A method, system and medium for improving post-disaster resilience of an integrated energy system of electricity and hydrogen

Abstract

The application discloses a kind of electric hydrogen comprehensive energy system post-disaster resilience promotion method, system and medium, including establishing the post-disaster power system collaborative recovery optimization model for hydrogen source guarantee, integrate mobile energy storage space-time scheduling and distributed power output constraint, the supply-demand correlation mechanism of electric-hydrogen conversion link is quantified;Hydrogen energy transport path optimization model based on centralized optimization is constructed, the global optimal configuration of hydrogen energy transport resources is realized by comprehensively vehicle scheduling cost, transport cost, hydrogenation station waiting time cost and hydrogen supply shortage penalty cost;For information localization and multi-transport unit coordination problem, multi-agent scheduling model is constructed based on distributed partially observable Markov decision process, and QMIX algorithm is used for strategy learning, and distributed hydrogen transport strategy is obtained.The method of the present application improves the power supply guarantee capability and hydrogen supply reliability of electric-hydrogen comprehensive energy system under extreme disaster, and enhances the overall resilience of the system.

Description

Technical Field

[0001] This invention relates to the fields of integrated energy system operation control, power system emergency recovery and intelligent dispatch technology, and in particular to a method, system and medium for improving the post-disaster resilience of an integrated electric-hydrogen energy system. Background Technology

[0002] With the intensification of global climate change and the increasing frequency of extreme weather events, large-scale energy system outages are becoming more frequent, severely impacting socio-economic development and people's lives. Against the backdrop of energy transition, hydrogen energy, as a clean and efficient secondary energy source, and its deep integration with the power system to form an integrated electric-hydrogen energy system, have become an important technological pathway to enhance the resilience of energy systems and ensure energy supply during extreme events.

[0003] However, the integrated electric-hydrogen energy system faces a dual challenge of coordination under extreme disasters: on the one hand, the entire chain of hydrogen production, storage, and use relies on a stable power supply, and a power grid failure will directly lead to the shutdown of hydrogen production stations, triggering a chain of disruptions in the hydrogen energy supply chain; on the other hand, hydrogen transportation is highly dependent on the transportation network, and factors such as road network damage, reduced traffic capacity, and poor information transmission after a disaster make it difficult for traditional dispatching methods to quickly respond to dynamically changing energy demand.

[0004] In existing technologies, research on post-disaster recovery of integrated electric-hydrogen energy systems largely focuses on the single system level, either considering only the load recovery of the power system or optimizing the scheduling strategy of the hydrogen energy supply chain separately, failing to fully consider the strong coupling relationship between the power and hydrogen energy systems. Traditional centralized optimization methods rely on the assumption of complete global information and a static environment, which leads to low computational efficiency and poor adaptability in scenarios with localized post-disaster information, complex multi-transport unit coordination, and dynamic demand fluctuations. Single-agent reinforcement learning methods face the challenge of environmental non-stationarity in multi-vehicle collaborative scheduling problems, easily getting trapped in local optima and struggling to achieve global collaborative optimization.

[0005] In view of this, there is an urgent need for a technical solution to enhance the post-disaster resilience of an integrated electric-hydrogen energy system that can achieve deep synergy between power restoration and hydrogen energy dispatch, while taking into account both global optimization and real-time responsiveness. Summary of the Invention

[0006] To address the aforementioned issues, this invention provides a method, system, and medium for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system. The aim is to solve the problems of insufficient synergy and poor dynamic adaptability of traditional technologies in extreme disaster scenarios, thereby achieving efficient and reliable post-disaster recovery of the integrated electric-hydrogen energy system.

[0007] In a first aspect, this invention provides a method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system, comprising the following steps: Based on electricity load data and hydrogen energy supply data, a collaborative recovery optimization model for electricity-hydrogen coupling is constructed with the objective function of minimizing the sum of electricity load reduction costs and hydrogen energy supply interruption penalty costs. Integrating spatiotemporal dynamic scheduling constraints for mobile energy storage, charging and discharging constraints for stationary energy storage, load shedding constraints, distributed power output constraints, and distribution network safety operation constraints, this approach quantifies electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; Based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, a hydrogen energy transportation route optimization model is constructed with the objective function of minimizing the sum of hydrogen long-tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen energy transportation resources. The hydrogen tube trailer is modeled as an independent intelligent agent, and a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning is constructed. A scheduling model is established based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost and insufficient hydrogen supply penalty cost as the global reward function. The QMIX algorithm, employing a centralized training and decentralized execution paradigm, is used for policy learning to achieve adaptive collaborative scheduling of hydrogen energy after disasters.

[0008] A further technical solution of the present invention is: the cooperative recovery optimization model of the electro-hydrogen coupling, expressed as follows: in, Represents the set of discrete time periods during disaster recovery; Represents the set of power load nodes; Represents the set of hydrogen production station nodes; Represents a node The weighting factor for electricity load reduction costs Indicates hydrogen production station The weighting coefficient for hydrogen energy reduction costs; Indicates the time period ,node Forced reduction in electricity load, Indicates the time period Hydrogen production station Hydrogen supply gap; through adjustment and The relative size of the loads determines the differentiated recovery strategy that prioritizes critical loads or hydrogen production stations.

[0009] A further technical solution of the present invention is as follows: The spatiotemporal dynamic scheduling constraints of the mobile energy storage include travel time calculation constraints, travel distance correction constraints, vehicle speed attenuation constraints, node connection uniqueness constraints, travel period continuity constraints, charging and discharging state coupling constraints, charging and discharging power constraints, and state of charge constraints, wherein: The constraint expression for calculating the travel time is: ,in, For mobile energy storage during the period From node arrive The actual migration time; This is the corrected distance between nodes; The actual vehicle speed under disaster conditions; The expression for the distance correction constraint is: ,in, For nodes and Topological distance between them; For time period node Reference vehicle speeds on surrounding roads quantify the indirect impact of reduced road traffic efficiency caused by disasters on travel distance; The expression for the vehicle speed attenuation constraint is: ,in, The ideal driving speed. To quantify the level of disasters, , Indicates no disaster. This indicates that the transportation network is completely paralyzed; The expression for the charging / discharging state coupling constraint is: ,in, For MESS at node Time period The charging icon For MESS at node Time period The discharge indicator, For the set of distribution network nodes, For MESS in the time period From node arrive The scheduling connection status; The expression for the charge state constraint is: ,in, For MESS at node Time period Real-time state of charge, For charging efficiency, For discharge efficiency, For MESS at node Time period charging power, For MESS at node Time period The discharge power.

[0010] A further technical solution of the present invention is as follows: the supply and demand correlation mechanism of the electro-hydrogen conversion process is quantified through constraints on the electro-hydrogen conversion efficiency of the electrolyzer, the total power demand of the hydrogen production station, the hydrogen energy balance constraint of the hydrogen storage tank, and the hydrogen energy gap quantification constraint, wherein: The constraint expression for the electro-hydrogen conversion efficiency of the electrolyzer is: ,in, For hydrogen production station Time period Electrolytic cell output, The electro-hydrogen conversion coefficient is determined by the efficiency of the electrolyzer. This refers to the electrical input power of the electrolytic cell. This is a temperature correction factor; The total power demand constraint expression for the hydrogen production station is: ,in, This refers to the power consumption of the hydrogen compression process. To meet the total electricity demand for hydrogen production; The hydrogen energy balance constraint expression for the hydrogen storage tank is: ,in, Indicates hydrogen production station Time period Hydrogen storage capacity, This refers to the external supply of hydrogen. The quantitative constraint expression for the hydrogen energy gap is: in, For hydrogen production station Time period To meet external hydrogen demand, the on-site storage capacity must be sufficient; otherwise, a hydrogen supply gap will occur. This represents the minimum amount of hydrogen stored.

[0011] A further technical solution of the present invention is: the hydrogen energy transportation route optimization model is specifically expressed as: Where K represents the collection of hydrogen tube trailers, For the set of hydrogen refueling station nodes, The fixed dispatch cost per vehicle. This is a flag indicating whether vehicle k has been dispatched. For the variable cost per unit distance of the vehicle, This is a flag indicating whether vehicle k has traveled from node i to node j. Let i be the shortest distance from node i to j. The unit waiting time cost at hydrogen refueling stations The waiting time for vehicle k at the hydrogen refueling station r. The penalty cost coefficient for hydrogen refueling station r. The minimum supply threshold for hydrogen refueling station r, The amount of hydrogen supplied to vehicle k at hydrogen refueling station r.

[0012] A further technical solution of the present invention includes: vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, specifically comprising: The vehicle origin constraint, which stipulates that dispatched vehicles must originate from the hydrogen production station, is expressed as follows: Where G represents the set of hydrogen production station nodes. This is a flag indicating whether vehicle k has traveled from node g to node j; The vehicle destination constraint is used to unify the destination of all dispatched vehicles as virtual nodes, and its expression is: Where V is the set of virtual nodes, and the distance between the virtual nodes and all hydrogen refueling stations is set to 0. This is a flag indicating whether vehicle k has traveled from node i to node v; Vehicle path continuity constraints are used to ensure the integrity of vehicle travel paths, and are expressed as follows: The "Disaster-damaged road section inaccessibility" restriction is used to exclude road sections damaged after a disaster, and its expression is: in, This refers to a collection of road sections that are impassable due to disasters. The initial hydrogen transport capacity constraint, used to specify that dispatched vehicles must be fully loaded when departing from the hydrogen production station, is expressed as: in, This refers to the initial hydrogen load carried by the vehicle when it departs from the hydrogen production station. This represents the vehicle's maximum hydrogen carrying capacity. The minimum supply threshold constraint, used to ensure the basic operational needs of hydrogen refueling stations, is expressed as follows: .

[0013] A further technical solution of the present invention is: the scheduling model based on the distributed partially observable Markov decision process uses octuples. Define, where: Collection of intelligent agents This represents the number of agents participating in the task, with a value ranging from [value range missing]. Define the scale of multi-agent collaboration; Global state space Includes road network status Vehicle status Demand Status ,Right now ; Joint Action Space The Cartesian product of all agent actions, and the action of a single agent. This indicates the target hydrogen refueling station selected at the current moment or the waiting action; State transition function Describes the current environmental state. Under these circumstances, the intelligent agents execute joint actions. Afterwards, the environment shifted to a new state. The probability distribution; Global reward function Generated by environmental feedback, used to measure the agent's state in the environment. Execute joint actions The benefits are shared by all agents through the same global reward function. To optimize the overall objective; Local observation space With observation function Define a single intelligent agent Based on the observation function Perceptible local observation information The agent's observable range is 80% of the map's diagonal distance; Strategy This represents the action distribution of a single agent based on its observation history. Represents intelligent agents Action-observation history is the sequence of all observations and actions performed by the agent from the initial moment to the current moment. Discount factor The weights used to balance current and future rewards.

[0014] A further technical solution of the present invention is: the monotonic value decomposition mechanism of the QMIX algorithm guarantees: This makes the globally optimal joint action equivalent to the combination of the optimal actions of each agent. The algorithm adopts a centralized training and decentralized execution paradigm. During the training phase, global state information is used to dynamically generate non-negative weights and biases of the hybrid network through a hypernetwork, non-linearly fusing the Q-values ​​of individual agents into a global Q-value. During the execution phase, each agent makes independent decisions based only on its own local observations. Represents a set of intelligent agents. The global Q-value is represented by u, which represents the combined action formed by each agent performing its chosen action. Represents the global state. For the state of the kth agent, For the action targeting the k-th agent.

[0015] According to a second aspect of the present disclosure, a disaster resilience enhancement system for an integrated electric-hydrogen energy system is provided, comprising: The collaborative recovery model construction module is used to construct an electric-hydrogen coupled collaborative recovery optimization model based on electricity load data and hydrogen energy supply data, with the objective function being to minimize the sum of electricity load reduction costs and hydrogen energy supply interruption penalty costs. The constraint integration and solution module integrates spatiotemporal dynamic scheduling constraints for mobile energy storage, charging and discharging constraints for stationary energy storage, load shedding constraints, distributed power generation output constraints, and distribution network safety operation constraints to quantify electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; The centralized scheduling module is used to construct a hydrogen transportation route optimization model based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, with the objective function being to minimize the sum of hydrogen tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen transportation resources. The distributed strategy modeling module is used to model the hydrogen tube trailer as an independent intelligent agent, construct a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning, and establish a scheduling model based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost and insufficient hydrogen supply penalty cost as the global reward function. The algorithm training and execution module is used to learn policies using the QMIX algorithm under the paradigm of centralized training and decentralized execution, so as to realize the adaptive and collaborative scheduling of hydrogen energy after disasters.

[0016] According to a third aspect of the present disclosure, a non-transitory computer-readable storage medium is provided, on which computer instructions are stored, which, when executed by a processor, implement the steps of the post-disaster resilience enhancement method for an integrated electric-hydrogen energy system as described above.

[0017] The beneficial effects of this invention are as follows: On the one hand, this invention constructs a collaborative recovery optimization model for electricity-hydrogen coupling, incorporating the power supply guarantee of hydrogen production stations into the power recovery objective function. By flexibly adjusting the recovery priority under different scenarios through differentiated weight coefficients, it achieves deep collaborative optimization of the electricity and hydrogen energy systems. On the other hand, this invention adopts a hierarchical hydrogen energy scheduling architecture. In scenarios with complete information, it obtains the global optimal solution through centralized optimization. In complex scenarios with localized and dynamically changing information, it achieves rapid adaptive decision-making through multi-agent reinforcement learning, balancing the global optimality and real-time responsiveness of scheduling. Simultaneously, this invention employs the QMIX algorithm for multi-agent policy learning, ensuring the consistency between individual optimality and global optimality through a monotonic value decomposition mechanism. This effectively overcomes the non-stationarity problem of traditional single-agent reinforcement learning in multi-vehicle collaborative scenarios, improving the collaborative efficiency of multiple transportation units. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0019] Figure 1 This is an overall flowchart of the method for improving the post-disaster resilience of the integrated electric-hydrogen energy system of the present invention; Figure 2 This is an architecture diagram of the electro-hydrogen coupling synergistic recovery optimization model of the present invention; Figure 3 This is a flowchart of the distributed hydrogen transportation strategy solution based on the QMIX algorithm of this invention; Figure 4 This is a schematic diagram of the post-disaster resilience enhancement system of the integrated electric-hydrogen energy system of the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present invention are shown in the drawings, not the entire structure.

[0021] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of these steps can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the steps can be rearranged. The process can be terminated when its operation is complete, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0022] This invention achieves efficient recovery of integrated electric-hydrogen energy systems under extreme disasters by constructing an electric-hydrogen coupled collaborative recovery model and a hierarchical hydrogen energy dispatch architecture. The invention utilizes mixed-integer second-order cone programming for power-side collaborative recovery optimization, mixed-integer linear programming for centralized hydrogen transport path optimization, and QMIX-based multi-agent deep reinforcement learning for distributed hydrogen transport strategy generation. This invention effectively solves the problems of insufficient coordination and poor dynamic adaptability of traditional methods in complex scenarios with localized post-disaster information, multi-resource coupling, and high real-time requirements, significantly improving the post-disaster recovery capability and energy supply reliability of integrated electric-hydrogen energy systems.

[0023] like Figure 1 As shown, the present invention employs a method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system based on deep reinforcement learning, comprising the following steps: S1. Based on electricity load data and hydrogen energy supply data, a collaborative recovery optimization model for electricity-hydrogen coupling is constructed with the objective function of minimizing the sum of electricity load reduction costs and hydrogen energy supply interruption penalty costs. S2. Integrates spatiotemporal dynamic scheduling constraints for mobile energy storage, charging and discharging constraints for fixed energy storage, load shedding constraints, distributed power generation output constraints, and distribution network safety operation constraints to quantify electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; S3. Based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, a hydrogen energy transportation route optimization model is constructed with the objective function of minimizing the sum of hydrogen long-tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen energy transportation resources. S4. Model the hydrogen long-tube trailer as an independent intelligent agent, construct a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning, and establish a scheduling model based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost, and hydrogen shortage penalty cost as the global reward function. S5. The QMIX algorithm under the centralized training and decentralized execution paradigm is used for policy learning to achieve adaptive and collaborative scheduling of hydrogen energy after disasters.

[0024] like Figure 2 As shown, steps S1-S2 specifically include: S11. Construct the objective function of the electro-hydrogen coupling collaborative recovery optimization model, comprehensively considering the cost of electricity load reduction and the cost of hydrogen supply interruption penalty, and achieve flexible adjustment of the power supply priority of important loads and hydrogen production stations through differentiated weight coefficients: in, This is a set of discrete time periods during disaster recovery, and the time interval can be set to 1 hour according to actual needs. A set of power load nodes; For the set of hydrogen production station nodes; For nodes The power load reduction cost weighting coefficient, the larger the value, the higher the load priority of that node; For hydrogen production station The hydrogen energy cost reduction weighting coefficient, the larger the value, the higher the priority of hydrogen production station power supply; Indicates the time period ,node Forced reduction in electricity load; Indicates the time period hydrogen production station Hydrogen supply gap; through adjustment and The relative size of the loads determines the differentiated recovery strategy that prioritizes critical loads or hydrogen production stations.

[0025] S21. Set spatiotemporal dynamic scheduling constraints for mobile energy storage: S211. Travel time calculation constraint: The migration time of mobile energy storage is calculated by the ratio of the corrected distance to the actual vehicle speed, as shown in the following formula: in, The time required for the mobile energy storage k to travel from node i to node j is given in hours. Let be the shortest distance from node i to j, in km; The speed of the mobile energy storage device k during time period t is expressed in km / h. S212. Travel distance correction constraint, to quantify the indirect impact of reduced road traffic efficiency caused by disasters on travel distance, the formula is as follows: in, For nodes and Topological distance between them For time period node Reference vehicle speeds on surrounding roads quantify the indirect impact of reduced road traffic efficiency caused by disasters on travel distance; S213. Speed ​​decay constraint: Establishes a quantitative relationship between disaster level and mobile energy storage travel speed. This allows for dynamic adjustment of the mobile energy storage's travel capacity based on different disaster intensities, making the model more closely reflect actual post-disaster traffic conditions. The formula is as follows: in, The speed of mobile energy storage under ideal operating conditions is expressed in km / h. The disaster impact coefficient; , Indicates no disaster. This indicates that the transportation network is completely paralyzed; S214. Node connection uniqueness constraint ensures that a single mobile energy storage device can only be deployed on one distribution network node at a time, conforming to the spatial uniqueness principle of physical devices and avoiding logical contradictions such as the same device working on multiple nodes simultaneously. The formula is as follows: in, For MESS at node Time period The pre-layout state, For MESS in the time period From node arrive The scheduling connection status, A collection of nodes that can be deployed for Mobile Energy Storage System (MESS). For nodes The set of adjacent nodes, i.e., traffic-reachable nodes; S215. Continuity constraint during the migration period: This ensures that charging and discharging operations can only commence after the mobile energy storage has completed its migration and equipment configuration, avoiding infeasible dispatch schemes where power supply begins before migration is complete. The formula is as follows: in, For time period From node arrive The passage time, Fixed device configuration time for MESS; S216. Charging and discharging state coupling constraint: Establish the coupling relationship between the mobile energy storage deployment state and the charging and discharging state: Charging and discharging operations can only be performed on node i when the mobile energy storage is successfully deployed, logically eliminating the unreasonable situation of charging and discharging before deployment. The formula is as follows: in, For MESS at node Time period Charge and discharge indicators, For distribution network nodes; S217. Charge and discharge power constraints: These limits the upper limits of the charging and discharging power of mobile energy storage to prevent over-power operation and prohibit simultaneous charging and discharging operations, thus protecting the safety and lifespan of the battery energy storage system. The formula is as follows: in, For MESS at node Time period The charging and discharging power, This is the upper limit of charging and discharging power. The maximum charge and discharge energy of MESS must be within the energy storage capacity of the device, and energy conservation must be ensured in conjunction with power constraints.

[0026] S218. State of charge constraints, including the energy conservation equation for mobile energy storage, accurately tracking real-time changes in battery charge; and limiting the state of charge within a safe range to prevent irreversible damage to the battery from overcharging or over-discharging, complying with the operating specifications of energy storage devices, as shown in the following formula: in, For MESS at node Time period Real-time state of charge, For charging and discharging efficiency, To ensure energy conservation and equipment safety, MESS sets minimum / maximum safe SOC.

[0027] S22. Set fixed energy storage charging and discharging constraints: S211. Charge and discharge power constraint, consistent with the principle of charge and discharge power constraint for mobile energy storage, limits the charge and discharge power range of fixed energy storage and prohibits simultaneous charge and discharge to ensure safe operation of the equipment. The formula is as follows: in, For MESS at node Time period The charging and discharging power, For FESS charging and discharging indication, the device's charging and discharging power must be within the capacity range.

[0028] S212. State of charge constraints are used to achieve energy conservation and safe operation control of stationary energy storage, ensuring that stationary energy storage can continuously and stably participate in grid regulation. The formula is as follows: in, For FESS at the node Time period Real-time state of charge, For charging and discharging efficiency, To ensure energy conservation and equipment safety, MESS sets minimum / maximum safe SOC.

[0029] S23. Set load shedding constraints, stipulating that the load shedding amount cannot exceed the actual load of the node to avoid logical errors caused by negative loads; stipulate that the reactive power shedding amount and the active power shedding amount must be reduced proportionally according to the reactive power-active power ratio when the node is operating normally, to ensure that the power factor of the load remains unchanged after shedding, maintain grid voltage stability, and avoid voltage fluctuation problems caused by reactive power excess or deficiency. The formula is as follows: in, For nodes Time period The reduction in active power load should be less than the maximum active power load at the node. . For nodes Time period The amount of reactive load reduction, For nodes Maximum reactive load; S24. Set output constraints for distributed power sources, limiting their active and passive power output ranges, and restricting them to output power only when connected to the grid; control the power factor of distributed power sources within the allowable range, wherein photovoltaic power generation systems operate at a constant power factor, with equal upper and lower limits, as shown in the following formula: in, For nodes Active power output of distributed power sources For nodes The connection status of distributed power sources with the power grid. This represents the upper limit of active power output of distributed power sources. For nodes Reactive power output of distributed power sources This represents the upper limit of reactive power output of distributed power sources. These are the upper and lower limits of the power factor for distributed generation. S25. Set distribution network operation constraints: S251. Active power balance constraint is one of the core constraints for distribution network operation. The left side of the equation is the sum of the active power output of all active power sources at node i, and the right side is the sum of the active load, charging power, line active power loss, and outflowing active power at node i. It ensures the balance of active power supply and demand in the entire distribution network. The formula is as follows: in, For the line from arrive active power, For the line The resistance, For the line The current, For nodes Distributed power generation has active power output. node The maximum active load, For nodes The hydrogen production station has active power demand. For nodes Active load reduction.

[0030] S252. Reactive power balance constraint, corresponding to active power balance constraint, the left side of the equation is the sum of the reactive power output of all reactive power sources at node i, and the right side is the sum of the reactive load, charging power, line reactive power loss and outflow reactive power at node i, ensuring the reactive power supply and demand balance of the distribution network and maintaining the grid voltage level: in, For the line from arrive reactive power, For the line Reactance, For nodes Distributed power generation reactive power output, node Maximum reactive load, For nodes Reactive load reduction; S253. Voltage Equation Relaxation Constraints: The original distribution network voltage equations are nonlinear, making direct solution difficult. This constraint relaxes the nonlinear terms in the voltage equations into linear constraints by introducing a large constant M. When the line is in operation, the constraint degenerates into the original voltage equations; when the line is out of operation, the constraint is automatically satisfied, thus significantly reducing computational complexity while ensuring accuracy. The formula is as follows: in, For nodes voltage amplitude, For the line The operational status, Since it is a large constant, the nonlinear terms in the voltage equation are relaxed into linear constraints, reducing computational complexity.

[0031] S254. Node voltage and line current safety constraints limit node voltage and line current to within permissible ranges, preventing voltage exceeding limits from damaging electrical equipment and preventing tripping accidents caused by line overload. The formula is as follows: S255. Second-order cone relaxation constraints transform the nonlinear current-power-voltage relationship into a second-order cone form, making the entire model a mixed-integer second-order cone programming problem. This allows for efficient solving using commercial solvers such as Gurobi, significantly improving computational speed while maintaining solution quality. The formula is as follows: S26. Quantitative Electricity Supply and demand linkage mechanism in hydrogen conversion: S261. Electrolyte-to-hydrogen conversion efficiency constraint: Establish a quantitative relationship between the electrolyzer's power input and hydrogen production, reflecting the electrolyzer's energy conversion characteristics. The formula is as follows: in, For hydrogen production station Time period Electrolytic cell output, The electro-hydrogen conversion coefficient is determined by the efficiency of the electrolyzer. This refers to the electrical input power of the electrolytic cell. This is a temperature correction factor; S262. Temperature correction factor calculation: This quantifies the impact of ambient temperature on electrolyzer efficiency, enabling the model to more accurately calculate hydrogen production at different temperatures. The formula is as follows: S263. Electrolytic Cell Operating Power Range Constraints: This constraint stipulates that the operating power of the electrolytic cell must be within the allowable range. The minimum power is typically set at 25% of the maximum power, complying with the equipment protection regulations for industrial electrolytic cells. This avoids decreased equipment efficiency and shortened lifespan due to low-load operation. The formula is as follows: in, These are the minimum and maximum operating power of the electrolytic cell, with the minimum power set at 25% of the maximum power, in accordance with industrial equipment protection regulations. S264. Hydrogen Compression Power Constraint: The hydrogen produced at the hydrogen production station needs to be compressed to a high-pressure state before it can be stored and transported. This constraint quantifies the power consumption of the hydrogen compression process and is an important component of the total power demand of the hydrogen production station. The formula is as follows: in, This refers to the power consumption of the hydrogen compression process. This is the compression power consumption coefficient, which is related to the compressor efficiency; S265. Total Power Demand Constraint for Hydrogen Production Station: The total power demand of the hydrogen production station is decomposed into two parts: electricity consumption for the electrolyzer and electricity consumption for hydrogen compression. This reflects the power load characteristics of the hydrogen production station, as shown in the following formula: in, To meet the total electricity demand of the hydrogen production station, This represents the power consumption of the hydrogen compression process.

[0032] S266. Maximum power supply capacity constraint of distribution network nodes: This limits the total power demand of the hydrogen production station to the maximum power supply capacity of the distribution network nodes, preventing excessive power load from the hydrogen production station and thus overloading the grid nodes, ensuring the safe and stable operation of the distribution network. The formula is as follows: in, For the distribution network at the node The maximum power supply capacity limits the total power demand to the maximum power supply capacity of the distribution network nodes, thus preventing new loads from exceeding the grid's carrying capacity.

[0033] S267. Hydrogen supply and demand balance constraints within the hydrogen production station, including the mass balance equation for hydrogen storage tanks, accurately tracking changes in hydrogen storage levels within the station; and converting hydrogen storage levels into storage tank pressure based on the ideal gas law, limiting the pressure within a safe range to prevent overpressure operation of the storage tanks and potential safety accidents. The formula is as follows: in, Indicates hydrogen production station Time period Hydrogen storage capacity, For the external supply of hydrogen, For the hydrogen storage tank pressure, Where is the hydrogen gas constant. It is the compression factor; S268. Quantitative Constraints on Hydrogen Supply Gap: The hydrogen supply gap for a hydrogen production station is defined as follows: when the sum of the station's hydrogen production and available hydrogen storage can meet external demand, the gap is 0; when it cannot, the gap equals the difference between demand and available supply. A non-zero gap will trigger a penalty term in the objective function, guiding the model to prioritize ensuring the hydrogen supply to the hydrogen production station, achieving coordinated optimization of power restoration and hydrogen stability. The formula is as follows: in, For hydrogen production station Time period To meet external hydrogen demand, the on-site storage capacity must satisfy that demand; otherwise, a hydrogen supply gap will occur. When the sum of hydrogen production and available hydrogen storage cannot meet the demand, Non-zero values ​​trigger the penalty term in the objective function to quantify the loss from hydrogen supply interruption and achieve coordinated optimization of electricity and hydrogen. This represents the minimum amount of hydrogen stored. S268. Hydrogen production station operation status constraints achieve coupled control of the hydrogen production station's operation status and total power demand. When the hydrogen production station is shut down, the total power demand is 0; when the hydrogen production station is running, the total power demand must be within the allowable range, as shown in the following formula: in, In running state, Indicates that the program is running. Indicates that the machine is out of service. It represents the minimum / maximum total power demand, enabling coupled control of operating status and power; S269. Hydrogen Production Station Restart Interval Constraint: Hydrogen production stations are large industrial equipment, and frequent start-ups and shutdowns can lead to equipment fatigue, decreased efficiency, and even damage. This constraint stipulates that a hydrogen production station must remain shut down for at least two time intervals before restarting, complying with the hot-start procedure for hydrogen production units and extending equipment lifespan. The formula is as follows: in, This indicates that at least two time intervals are required after a shutdown before restarting to avoid equipment fatigue caused by frequent start-stop cycles, which complies with the hot start procedure for hydrogen production units. S27. Transform the above objective function and constraints into a mixed-integer second-order cone programming model, and use the Gurobi commercial solver to solve it, obtaining the post-disaster power restoration plan and the hydrogen production station power supply plan.

[0034] Step S3 specifically includes: S31. Construct the objective function of the hydrogen transportation route optimization model, comprehensively considering the fixed scheduling cost of hydrogen long-tube trailers, the variable transportation cost per unit distance, the waiting time cost at hydrogen refueling stations, and the penalty cost for insufficient hydrogen supply, to achieve a balance between transportation economy and hydrogen supply reliability: Wherein, K represents the collection of hydrogen tube trailers; A set of hydrogen refueling station nodes; The fixed dispatch cost per vehicle is expressed in yuan per vehicle. It is a 0-1 variable, indicating whether vehicle k is scheduled; The variable cost per unit distance for vehicles is expressed in yuan / km. It is a 0-1 variable, indicating whether vehicle k travels from node i to node j; Let be the shortest distance from node i to j, in km; The unit waiting time cost at hydrogen refueling stations is expressed in yuan / hour. The waiting time for vehicle k at a hydrogen refueling station (r), in hours (h). The penalty cost coefficient for hydrogen refueling stations is r, expressed in yuan / kg; The minimum supply threshold for hydrogen refueling stations is typically set at 80% of demand, expressed in kg. The amount of hydrogen supplied from vehicle k to hydrogen refueling station r, in kg; S32. Establish a constraint system, constructing a complete constraint framework from four dimensions: path feasibility, hydrogen balance, time constraints, and cost rationality, to ensure that the scheduling scheme is physically feasible and economically reasonable: S321. Path constraints are used to regulate the travel path of hydrogen tube trailers, ensuring that the path conforms to post-disaster traffic conditions and hydrogen transportation logic: S3211. Vehicle Origin Constraint: The hydrogen production station is the sole hydrogen source in the hydrogen energy supply chain, and hydrogen-powered tube trailers can only load hydrogen at hydrogen production stations. This constraint mandates that all dispatched vehicles must depart from the hydrogen production station, ensuring that all transportation tasks originate from the hydrogen source, conforming to the physical logic of hydrogen energy transportation, as shown in the following formula: Wherein, G is the set of hydrogen production station nodes; It is a 0-1 variable, indicating whether vehicle k travels from hydrogen production station node g to node j; It is a 0-1 variable, indicating whether vehicle k is scheduled to perform a transportation task.

[0035] S3212. Vehicle Destination Constraint: Post-disaster hydrogen transportation is a multi-destination problem, where each vehicle does not need to return to the hydrogen production station after completing all hydrogen unloading tasks. This constraint transforms the original multi-destination vehicle routing problem into a single-destination problem by introducing virtual nodes, significantly simplifying the modeling complexity. The formula is as follows: Where V is the set of virtual nodes. This constraint uniformly sets the destination of all scheduled vehicles to virtual node v. The distance and travel time between the virtual node and all hydrogen refueling stations are set to 0. Once a vehicle completes its last hydrogen unloading task, it is considered to have arrived at the virtual node without needing to travel further, thus simplifying the modeling complexity of the multi-destination problem. S3213. Path Continuity Constraint: The vehicle's travel path must be a continuous line segment, without jumps or interruptions. This constraint ensures that for any non-starting and non-ending node j, the number of times vehicle k enters that node is equal to the number of times it leaves that node, avoiding the physical contradiction of path interruption or repeated node visits. The formula is as follows: in, This represents all road network nodes except for hydrogen production stations and virtual nodes, i.e., hydrogen refueling station nodes; S3214. No Loop Constraint: Vehicles should not encounter meaningless internal loops during transportation, otherwise it will increase travel distance and time, reducing transportation efficiency. This constraint ensures that no internal loops appear in the vehicle's travel path by forcing a monotonically increasing node access order, avoiding unnecessary detours. The formula is as follows: in, A is an auxiliary variable representing the order in which vehicle k visits node i; A is the set of feasible road segments; M is a sufficiently large positive integer. S3215. Hydrogen production station access order constraint: This constraint stipulates that the hydrogen production station is the first node visited by vehicle k, and its access order number is fixed at 0. This complements the vehicle origin constraint and further clarifies the starting logic of the path. The formula is as follows: S3216. Disaster-damaged road section prohibition constraint. This constraint forces all vehicles to exclude disaster-damaged road sections from their travel routes to ensure the physical feasibility of the dispatching plan. The formula is as follows: in, This is a collection of road sections that are impassable due to disasters.

[0036] S322. Set hydrogen quantity constraints to regulate hydrogen loading, unloading, and storage during transportation, ensuring hydrogen balance and compliance with equipment capacity limits: S3221. Initial hydrogen transport volume constraint: To maximize the hydrogen utilization rate of a single transport and reduce the unit hydrogen transport cost, the dispatched vehicles must be fully loaded when departing from the hydrogen production station, as shown in the following formula: in, The amount of hydrogen remaining when vehicle k leaves hydrogen production station node g; This represents the vehicle's maximum hydrogen carrying capacity. S3222. Minimum Supply Threshold Definition: In the event of a disaster leading to insufficient transportation capacity, priority is given to ensuring the basic operational needs of hydrogen refueling stations, preventing a complete disruption of hydrogen refueling stations and the resulting social impact. This constraint sets the minimum supply threshold at 80% of demand to guarantee the basic operational needs of hydrogen refueling stations, as shown in the following formula: in, The amount of hydrogen required by the hydrogen refueling station r; The minimum supply threshold for hydrogen refueling station r; S3223. Minimum Hydrogen Supply Constraint for Hydrogen Refueling Stations. This constraint mandates that each hydrogen refueling station receive a total hydrogen supply that meets at least its minimum supply threshold; otherwise, a penalty cost in the objective function will be triggered, as shown in the following formula: in, The amount of hydrogen supplied to vehicle k from hydrogen refueling station r; S3224. Vehicle dispatch quantity constraint: This constraint ensures that the total capacity of dispatched vehicles can meet the minimum supply demand of all hydrogen refueling stations, avoiding systemic hydrogen supply gaps caused by insufficient vehicle numbers. The formula is as follows: S3225. Single-station hydrogen unloading capacity limit constraint: This constraint stipulates that hydrogen can only be unloaded at a hydrogen refueling station when vehicle k visits the station r, and the amount of hydrogen unloaded in a single transaction must not exceed the equipment's capacity limit. The formula is as follows: in, This is the maximum amount of hydrogen that a vehicle can unload in a single trip; S3226. Hydrogen unloading amount and remaining hydrogen amount constraint. This constraint ensures that the amount of hydrogen unloaded by the vehicle to the hydrogen refueling station r does not exceed its remaining hydrogen amount upon arrival at the station, avoiding the physical contradiction of negative hydrogen amount. The formula is as follows: in, The amount of hydrogen remaining when vehicle k leaves node i; S3227. Remaining Hydrogen Supply Update Constraint: This constraint stipulates that after vehicle k travels from node i to hydrogen refueling station node j and completes hydrogen unloading, its remaining hydrogen supply is equal to the remaining hydrogen supply upon arrival at the station minus the unloaded hydrogen supply, achieving a dynamic balance of hydrogen supply during transportation. The formula is as follows: S3227. Non-negative hydrogen transport capacity constraint: This constraint ensures that the remaining hydrogen capacity of the vehicle remains non-negative throughout the entire transportation process, conforming to physical reality. The formula is as follows: S323. Set time constraints to regulate time consumption during transportation, ensuring transportation safety and timely hydrogen supply: S3231. Total Travel Time Calculation Constraint. This constraint calculates the total travel time of the vehicle by summing the travel times of each road segment, as shown in the following formula: in, The total travel time for vehicle k to complete all transportation tasks; is the shortest distance from node i to j; v is the average speed of the vehicle. S3232. Maximum Driving Time Constraint: This constraint limits the continuous operating time of a single vehicle to ensure transportation safety and driver rest needs. The formula is as follows: in, The maximum permitted driving time for the vehicle, in hours (h). S3233. Single-station hydrogen unloading time constraint: This constraint controls the hydrogen unloading operation time of a vehicle at a single hydrogen refueling station, avoiding overall scheduling delays due to excessively long unloading times. The formula is as follows: in, The hydrogen unloading rate of the vehicle; This is the upper limit for hydrogen unloading time at a single station; S3234. Maximum waiting time constraint at hydrogen refueling stations. This constraint limits the waiting time at hydrogen refueling stations due to hydrogen shortage, avoiding service interruptions caused by hydrogen supply delays. The formula is as follows: in, The waiting time for vehicle k at the hydrogen refueling station (r); This is the maximum allowed waiting time at a hydrogen refueling station; S324. Set non-negativity constraints on costs to ensure the economic rationality of cost calculations: S3241. Fixed Cost Non-Negative Constraint: This constraint ensures that the fixed scheduling cost of vehicles is non-negative, which is consistent with economic logic. The formula is as follows: S3242. Variable cost non-negativity constraint: This constraint ensures that the vehicle's variable transportation cost is non-negative and positively correlated with the travel distance, as shown in the following formula: S3243. Non-negativity constraint on penalty cost: This constraint ensures that the penalty cost for insufficient hydrogen supply at a hydrogen refueling station is non-negative. A penalty is only incurred when the actual hydrogen supply falls below a minimum threshold. The formula is as follows: S33. Transform the above objective function and constraints into a mixed-integer linear programming model, call the Gurobi commercial solver to solve it, and obtain the globally optimal scheduling scheme for post-disaster hydrogen energy transportation.

[0037] Step S4. Construct a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning, model the hydrogen tube trailer as an independent agent, and establish a scheduling model based on a distributed partially observable Markov decision process: S41. Construct a scheduling model based on the Decentralized Partially Observable Markov Decision Process (Dec-POMDP). This model consists of octets. Formal definition, applicable to fully cooperative task scenarios involving post-disaster information localization and multi-agent collaboration: S411. Define a set of agents, modeling each hydrogen tube trailer as an independent decision-making agent, with each agent possessing independent observation, decision-making, and execution capabilities. This modeling approach gives the system good scalability; when the number of transport vehicles changes, there is no need to redesign the entire scheduling model. The formula is as follows: Where n is the total number of hydrogen tube trailers participating in the transportation mission.

[0038] S412. Define the global state space. The global state space encompasses all environmental information relevant to the task and is the basis for the agent's decision-making. It is composed of the road network state. Vehicle status and demand status It is composed of a three-dimensional Cartesian product, as shown in the following formula: S4121. Road network state space, reflecting the topological connectivity of the post-disaster transportation network in real time, includes the traffic status information of all nodes and road segments. The formula is as follows: in, It is a collection of road network nodes, including hydrogen production stations, hydrogen refueling stations, and road hubs; The set of edges is divided into feasible edge sets. and sets of impassable edges caused by disasters The road network status reflects the topological connectivity of the post-disaster transportation network in real time; S4122. Vehicle State Space: Describes the real-time operating status of all hydrogen tube trailers, including location, remaining hydrogen quantity, and pending tasks. The formula is as follows: in, Let k be the real-time location of vehicle k. Let be the remaining hydrogen quantity in vehicle k. This represents the vehicle's maximum hydrogen carrying capacity. For vehicle k, there is a list of tasks to be performed, with each task tuple as an example. This indicates that it needs to be a hydrogen refueling station. supply kg of hydrogen; S4123. Demand state space, describing the hydrogen supply demand status of all hydrogen refueling stations, including remaining demand and waiting time information. The formula is as follows: in, For the demand vector of hydrogen refueling stations, The remaining demand for hydrogen refueling station r; Let this be the waiting time vector at the hydrogen refueling station. The cumulative waiting time for the hydrogen refueling station r since it issued the hydrogen supply request is expressed in hours. S413. Define the joint action space, which is the Cartesian product of the action spaces of all agents. The action space of a single agent includes two types: selecting a target hydrogen refueling station and waiting. Selecting a target hydrogen refueling station means that the vehicle goes to that refueling station to perform a hydrogen unloading task. The waiting action is applicable when there are no available tasks or insufficient remaining hydrogen, as shown in the following formula: Where R represents the set of all hydrogen refueling stations; wait indicates a waiting action, applicable to situations where there are no available tasks or insufficient remaining hydrogen. S4131. Action Feasibility Constraint: This constraint ensures that the remaining demand of the target hydrogen refueling station r selected by the agent does not exceed the current remaining hydrogen supply of the vehicle, thus preventing the task from failing due to insufficient capacity. The formula is as follows: S414. Define a state transition function. The state transition function describes the change in the state of the environment after performing a joint action. It follows the Markov property, that is, the next state is determined only by the current state and the joint action performed, and is independent of the historical state. The formula is as follows: S4141. Vehicle State Transition describes the changes in the vehicle's state after performing an action, including updates to its location, remaining hydrogen supply, and task list. After performing an action, the vehicle's location is updated to the endpoint of the shortest feasible path planned by the environment. After completing the hydrogen supply task, the remaining hydrogen supply is reduced by the amount already supplied, and completed tasks are removed from the pending task list, as shown in the following formula: in, Based on the current road network conditions, the environment plans the shortest feasible path from the current location to the target hydrogen refueling station for vehicle k; This is the set of hydrogen refueling stations for vehicle k that will complete its hydrogen supply in the current time step. The amount of hydrogen supplied to vehicle k from hydrogen refueling station r; S4142. Demand State Transition describes changes in the demand state of hydrogen refueling stations, including waiting time and updates to remaining demand. The formula is as follows: in, The scheduling time step is expressed in hours (h). S415. Define a global reward function, which takes the form of a negative cost. This function converts various economic and time costs during the scheduling process into negative reward signals, guiding the agent to minimize the overall scheduling cost through trial and error learning. All agents share the same global reward function, ensuring the consistency of the objectives for multi-vehicle collaborative optimization. S4151. The global reward function is generally defined, and its proportion in the total reward is adjusted by weighting coefficients. This allows for flexible adjustment based on the resilience priority of different disaster scenarios. A negative sign indicates that the higher the cost, the lower the reward received by the agent. The formula is as follows: in, These are non-negative weighting coefficients that adjust the proportion of driving cost, waiting cost, and penalty cost in the total reward, and can be flexibly adjusted according to the resilience priority of different disaster scenarios; the negative sign indicates that the higher the cost, the lower the reward obtained by the agent.

[0039] S4152. Total operating cost calculation includes both fixed dispatch costs and variable transportation costs for all dispatched vehicles. The formula is as follows: in, The fixed dispatch cost per vehicle includes fixed costs such as equipment depreciation and basic staff salaries. The variable transportation cost per unit distance for vehicles includes marginal costs such as fuel consumption and tire wear. Distance between road sections; S4153. Total waiting cost calculation is the weighted sum of waiting times at all hydrogen refueling stations, guiding the agent to prioritize responding to requests from stations with longer waiting times. The formula is as follows: in, To quantify the unit waiting time cost of hydrogen refueling stations, and to quantify the operational losses, customer churn risks, and social impacts caused by hydrogen shortages during waiting times. S4154. Calculation of Penalty Costs for Insufficient Hydrogen Supply: Penalty costs are triggered only when the actual hydrogen supply falls below a minimum threshold. The penalty coefficient can be set differently based on the importance of the hydrogen refueling station. The formula is as follows: in, The penalty coefficient for insufficient hydrogen supply at a hydrogen refueling station (r) can be set differently based on the importance of the service targets of the hydrogen refueling station (such as hospitals, transportation hubs, and industrial parks). This represents the minimum supply threshold for hydrogen refueling station r.

[0040] Local observation space With observation function Define a single intelligent agent Based on the observation function Perceptible local observation information The agent's observable range is 80% of the map's diagonal distance; Strategy This represents the action distribution of a single agent based on its observation history. Represents intelligent agents Action-observation history is the sequence of all observations and actions performed by the agent from the initial moment to the current moment. Step S5. The QMIX algorithm, employing a centralized training and distributed execution paradigm, is used for policy learning to achieve adaptive collaborative scheduling of hydrogen energy after a disaster. For example... Figure 3As shown, the algorithm process is divided into two core stages: offline centralized training and online distributed execution. It solves the credit allocation problem in multi-agent fully cooperative scenarios through a monotonic value decomposition mechanism, ensuring that the combination of optimal actions for individuals is equivalent to the globally optimal actions. At the same time, it is perfectly adapted to the actual scenario of localized post-disaster information. S51. Design the network structure for the QMIX algorithm, which consists of two parts: an individual agent network and a hybrid network. The individual agent network is responsible for calculating the individual Q-values ​​based on local observations, while the hybrid network is responsible for nonlinearly fusing the individual Q-values ​​into a global Q-value. S511. Construct an individual agent network, employing the same two-layer fully connected shared network structure as the IDQN algorithm. All agents share the same set of network parameters to reduce the number of parameters and improve training efficiency. The input is the local state of a single trailer. The output is the individual Q-value for each action. ;in, These are the individual network parameters. The input layer dimension is determined by the local state elements, the hidden layer contains 256 ReLU activated neurons, and the output layer dimension is equal to the action space size. Reusing the pre-trained network parameters of IDQN can accelerate the convergence process of QMIX; S512. Construct a hybrid network, taking the global state s as input, and dynamically generate non-negative weights and biases through a super network to nonlinearly map the individual Q-value vectors to the global Q-values; S5121. Hybrid network forward computation utilizes a supernetwork to dynamically generate weights and biases based on the global state, enabling the hybrid network to adapt to different environmental conditions. The formula is as follows: in, The global Q value, This is the global state. For the non-negative weights of the hybrid network, For bias, It is a linear activation function, ensuring a linear and monotonic relationship between the global Q-value and the individual Q-value; S5122. Monotonicity Constraint: This constraint is the core of the QMIX algorithm. It ensures that the global Q-value is monotonically non-decreasing with respect to the Q-values ​​of all individuals. That is, increasing the Q-value of any individual agent will not lead to a decrease in the global Q-value, thus ensuring consistency between individual optima and global optima. The formula is as follows: S5123. Global optimal action decomposition, derived from monotonicity constraints, shows that the globally optimal joint action can be obtained by each agent independently choosing the action with the highest individual Q value, providing a theoretical basis for distributed execution. The formula is as follows: This formula shows that the globally optimal joint action can be obtained by each agent independently choosing the action with the largest individual Q value, providing a theoretical basis for distributed execution. S52. Execute a centralized training process. During the training phase, global information is used to optimize the joint strategy, an experience replay mechanism is used to mitigate sample correlation, and target network technology is employed to improve training stability. S521. Initialize parameters, initialize network parameters. Including agent network parameters Hybrid network parameters Target network parameters Initialize the experience replay buffer ; S522. Start the training round loop. For each training round, first initialize the environment state. Reset the positions, remaining hydrogen, and task lists of all agents; then enter a time step loop until the maximum trajectory length is reached or all tasks are completed. S5221. Execute action selection. For each agent k, select an action according to an ε-greedy policy: with probability... Select the action that maximizes the Q value under the current local observation. With probability Randomly select an action to ensure that the algorithm can explore unknown scheduling strategies and avoid getting trapped in local optima; S5222. Environmental interaction occurs, with all agents executing their chosen actions to form a joint action. The environment updates the global state according to the state transition function and generates a global reward. and return to the next state. ; S5223. Store experience data and combine experience tuples. Stored in the experience replay buffer D for subsequent model updates; S5224. Perform model update. When the size of the experience replay buffer is greater than the batch size B, randomly sample B experience samples from the buffer to form the batch data. For each sample i, the current global Q-value is calculated using a hybrid network. Then, the TD target is calculated through the target network using the following formula: in, The target network output is used to provide a stable training objective; then the loss function is calculated according to the following format: Finally, all network parameters are updated using the Adam gradient descent algorithm. Minimize the loss function; S5225. Update the target network parameters. Every 100 time steps, copy the parameters of the current hybrid network to the target network. This ensures that the update frequency of the target network parameters is lower than that of the main network, thus avoiding oscillations or divergence during training. S53. Execute the distributed execution process. After training, the execution phase begins. At this stage, no global information is required. Each hydrogen tube trailer makes autonomous decisions based solely on its local observations. Each agent independently selects the action with the highest individual Q value. Through the cooperative strategy learned during the training phase, globally optimal hydrogen transportation scheduling is implicitly achieved. No central node control is required, resulting in extremely strong robustness and real-time response capabilities.

[0041] Another embodiment provides a disaster resilience enhancement system for an integrated electric-hydrogen energy system, such as... Figure 4 As shown, system 400 includes: The collaborative recovery model construction module 410 is used to construct an electric-hydrogen coupled collaborative recovery optimization model based on power load data and hydrogen energy supply data, with the objective function being to minimize the sum of power load reduction costs and hydrogen energy supply interruption penalty costs. The constraint integration and solution module 420 is used to integrate spatiotemporal dynamic scheduling constraints of mobile energy storage, charging and discharging constraints of stationary energy storage, load shedding constraints, distributed power generation output constraints, safe operation constraints of distribution networks, and operation and electricity-hydrogen conversion constraints of hydrogen production stations, quantifying the constraints of electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; The centralized scheduling module 430 is used to construct a hydrogen energy transportation route optimization model based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, with the objective function being to minimize the sum of hydrogen tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen energy transportation resources. The distributed strategy modeling module 440 is used to model the hydrogen tube trailer as an independent intelligent agent, construct a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning, and establish a scheduling model based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost and insufficient hydrogen supply penalty cost as the global reward function. The algorithm training and execution module 450 is used to learn policies using the QMIX algorithm under the centralized training and decentralized execution paradigm to achieve adaptive and collaborative scheduling of hydrogen energy after disasters.

[0042] In addition to the modules described above, the post-disaster resilience enhancement system 400 of the integrated electric-hydrogen energy system may also include other components. However, since these components are not related to the content of the embodiments of this disclosure, their illustrations and descriptions are omitted here.

[0043] Other specific working processes of the method for enhancing the post-disaster resilience of the integrated electric-hydrogen energy system using the aforementioned post-disaster resilience enhancement system 400 are described in the above-described embodiment of the method for enhancing the post-disaster resilience of the integrated electric-hydrogen energy system, and will not be repeated here.

[0044] This invention also provides a non-transitory computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the post-disaster resilience enhancement method for the integrated electric-hydrogen energy system described in the above embodiments. The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0045] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0046] The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0047] Furthermore, other specific operating procedures of a non-temporary computer-readable storage medium are described in the above-described embodiment of the method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system, and will not be repeated here.

[0048] In this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a step or method that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such a step or method.

[0049] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system, characterized in that, The method includes the following steps: Based on electricity load data and hydrogen energy supply data, a collaborative recovery optimization model for electricity-hydrogen coupling is constructed with the objective function of minimizing the sum of electricity load reduction costs and hydrogen energy supply interruption penalty costs. Integrating spatiotemporal dynamic scheduling constraints for mobile energy storage, charging and discharging constraints for stationary energy storage, load shedding constraints, distributed power output constraints, and distribution network safety operation constraints, this approach quantifies electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; Based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, a hydrogen energy transportation route optimization model is constructed with the objective function of minimizing the sum of hydrogen long-tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen energy transportation resources. The hydrogen tube trailer is modeled as an independent intelligent agent, and a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning is constructed. A scheduling model is established based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost and insufficient hydrogen supply penalty cost as the global reward function. The QMIX algorithm, employing a centralized training and decentralized execution paradigm, is used for policy learning to achieve adaptive and collaborative scheduling of hydrogen energy after disasters.

2. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The collaborative recovery optimization model for the electro-hydrogen coupling is expressed as follows: in, Represents the set of discrete time periods during disaster recovery; Represents the set of power load nodes; Represents the set of hydrogen production station nodes; Represents a node The weighting factor for electricity load reduction costs Indicates hydrogen production station The weighting coefficient for hydrogen energy reduction costs; Indicates the time period ,node Forced reduction in electricity load, Indicates the time period Hydrogen production station Hydrogen supply gap; through adjustment and The relative size of the loads determines the differentiated recovery strategy that prioritizes critical loads or hydrogen production stations.

3. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The spatiotemporal dynamic scheduling constraints for mobile energy storage include travel time calculation constraints, travel distance correction constraints, vehicle speed attenuation constraints, node connection uniqueness constraints, travel period continuity constraints, charge / discharge state coupling constraints, charge / discharge power constraints, and state of charge constraints, among which: The constraint expression for calculating the travel time is: ,in, For mobile energy storage during the period From node arrive The actual migration time; This is the corrected distance between nodes; The actual vehicle speed under disaster conditions; The expression for the distance correction constraint is: ,in, For nodes and Topological distance between them; For time period node Reference vehicle speeds on surrounding roads quantify the indirect impact of reduced road traffic efficiency caused by disasters on travel distance; The expression for the vehicle speed attenuation constraint is: ,in, The ideal driving speed. To quantify the level of disaster, , Indicates no disaster. This indicates that the transportation network is completely paralyzed; The expression for the charging / discharging state coupling constraint is: ,in, For mobile energy storage at nodes Time period The charging icon For mobile energy storage at nodes Time period The discharge indicator, For the set of distribution network nodes, For mobile energy storage during the period From node arrive The scheduling connection status; The expression for the charge state constraint is: ,in, For mobile energy storage at nodes Time period Real-time state of charge, For charging efficiency, For discharge efficiency, For mobile energy storage at nodes Time period charging power, For mobile energy storage at nodes Time period The discharge power.

4. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The supply and demand relationship mechanism of the electro-hydrogen conversion process is quantified through constraints on the electrolyzer's electro-hydrogen conversion efficiency, the total power demand of the hydrogen production station, the hydrogen energy balance of the hydrogen storage tank, and the quantitative constraint on the hydrogen energy gap, among which: The expression for the constraint on the electro-hydrogen conversion efficiency of the electrolyzer is: ,in, For hydrogen production station Time period Electrolytic cell output, The electro-hydrogen conversion coefficient is determined by the efficiency of the electrolyzer. This refers to the electrical input power of the electrolytic cell. This is a temperature correction factor; The total power demand constraint expression for the hydrogen production station is: ,in, This refers to the power consumption of the hydrogen compression process. To meet the total electricity demand for hydrogen production; The hydrogen energy balance constraint expression for the hydrogen storage tank is: ,in, Indicates hydrogen production station Time period Hydrogen storage capacity, This refers to the external supply of hydrogen. The quantitative constraint expression for the hydrogen energy gap is: in, For hydrogen production station Time period To meet external hydrogen demand, the on-site storage capacity must be sufficient; otherwise, a hydrogen supply gap will occur. This represents the minimum amount of hydrogen stored.

5. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The hydrogen energy transportation route optimization model is specifically expressed as follows: Where K represents the collection of hydrogen tube trailers, For the set of hydrogen refueling station nodes, The fixed dispatch cost per vehicle. This is a flag indicating whether vehicle k has been dispatched. For the variable cost per unit distance of the vehicle, This is a flag indicating whether vehicle k has traveled from node i to node j. Let i be the shortest distance from node i to j. The unit waiting time cost at hydrogen refueling stations The waiting time for vehicle k at the hydrogen refueling station r. The penalty cost coefficient for hydrogen refueling station r. The minimum supply threshold for hydrogen refueling station r, The amount of hydrogen supplied to vehicle k at hydrogen refueling station r.

6. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 5, characterized in that, Vehicle routing constraints, initial hydrogen transport volume constraints, and supply threshold constraints, specifically including: The vehicle origin constraint, which stipulates that dispatched vehicles must originate from the hydrogen production station, is expressed as follows: Where G represents the set of hydrogen production station nodes. This is a flag indicating whether vehicle k has traveled from node g to node j; The vehicle destination constraint is used to unify the destination of all dispatched vehicles as virtual nodes, and its expression is: Where V is the set of virtual nodes, and the distance between the virtual nodes and all hydrogen refueling stations is set to 0. This is a flag indicating whether vehicle k has traveled from node i to node v; Vehicle path continuity constraints are used to ensure the integrity of vehicle travel paths, and are expressed as follows: The "Disaster-damaged road section inaccessibility" restriction is used to exclude road sections damaged after a disaster, and its expression is: in, This refers to a collection of road sections that are impassable due to disasters. The initial hydrogen transport capacity constraint, used to specify that dispatched vehicles must be fully loaded when departing from the hydrogen production station, is expressed as: in, This refers to the initial hydrogen load carried by the vehicle when it departs from the hydrogen production station. This represents the vehicle's maximum hydrogen carrying capacity. The minimum supply threshold constraint, used to ensure the basic operational needs of hydrogen refueling stations, is expressed as follows: 。 7. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The scheduling model based on the distributed partially observable Markov decision process uses octuples. Define, where: Collection of intelligent agents This represents the number of agents participating in the task, with a value ranging from [value range missing]. Define the scale of multi-agent collaboration; Global state space Includes road network status Vehicle status Demand Status ,Right now ; Joint Action Space The Cartesian product of all agent actions, and the action of a single agent. This indicates the target hydrogen refueling station selected at the current moment or the waiting action; State transition function Describes the current environmental state. Under these circumstances, the intelligent agents execute joint actions. Afterwards, the environment shifted to a new state. The probability distribution; Global reward function Generated by environmental feedback, used to measure the agent's state in the environment. Execute joint actions The benefits are shared by all agents through the same global reward function. To optimize the overall objective; Local observation space With observation function Define a single intelligent agent Based on the observation function Perceptible local observation information The agent's observable range is 80% of the map's diagonal distance. For the action of the k-th agent; Strategy This represents the action distribution of a single agent based on its observation history. Represents intelligent agents Action-observation history is the sequence of all observations and actions performed by the agent from the initial moment to the current moment. Discount factor The weights used to balance current and future rewards.

8. The method for enhancing the post-disaster resilience of an integrated electric-hydrogen energy system according to claim 1, characterized in that, The monotonic value decomposition mechanism of the QMIX algorithm guarantees: This makes the globally optimal joint action equivalent to the combination of the optimal actions of each agent. The algorithm adopts a centralized training and decentralized execution paradigm. During the training phase, global state information is used to dynamically generate non-negative weights and biases of the hybrid network through a hypernetwork, non-linearly fusing the Q-values ​​of individual agents into a global Q-value. During the execution phase, each agent makes independent decisions based only on its own local observations. Represents a set of intelligent agents. The global Q-value is represented by u, which represents the combined action formed by each agent performing its chosen action. Represents the global state. For the state of the kth agent, For the action targeting the k-th agent.

9. A post-disaster resilience enhancement system for an integrated electric-hydrogen energy system, characterized in that, include: The collaborative recovery model construction module is used to construct an electric-hydrogen coupled collaborative recovery optimization model based on electricity load data and hydrogen energy supply data, with the objective function being to minimize the sum of electricity load reduction costs and hydrogen energy supply interruption penalty costs. The constraint integration and solution module integrates spatiotemporal dynamic scheduling constraints for mobile energy storage, charging and discharging constraints for stationary energy storage, load shedding constraints, distributed power generation output constraints, and distribution network safety operation constraints to quantify electricity. The supply and demand relationship mechanism in the hydrogen conversion process is solved to obtain a power restoration scheme; The centralized scheduling module is used to construct a hydrogen transportation route optimization model based on vehicle route constraints, initial hydrogen transport volume constraints, and supply threshold constraints, with the objective function being to minimize the sum of hydrogen tube trailer scheduling costs, hydrogen refueling station waiting time costs, and insufficient hydrogen supply penalty costs. This model aims to achieve the global optimal allocation of hydrogen transportation resources. The distributed strategy modeling module is used to model the hydrogen tube trailer as an independent intelligent agent, construct a distributed hydrogen transportation strategy based on multi-agent deep reinforcement learning, and establish a scheduling model based on a distributed partially observable Markov decision process. The scheduling model uses the weighted sum of driving cost, waiting cost and insufficient hydrogen supply penalty cost as the global reward function. The algorithm training and execution module is used to learn policies using the QMIX algorithm under the paradigm of centralized training and decentralized execution, so as to realize the adaptive and collaborative scheduling of hydrogen energy after disasters.

10. A non-transitory computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method for improving the post-disaster resilience of the integrated electric-hydrogen energy system as described in any one of claims 1 to 8.

Images