are basically the same as the power constraints in S1. The following are additional constraints on the distribution network-industrial park interconnection coupling: In the formula, Pgrid g,t This is expressed as the switching power / hydrogen / ammonia / methanol quantity of the distribution network-park interconnection line; Pexchange l,max For connecting lines l Maximum allowed transfer amount; N l For connecting to the park system; Pgrid g,max , Pgrid g,max These are the upper and lower limits of the exchange power between the park and the distribution network, respectively. E gridESS t、 E gridESS, ch t、 E gridESS and dis t represent the energy storage capacity, charging power, and discharging power of the distribution network-level energy storage, respectively. or gridESS, ch t、 or gridESS and dis t represent the charging and discharging efficiencies of distribution network-level energy storage, respectively.

[0062] Solve the collaborative optimization operation model of the integrated energy distribution network system to obtain the production and consumption scheduling strategy of hydrogen, ammonia and alcohol, and achieve the optimization of safe and economical system operation.

[0063] It should be noted that the inputs to the integrated energy distribution network system collaborative optimization operation model include: Safety domain space indicators, such as the previously calculated volume coefficient and sensitivity, reflect the scope and characteristics of the park's safety domain, providing the model with information about the system's safe operating space.

[0064] Energy market data: This includes price information for various energy sources (electricity, hydrogen, ammonia, and alcohols), such as the cost of purchasing electricity, hydrogen, and alcohols / ammonia, which are used to calculate the energy purchase cost in the objective function.

[0065] System operating parameters: such as the power output of new energy power plants, the control of gas turbines, the transmission capacity of distribution network-industrial park interconnection lines, and the charging and discharging efficiency of distribution network-level energy storage, etc. These parameters are used to construct the constraints of the model and limit the operating range of the system.

[0066] The correlation coefficients between the objective function and the constraints: such as the weighting coefficients of various costs in the objective function, the cost coefficients of the safety domain space index, and the upper and lower limit parameters in the constraints, these coefficients determine the specific form of the objective function and the constraints.

[0067] The output includes: Production and consumption scheduling strategy for electricity, hydrogen, ammonia, and alcohol: Clearly define the production and consumption arrangements for electricity, hydrogen, ammonia, and alcohol in different time periods, including the output plan of production equipment, energy purchase and sales decisions, etc., in order to achieve safe and economical operation of the system.

[0068] The optimal objective function value, i.e. the minimized total cost, reflects the economic cost of system operation under the scheduling strategy and can be used to evaluate the merits of the scheduling strategy.

[0069] Relevant status information of the model solution, such as whether the solution has converged, the number of iterations, and the computation time, can help evaluate the efficiency and reliability of the solution process.

[0070] In another possible implementation, the optimization of safe and economical system operation can also be achieved by dynamically adjusting the scheduling strategy. Based on factors such as real-time energy market prices and equipment operating status, the production and consumption scheduling strategy of hydrogen, ammonia, and methanol can be adjusted in real time to achieve economical operation optimization while ensuring system safety.

[0071] In another possible implementation, the optimization of safe and economical system operation can also be achieved by introducing intelligent algorithms for multi-objective optimization. By using intelligent algorithms such as genetic algorithms and particle swarm optimization, multiple objectives such as system safety, economy, and environmental protection can be considered simultaneously to find the optimal scheduling strategy.

[0072] Example 3 illustrates a schematic scheme for a method of safety domain mapping and optimization for an integrated deep learning-based electro-hydrogen-amine-ethanol energy system. It should be noted that the technical solution of the integrated deep learning-based safety domain mapping and optimization system for an integrated deep learning-based electro-hydrogen-amine-ethanol energy system belongs to the same concept as the aforementioned method. Details not described in detail in the technical solution of the integrated deep learning-based safety domain mapping and optimization system for an integrated deep learning-based electro-hydrogen-amine-ethanol energy system can be found in the description of the aforementioned method.

[0073] This embodiment also provides a safety domain mapping and optimization system for an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, including: The park security domain modeling module is used to establish multiple park system security domain models that consider the energy of electric hydrogen, ammonia, and alcohol based on the production and consumption data, topology parameters, and historical operation data of the electric hydrogen, ammonia, and alcohol industry, and considering the constraints of multi-energy flow balance related security equations and inequalities. The model solving module is used to solve the campus system security domain model by coupling multi-energy constraint generation based on the campus system security domain model and calculating the final vertex set based on the improved vertex search algorithm. The feature learning model training module is used to construct a feature set based on the obtained security domain of the park system, with the green park system operation status features and security domain boundary information as the mapping relationship, and to design and train an LSTM-BiLSTM-BiGRU ensemble learning model based on the feature set. The index calculation module is used to evaluate the trained LSTM-BiLSTM-BiGRU ensemble learning model, and calculate the spatial index of the park security domain after obtaining the optimal model. The production and consumption scheduling optimization module is used to set an objective function based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, establish a collaborative optimization operation model of the integrated energy distribution network system, solve for the production and consumption scheduling strategy of hydrogen, ammonia and alcohol, and realize the optimization of safe and economical operation of the system.

[0074] This embodiment also provides an electronic device applicable to a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a method for mapping and optimizing the safety domain of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, as proposed in the above embodiments.

[0075] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a method for security domain mapping and optimization of an integrated deep learning-based electrohydrogen-ammonia-ethanol energy system as proposed in the above embodiment.

[0076] The storage medium proposed in this embodiment belongs to the same inventive concept as the integrated deep learning-based security domain mapping and optimization method for an electrohydrogen-ammonia-ethanol integrated energy system proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0077] Example 4, refer to Figure 2-Figure 6As an embodiment of the present invention, a method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system is provided. To verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0078] This embodiment demonstrates the construction and solution process of the safety domain model for an industrial park using the described technology. Taking a typical industrial park containing a photovoltaic power station, electrolyzers, hydrogen storage tanks, and ammonia synthesis equipment as an example, the implementation method is as follows: First, collect equipment parameters such as the rated power of the electrolyzers, the capacity of the hydrogen storage tanks, and the maximum yield of the ammonia synthesis reactor, as well as historical electricity, hydrogen, and ammonia load data for the park. Then, establish a multi-energy flow safety domain model that includes power flow, hydrogen production-storage-transport balance, ammonia synthesis reaction kinetics, and ramp-up constraints. Finally, based on an improved vertex search algorithm, solve for the boundary vertex set of the safety domain model in a high-dimensional decision space. The structure of a comprehensive industrial park system-distribution network system in a certain region is as follows: Figure 2 As shown in the diagram, the integrated LSTM-BiLSTM-BiGRU method structure is as follows: Figure 3 As shown, the power output and load demand curves of a certain region are respectively shown in the figure. Figure 4 , Figure 5 As shown, based on the solution considering all relevant constraints, a 3D visualized security domain diagram of load-new energy-power flow is presented. Figure 6 As shown, the security domain space of the campus system proposed in this invention can be obtained, which can provide a security support foundation for the distribution network system.

[0079] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, characterized in that, include: Based on the production and consumption data, topology parameters, and historical operation data of the electro-hydrogen, ammonia, and alcohol industry, and considering the constraints of multi-energy flow balance related safety equations and inequalities, multiple park system safety domain models considering electro-hydrogen, ammonia, and alcohol energy are established. Based on the campus system security domain model, multi-energy constraint generation is coupled, and the final vertex set is calculated based on the improved vertex search algorithm to solve the campus system security domain model; Based on the obtained security domain of the park system, a feature set is constructed with the green park system operation status characteristics and security domain boundary information as the mapping relationship. Based on the feature set, an LSTM-BiLSTM-BiGRU ensemble learning model is designed and trained. The trained LSTM-BiLSTM-BiGRU ensemble learning model is evaluated, and the optimal model is obtained. The spatial index of the campus security domain is then calculated. Based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, an objective function is set, and a collaborative optimization operation model of the integrated energy distribution network system is established. The production and consumption scheduling strategy of hydrogen, ammonia, and methanol is obtained by solving the model, thereby achieving optimization of the safe and economical operation of the system.

2. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 1, characterized in that, Based on the production and consumption data, topology parameters, and historical operating data of the electro-hydrogen, ammonia, and methanol industry, and considering the constraints of multi-energy flow balance-related safety equations and inequalities, multiple safety domain models for the industrial park system considering electro-hydrogen, ammonia, and methanol energy are established, including: Construct a safe operation domain model for a park system that integrates the conversion and flow of multiple energy sources, including electricity, hydrogen, ammonia, and methanol, including the physical characteristics and operational constraints of the coupled power subsystem, hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem; The hydrogen energy subsystem model integrates the electrolytic hydrogen production, hydrogen storage, and dynamic transportation processes, and uses a position state matrix to describe the scheduling logic of hydrogen transportation vehicles between hydrogen production stations and chemical synthesis facilities in order to determine the spatiotemporal allocation of hydrogen. The ammonia synthesis subsystem model, based on hydrogen provided by the hydrogen energy subsystem and combined with nitrogen input, determines ammonia production through a chemical reaction process with dynamic adjustment characteristics.

3. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 2, characterized in that, The establishment of multiple park system security domain models considering the energy of electric hydrogen, ammonia, and methanol, based on the industrial production and consumption data, topology parameters, and historical operating data of the electric hydrogen, ammonia, and methanol industry, and considering the constraints of multi-energy flow balance correlation security equations and inequalities, also includes: The methanol synthesis subsystem model is based on hydrogen provided by the hydrogen energy subsystem and carbon dioxide captured and compressed from a coal-fired power plant, and determines the methanol yield through a catalytic synthesis reaction. The power subsystem model provides power to the power consumption units in the hydrogen energy subsystem, ammonia synthesis subsystem, and methanol synthesis subsystem, and meets grid operation constraints including unit output, ramping, carbon emissions, and power flow safety.

4. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 3, characterized in that, The process of solving the campus system security domain model by coupling multi-energy constraint generation with an improved vertex search algorithm to calculate the final vertex set includes: Define high-dimensional decision variables that include flexible resource and power decision variables of hydrogen-ammonia-alcohol; calculate the initial vertices of the safety domain of the multidimensional integrated park, define the boundary normal vector, take random vertices to form a closed region and determine the initial normal, and calculate the initial vertices.

5. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 4, characterized in that, The method of solving the campus system security domain model by coupling multi-energy constraint generation and calculating the final vertex set based on an improved vertex search algorithm also includes: The initial vertex coordinates are transformed, and new vertex coordinates are calculated based on the transformed vertices. The calculation is repeated to obtain the safety boundary. A solution termination condition is set. When the second norm of the normal vector of a certain line is less than the set value, the search for boundary vertices is stopped. The obtained vertices determine the boundary of the multidimensional constrained safety domain, and the safety domain of the park system is obtained.

6. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 5, characterized in that, Based on the solved security domain of the park system, a feature set is constructed with the green park system operation status characteristics and security domain boundary information as the mapping relationship. The LSTM-BiLSTM-BiGRU ensemble learning model is designed and trained based on this feature set, including: Using the park system security domain as the output label and high-dimensional space decision variables containing hydrogen-ammonia-alcohol flexible resource and power decision variables as input features, the input feature data is preprocessed to construct a feature set; the constructed feature set is divided into a training set and a test set; an LSTM-BiLSTM-BiGRU ensemble learning model is designed, which is composed of LSTM, BiLSTM and BiGRU modules in sequence; the designed ensemble learning model is trained using the training set.

7. The method for safety domain mapping and optimization of an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system as described in claim 6, characterized in that, The process includes evaluating the trained LSTM-BiLSTM-BiGRU ensemble learning model, obtaining the optimal model, and then calculating the spatial indices of the campus security domain, including: The regression data of the ensemble learning model is compared with the actual data. The mean absolute error, mean relative error, and mean squared error are selected as model evaluation indicators. The difference between the regression data and the actual data is measured by the model evaluation indicators. The trained ensemble learning model is evaluated, and the optimal model is selected from the evaluation results. The spatial index of the park security domain is calculated based on the obtained optimal model.

8. A safety domain mapping and optimization system for an integrated deep learning-based electro-hydrogen-ammonia-ethanol energy system, employing the method described in any one of claims 1 to 7, characterized in that, include: The park security domain modeling module is used to establish multiple park system security domain models that consider the energy of electric hydrogen, ammonia, and alcohol based on the production and consumption data, topology parameters, and historical operation data of the electric hydrogen, ammonia, and alcohol industry, and considering the constraints of multi-energy flow balance related security equations and inequalities. The model solving module is used to solve the campus system security domain model by coupling multi-energy constraint generation based on the campus system security domain model and calculating the final vertex set based on the improved vertex search algorithm. The feature learning model training module is used to construct a feature set based on the obtained security domain of the park system, with the green park system operation status features and security domain boundary information as the mapping relationship, and to design and train an LSTM-BiLSTM-BiGRU ensemble learning model based on the feature set. The index calculation module is used to evaluate the trained LSTM-BiLSTM-BiGRU ensemble learning model, and calculate the spatial index of the park security domain after obtaining the optimal model. The production and consumption scheduling optimization module is used to set an objective function based on the safety domain space index and with the goal of safe and economical operation of the distribution network system, establish a collaborative optimization operation model of the integrated energy distribution network system, solve for the production and consumption scheduling strategy of hydrogen, ammonia and alcohol, and realize the optimization of safe and economical operation of the system.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images