Port multi-energy supply and demand intelligent collaborative management and control method based on central decision and double-end execution

Abstract

The invention relates to a central decision and double-end execution port multi-energy supply and demand intelligent collaborative management and control method, which comprises the following steps: S1, constructing an operation environment based on an actual scene of a port; s2, constructing a port multi-energy supply and demand intelligent collaborative management and control framework; s3, collecting data of an energy side and a production side, and synchronizing the data to the multi-energy supply and demand coordination agent; s4, constructing a total data pool and performing multi-energy supply and demand prediction; s5, generating a scheduling instruction through an optimization algorithm, and transmitting the instruction to a port energy management and control agent and a port production management and control agent; s6, performing cooperative control on each energy unit of the port; s7, dynamically optimizing the production task and the equipment time sequence; and S8, the multi-energy supply and demand coordination agent corrects the prediction result based on the execution deviation, and synchronously adjusts a subsequent scheduling decision. The stability, the energy utilization efficiency and the low-carbon operation level of port multi-energy supply and demand cooperative operation are effectively improved, and good engineering application value is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent port management and control, and more specifically, to a method for intelligent collaborative management and control of multi-energy supply and demand in ports, which involves central decision-making and dual-end execution. Background Technology

[0002] Port energy systems are gradually shifting from traditional single-energy supply models to integrated energy systems with multiple coupled energy sources. Existing ports typically connect to green energy resources such as wind and solar power, as well as external grey energy resources, and are equipped with energy storage devices and hydrogen production and refueling systems to meet the diverse electricity and hydrogen demands of port operations. Meanwhile, the port production system exhibits diverse load types, including quay cranes, yard cranes, and port vehicles. Operational plans and load demands dynamically change with the production rhythm, exhibiting significant time-varying and uncertainties. Under these conditions, the degree of coordination between the energy supply side and the production demand side directly affects port operational efficiency, energy utilization levels, and carbon emission control effectiveness.

[0003] However, existing port energy management and production scheduling technologies mostly adopt a subsystem independent operation control mode. This means that the energy side focuses on the local operation control of the electricity or hydrogen energy system, while the production side prioritizes operational efficiency or plan completion during scheduling. This approach makes it difficult to achieve dynamic matching and coordinated regulation between energy supply and production demand, easily leading to problems such as irrational energy allocation, insufficient green energy utilization, and delayed regulatory response. Furthermore, existing technologies have limited utilization of feedback from energy and production execution results, generally lacking a closed-loop dynamic optimization mechanism based on real-time status feedback. The overall adaptive regulation capability and intelligence level of the system still need improvement. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a port multi-energy supply and demand intelligent collaborative management and control method with central decision-making and dual-end execution. It can simultaneously target the port energy system and production system, realize unified perception of supply and demand information, collaborative decision-making and execution feedback linkage, and improve the coordination, stability and overall operational efficiency of the matching between port energy supply and production demand.

[0005] The technical solution adopted by this invention to solve its technical problem is: to construct a port multi-energy supply and demand intelligent collaborative management and control method with central decision-making and dual-end execution, including the following steps: S1. Based on the actual operation scenario of the port, construct an operating environment that includes port microgrid, coordinated supply of green electricity and gray electricity, hydrogen production and hydrogen refueling system, and multiple types of production energy loads; S2. Construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends. S3, the port energy management and control intelligent agent and the port production management and control intelligent agent collect data from the energy side and the production side respectively, and simultaneously provide the data to the multi-energy supply and demand coordination intelligent agent; S4. The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool, predicts production demand and energy supply in a specific future period, and outputs the results in the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions. S5. Based on the production demand forecast and energy supply forecast obtained in step S4, the multi-energy supply and demand coordination intelligent agent generates scheduling instructions through optimization algorithms and transmits the instructions to the port energy management intelligent agent and the port production management intelligent agent. S6. Under the constraints of the collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent performs collaborative control of various energy units in the port microgrid and hydrogen production system. S7. Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent dynamically optimizes production tasks and equipment timing. S8, the port energy management and control intelligent agent and the port production management and control intelligent agent will feed back the execution results of the scheduling instructions, the actual energy supply and consumption data and the production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent will correct the prediction results based on the execution deviation and adjust the subsequent scheduling decisions in a synchronous manner.

[0006] According to the above scheme, step S1, constructing the actual port operation scenario includes the following steps: S101. The port photovoltaic power generation device and wind power generation device will connect the generated green electricity to the microgrid. The port energy storage device is connected to the microgrid and will charge when there is a surplus of green electricity and discharge when there is a shortage of green electricity or during peak load. At the same time, the port microgrid will purchase gray electricity as supplementary energy as needed when local green electricity and energy storage cannot meet the energy demand through connection with the external power grid. S102. The electrical energy output by the microgrid is distributed according to three application scenarios, specifically: the first application scenario provides direct power support for the electric drive equipment of shore cranes, yard cranes, port mobile machinery, and some horizontal transport vehicles through the port sliding contact line power supply system and port charging piles; the second application scenario provides shore power services to docked ships through the shore power system; and the third application scenario supplies power to the port hydrogen production station, uses green electricity to electrolyze hydrogen, and stores and uses the resulting hydrogen as green hydrogen. S103. In the hydrogen energy supply chain, the port hydrogen production station converts electrical energy into hydrogen energy, and the hydrogen refueling station provides hydrogen refueling services to hydrogen-powered vehicles in the port as needed, using the green hydrogen produced or the gray hydrogen purchased from chemical plants around the port.

[0007] According to the above scheme, the method for constructing the intelligent collaborative management and control architecture for multi-energy supply and demand in ports in step S2 includes the following steps: S201, the multi-energy supply and demand coordination intelligent agent, as the central decision-making unit of the control architecture, is responsible for aggregating and analyzing real-time data from the port energy control intelligent agent and the port production control intelligent agent, completing energy supply and demand forecasting and collaborative decision-making, and dynamically correcting the forecast results and decision-making strategies. S202, the port energy management and control intelligent agent and the port production management and control intelligent agent are located at the two ends of the execution respectively. Through bidirectional information interaction with the multi-energy supply and demand coordination intelligent agent, on the one hand, it synchronizes the real-time status data of the energy side and the production side to the central hub, and on the other hand, it receives the scheduling instructions issued by the central hub and executes the corresponding control operations. The execution results are then fed back to the multi-energy supply and demand coordination intelligent agent in real time.

[0008] According to the above scheme, in step S3, Energy-side data includes: microgrid capacity, installed capacity of photovoltaic and wind turbines, rated capacity of energy storage, and upper limit of hydrogen production and refueling station capacity; real-time output of photovoltaic and wind turbines, remaining energy storage capacity, real-time output of hydrogen production and refueling, and purchase prices of grey electricity and grey hydrogen; meteorological forecast data, peak and off-peak electricity price periods, and coordinated data on hydrogen supply stability early warning. Production-side data includes: ship arrival schedules, cargo types and throughput, list of operating equipment, rated energy consumption or hydrogen consumption of equipment; current operation progress, real-time operating status of equipment, and energy demand for sudden operations; operation sequence, peak energy consumption, and operation efficiency fluctuation data within a custom time period. According to the above scheme, step S4, the method for predicting production demand and energy supply in a specific future period includes the following steps: S401, a multi-energy supply and demand coordination intelligent agent integrates energy-side and production-side data to build a full-scale predictive data pool; S402. Determine whether a fixed time interval has been reached or whether an emergency has occurred. Emergency conditions include energy supply failure, sudden operational demand, and sudden changes in weather conditions. If yes, execute steps S403 to S406. If no, execute only steps S404 and S406. S403. A hybrid forecasting model consisting of a time-series forecasting sub-model and a large language model is used to forecast basic energy demand on the production side. S404. An online incremental learning algorithm is adopted, which takes real-time port production operation data and actual energy consumption data for the corresponding period as input. Through continuous monitoring and learning updates of the deviation between predicted energy demand and actual energy consumption, the output of the production-side energy demand prediction model is dynamically corrected, and the output energy demand prediction result that matches the current operating status is given. S405. An energy supply prediction model composed of a machine learning prediction model and a large language model is used to predict the basic energy supply capacity. S406. Combining real-time port energy system operation data and actual energy supply data for the corresponding period, through continuous monitoring and evaluation of the deviation between the predicted supply capacity and the actual energy supply status, an online correction mechanism is adopted to dynamically correct the energy supply capacity prediction results and output supply capacity prediction results that match the current energy system operation status. S407. The production demand forecast and energy supply forecast results are integrated at the same spatiotemporal granularity, and the final output is the supply and demand balance prediction curve, supply and demand gap or surplus warning, and forecast confidence level for future periods. S408. Execute steps S401 to S407 repeatedly within the preset prediction time domain, and end the prediction loop when the prediction window ends.

[0009] According to the above scheme, step S5, the method for generating scheduling instructions through optimization algorithm, includes the following steps: S501. To meet the needs of coordinated regulation of multi-energy supply and demand in ports, construct an optimization target system oriented towards ensuring uninterrupted production operations, minimizing energy costs, and maximizing the proportion of green energy. S502. In conjunction with the actual operational constraints of the port, construct a production-side, energy-side, and collaborative constraint system. Production-side constraints include operation sequence constraints, task priority constraints, and equipment capacity constraints. Energy-side constraints include supply capacity constraints, remaining energy storage capacity constraints, and energy transmission constraints. Collaborative constraints include supply-demand balance constraints and green energy consumption constraints. S503. In order to generate collaborative scheduling instructions across the energy side and the production side while satisfying the optimization objectives and constraints, a hyperheuristic scheduling method based on reinforcement learning and large language model is adopted. Through adaptive selection and dynamic adjustment of different heuristic operators, multi-objective scheduling solution search and quality improvement are achieved. S504. Design corresponding encoding and decoding methods based on the scheduling characteristics of the port's energy and production sides, reflect the problem characteristics of the actual engineering environment, and use a preset initialization method to initialize the population. S505. By discretizing the system's operational status, a finite state set S is designed, and several scheduling strategies are intelligently generated using a large language model as heuristic operators to form an action set A. S506. Under the condition of meeting the preset triggering conditions, call the large language model to conduct a comprehensive analysis of the phased performance of the set of heuristic operators in the reinforcement learning process, and adaptively adjust the weight of each heuristic operator based on the analysis results. S507. During the scheduling optimization process, the reinforcement learning model is based on the current system state. and the established state-action value function Guided by a periodic large language model The greedy strategy selects a heuristic operator, and all individuals in the population locally adjust the scheduling scheme based on the selected operator; this is guided by a periodic large language model. In the -greedy strategy, when the interval random values ​​in At that time, based on the weight vector output by the large language model After sampling, a heuristic operator is selected, when the random value When selecting the current state Heuristic operator maximizing the state-action value function; guided by a periodic large language model. The -greedy strategy is expressed by the following formula, where Indicates the state Select action Decision-making rules Indicates the probability of exploration. This represents the weight vector output by the large language model. Choose actions based on probability. Indicates the state Choose the action that maximizes the state-action value function from all possible actions. : ; S508. After applying the selected heuristic operator to all individuals in the current population, calculate the immediate reward value based on the performance changes of the individuals before and after adjustment on multiple optimization objectives; and update the state-action value function of the reinforcement learning model according to the immediate reward, while completing the transition from the current decision state to the next state. S509. Using non-dominated sorting and related crowding criteria, individuals with better Pareto performance and diversity are selected from all current individuals to form the next generation population. S5010. Repeat steps S506 to S509 until the preset termination condition is met. After the iteration is completed, output the final scheduling scheme. The termination condition is based on a fixed number of iterations, that is, stopping after reaching the preset maximum number of loops, or it can be based on a time limit, stopping when the running time exceeds the preset duration.

[0010] According to the above scheme, step S508 specifically includes: S508a, the reward function is designed following a goal-oriented approach, employing a linearly weighted immediate reward, wherein... This indicates the percentage change in task delays. This indicates the rate of change in energy costs. This indicates an increase in the rate of green energy utilization. They represent , , The proportion in the reward function, and Population Individual Rewards Based solely on a comparison before and after a single operator execution, without relying on history:

[0011] S508b, Considering that the selected heuristic operator acts on all individuals in the population in the current iteration, where i is the individual index, i=1, 2, ..., NP, the state-action value function... The overall reward value used for the update Defined as all individuals in the current population Sum of instant rewards:

[0012] S508c, upon obtaining the total reward value Then, the current state-action value function is updated using the following formula, where Indicates the learning rate. As a discount factor, Indicates the next state The maximum value among all possible actions in the state-action value function. This means assigning the value on the right to the value on the left: .

[0013] According to the above scheme, the method for coordinated control of various energy units in the port microgrid and hydrogen production system in step S6 includes the following steps: S601. When green electricity is abundant, priority should be given to directly supplying green electricity to the port's sliding contact line, charging piles and shore power system production load. The remaining part should be allocated to port energy storage charging, and the other part should be sent to the port's hydrogen production station to produce green hydrogen. At the same time, the purchase of gray electricity should be suspended to reduce costs. S602. When green electricity is insufficient but energy storage is sufficient, green electricity is given priority to supply high-priority production loads. At the same time, energy storage discharge is activated to supplement the green electricity gap and meet other production loads. If energy storage discharge is still insufficient, a small amount of grey electricity is purchased to supplement the energy. S603. Under peak load conditions, green electricity, energy storage discharge and gray electricity are used simultaneously to ensure peak load, while temporary power restrictions or staggered power use are implemented for non-core equipment. S604. Under off-peak load conditions, green electricity is prioritized to charge energy storage, and the remaining green electricity is used entirely for hydrogen production. At the same time, the purchase of gray electricity is suspended to avoid energy waste. S605. In an emergency, immediately switch to full gray electricity supply, simultaneously trigger fault alarms of energy storage or green energy equipment, and notify maintenance personnel for emergency repairs. S606. In the hydrogen energy supply chain, green hydrogen should be used first, and gray hydrogen should be used as a supplement when green hydrogen is insufficient.

[0014] According to the above scheme, the method for dynamically optimizing production tasks and equipment timing in step S7 includes the following steps: S701. When energy supply is tight, prioritize tasks, prioritize high-efficiency tasks, suspend or postpone low-efficiency tasks, and schedule equipment to operate in staggered shifts to avoid starting multiple high-energy-consuming devices at the same time. Start them in batches at fixed time intervals to smooth out peak electricity or hydrogen consumption. S702. When energy supply is sufficient, scheduling operations are carried out in parallel, multiple high-energy-consuming equipment are started in one go, and loading, unloading and transportation operations are carried out simultaneously to reduce the overall operation time, while preparatory tasks are executed in advance. S703. When the supply of green hydrogen is insufficient, some hydrogen-powered transport vehicles will be temporarily switched to electric vehicles, and hydrogen-powered vehicles will be refueled in different time periods to avoid queuing and congestion at hydrogen refueling stations. S704. When emergency power is interrupted, suspend non-safety-essential equipment, keep only safety-assurance equipment running, and rearrange task priorities. Once power is restored, prioritize the tasks with the highest risk of delay.

[0015] This invention also provides a port multi-energy supply and demand intelligent collaborative management and control system with central decision-making and dual-end execution, comprising: Operating environment module: Based on the actual operation scenario of the port, it is used to build an operating environment that includes port microgrid, green electricity and gray electricity supply, hydrogen production and hydrogen refueling system, and multiple types of production energy loads; Intelligent Collaborative Management and Control Architecture Module: Used to construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends. Energy and production side data collection modules: The port energy management and control intelligent agent and the port production management and control intelligent agent are used to collect energy and production side data respectively, and simultaneously provide them to the multi-energy supply and demand coordination intelligent agent; Predicting future production demand and energy supply module: The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool, predict production demand and energy supply in a specific future period, and output the results according to the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions. The scheduling instruction generation module: Based on the production demand forecast and energy supply forecast, the multi-energy supply and demand coordination intelligent agent generates scheduling instructions through optimization algorithms, aiming to achieve no delays in production operations, the lowest energy consumption cost, and the highest proportion of green energy, combined with the actual operational constraints of the port production system. The instructions are then transmitted to the port energy management intelligent agent and the port production management intelligent agent. Collaborative Control Module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent is used to coordinate the control of various energy units in the port microgrid and hydrogen production system, so as to realize the dynamic optimization and flexible switching of the energy supply structure. Dynamic optimization module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent is used to dynamically optimize production tasks and equipment timing. Under the premise of ensuring key production tasks and safe operation, it improves the port production system's adaptability to energy fluctuations and overall operating efficiency. Correction Module: The port energy management and control intelligent agent and the port production management and control intelligent agent feed back the execution results of scheduling instructions, actual energy supply and consumption data and production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent corrects the prediction model input and prediction results based on the execution deviation and adjusts subsequent scheduling decisions in a synchronous manner.

[0016] The intelligent collaborative management and control method for multi-energy supply and demand in ports, which integrates central decision-making and dual-end execution according to the present invention, has the following beneficial effects: 1. This invention constructs a unified decision-making center with a multi-energy supply and demand coordination intelligent agent as the core, which integrates the energy supply status and production load demand under the same framework for perception and comprehensive analysis. It overcomes the problems of the energy side and production side being independent and the supply and demand information being difficult to coordinate in the existing technology, and improves the overall integrity and effectiveness of supply and demand coordination under the condition of multiple energy and multiple loads coexisting. 2. In the unified decision-making of the multi-energy supply and demand coordination intelligent agent, this invention comprehensively considers production demand, energy supply structure and green energy availability, realizes dynamic optimization and flexible switching of multiple energy sources, reduces production operation delays, lowers energy use costs and improves the level of green energy consumption. 3. In this invention, the multi-energy supply and demand coordination intelligent agent synchronously issues scheduling instructions to the port energy management intelligent agent and the port production management intelligent agent, so that the energy scheduling behavior and the production scheduling behavior are consistent at the time and strategy level, effectively avoiding the problem of asynchronous execution caused by the decentralized generation of instructions, and enhancing the stability of energy supply and production demand matching. 4. In this invention, the multi-energy supply and demand coordination intelligent agent realizes dynamic correction and continuous optimization of supply and demand forecasting and scheduling decisions based on real-time feedback of energy execution results and production execution results, thereby improving the system's adaptability and operational reliability under conditions of energy fluctuations and production uncertainties. Attached Figure Description

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart of the intelligent collaborative management and control method for multi-energy supply and demand in ports, which integrates central decision-making and dual-end execution, according to the present invention. Figure 2 This is a schematic diagram of a port multi-energy supply and demand coordinated operation scenario according to the present invention; Figure 3 This is a diagram of the intelligent collaborative management and control architecture for port multi-energy supply and demand, which integrates central decision-making and dual-end execution in this invention. Figure 4 This is a flowchart of the demand forecasting process of the multi-energy supply and demand coordination intelligent agent of the present invention; Figure 5 This is a flowchart of the hyperheuristic scheduling method based on reinforcement learning and large language models of the present invention. Detailed Implementation

[0018] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] like Figure 1-5 As shown, the intelligent collaborative management and control method for multi-energy supply and demand in ports, which integrates central decision-making and dual-end execution, includes the following steps: S1. Based on the actual operation scenario of the port, construct an operating environment that includes port microgrid, coordinated supply of green and gray electricity, hydrogen production and refueling system, and various types of production energy loads.

[0020] like Figure 2 As shown, constructing the actual operating environment of a port includes the following steps: S101. The port photovoltaic power generation device and wind power generation device will connect the generated green electricity to the microgrid. The port energy storage device is connected to the microgrid and will charge when there is a surplus of green electricity and discharge when there is a shortage of green electricity or during peak load. At the same time, the port microgrid will purchase gray electricity as supplementary energy as needed when local green electricity and energy storage cannot meet the energy demand through connection with the external power grid. S102. The electrical energy output from the microgrid is distributed according to different application scenarios. On the one hand, it provides direct power support to electrically driven equipment such as shore cranes, yard cranes, port mobile machinery, and some horizontal transport vehicles through the port's sliding contact line power supply system and port charging piles. On the other hand, it provides shore power services to berthed ships through the shore power system. In addition, the microgrid also supplies power to the port's hydrogen production station, using green electricity to electrolyze hydrogen, and the resulting hydrogen is stored and used as green hydrogen. S103. In the hydrogen energy supply chain, the port hydrogen production station converts electrical energy into hydrogen energy, and the hydrogen refueling station provides hydrogen refueling services to hydrogen-powered vehicles in the port as needed, using the green hydrogen produced or the gray hydrogen purchased from chemical plants around the port.

[0021] S2. Construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends.

[0022] like Figure 3 As shown, the method for a port multi-energy supply and demand intelligent collaborative management and control architecture that integrates central decision-making and dual-end execution includes the following steps: S201, the multi-energy supply and demand coordination intelligent agent, as the central decision-making unit of the control architecture, is responsible for aggregating and analyzing real-time data from the port energy control intelligent agent and the port production control intelligent agent, completing energy supply and demand forecasting and collaborative decision-making, and dynamically correcting the forecast results and decision-making strategies. S202, the port energy management and control intelligent agent and the port production management and control intelligent agent are located at the two ends of the execution respectively. Through bidirectional information interaction with the multi-energy supply and demand coordination intelligent agent, on the one hand, it synchronizes the real-time status data of the energy side and the production side to the central hub, and on the other hand, it receives the scheduling instructions issued by the central hub and executes the corresponding control operations. The execution results are then fed back to the multi-energy supply and demand coordination intelligent agent in real time.

[0023] S3, the port energy management and control intelligent agent and the port production management and control intelligent agent respectively collect data from the energy side and the production side, and simultaneously transmit this data to the multi-energy supply and demand coordination intelligent agent. The collected energy-side data and production-side data are specifically as follows: Energy-side data includes static data such as microgrid capacity, photovoltaic and wind turbine installed capacity, energy storage rated capacity, and hydrogen production and refueling station capacity limits; dynamic data such as real-time photovoltaic and wind turbine output, remaining energy storage capacity, real-time hydrogen production and refueling output, and external purchase prices of grey electricity and grey hydrogen; and external data such as weather forecast data, peak and off-peak electricity price periods, and early warning coordination for hydrogen supply stability. Production-side data includes static data such as ship arrival schedules, cargo types and throughput, equipment lists, and equipment rated energy consumption or hydrogen consumption; dynamic data such as current operation progress, real-time equipment operating status, and energy demand for sudden operations; and historical data such as operation sequence, peak energy demand, and operation efficiency fluctuations within a custom time period.

[0024] S4. The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool. It predicts production demand and energy supply in a specific future period and outputs the results in the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions.

[0025] like Figure 4 As shown, the method for energy supply and demand forecasting by a multi-energy supply and demand coordination intelligent agent includes the following steps: S401, a multi-energy supply and demand coordination intelligent agent integrates energy-side and production-side data to build a full-scale predictive data pool; S402. Determine whether a fixed time interval has been reached or whether an emergency has occurred. Emergency conditions include, but are not limited to, energy supply failure, sudden operational needs, and sudden changes in weather conditions. If yes, execute steps S43 to S46. If no, execute only steps S44 and S46 and not steps S43 and S45. S403. A hybrid forecasting model, consisting of a time-series forecasting sub-model and a large language model, is used to forecast basic energy demand on the production side, specifically: S403a, the time series prediction sub-model adopts a Long Short-Term Memory (LSTM) network and uses a sliding time window method. It takes equipment operation data and energy consumption data from multiple consecutive historical moments as input vectors, models the long-term and short-term energy consumption change characteristics through recurrent memory units, and outputs the energy demand prediction results for a specific future period, thus depicting the time correlation and evolution trend of energy consumption demand of multiple equipment and processes in port production operations. S403b, the Large Language Model (LLM), takes the production-side energy demand forecast results output by the time series forecast sub-model, port equipment operation constraints, and operational experience as inputs. It infers and judges the consistency between the forecast results and the equipment operation constraints and operational experience, corrects the consistency of forecast results that do not meet the equipment operation conditions or operation organization requirements, and outputs the corrected production-side energy demand forecast results. Among them, port equipment operation constraints and operational experience include, but are not limited to, physical limit constraints of operating equipment, operating and maintenance constraints of operating equipment, operation organization and scheduling experience, and anomaly and risk avoidance experience. S403c, the production-side energy demand forecast results include, but are not limited to, total electricity demand, total hydrogen demand, energy demand sequence by equipment, peak demand time and peak duration. S404. An online incremental learning algorithm is adopted, which takes real-time port production operation data and actual energy consumption data for the corresponding period as input. Through continuous monitoring and learning updates of the deviation between predicted energy demand and actual energy consumption, the output of the production-side energy demand prediction model is dynamically corrected, and the output energy demand prediction result that matches the current operating status is given. S405. An energy supply forecasting model, composed of a machine learning forecasting model and a large language model, is used to forecast the basic energy supply capacity, specifically as follows: S405a The machine learning prediction model uses a random forest combined with time series decomposition method. It takes the historical output data and operating status data of wind power, photovoltaic, energy storage, grid power purchase and hydrogen production systems within a continuous time window as input. First, the historical output data is separated and modeled by the time series decomposition method to separate the trend term, periodic term and random fluctuation term. Then, the random forest model is used to learn the mapping relationship between each component and the relevant operating characteristics, and output the preliminary prediction results of the energy supply capacity in the future prediction period. S405b and the Large Language Model (LLM) take the energy supply capacity prediction results output by the machine learning prediction model, energy system operation constraints, and energy supply experience as input. It comprehensively reasones and judges the conformity and rationality between the prediction results and the energy system operation constraints and energy supply experience. It verifies and corrects prediction results that are unexecutable or have potential energy supply risks, and outputs energy supply capacity prediction results that have been verified for rationality. Among them, the port energy system operation constraints and energy supply experience include, but are not limited to, physical and safety constraints of energy equipment, multi-energy switching constraints, energy allocation experience, and energy supply anomaly and risk avoidance experience. S405c, The predicted results of the energy supply capacity include, but are not limited to, the total power supply capacity of green electricity, gray electricity and energy storage in different time periods, the total supply capacity of green hydrogen and gray hydrogen in different time periods, and the prediction of supply capacity gap or surplus.

[0026] S406. Combining real-time port energy system operation data and actual energy supply data for the corresponding period, through continuous monitoring and evaluation of the deviation between the predicted supply capacity and the actual energy supply status, an online correction mechanism is adopted to dynamically correct the energy supply capacity prediction results and output supply capacity prediction results that match the current energy system operation status. S407. The production demand forecast and energy supply forecast results are integrated at the same spatiotemporal granularity, and the final output is the supply and demand balance prediction curve, supply and demand gap or surplus warning, and forecast confidence level for future periods. S408. Execute steps S401 to S407 repeatedly within the preset prediction time domain, and end the prediction loop when the prediction window ends.

[0027] S5. Based on the production demand forecast and energy supply forecast results obtained in step S4, the multi-energy supply and demand coordination intelligent agent aims to ensure uninterrupted production operations, minimize energy consumption costs, and maximize the proportion of green energy. Combining the actual operational constraints of the port production system, it generates scheduling instructions through optimization algorithms and transmits the instructions to the port energy management intelligent agent and the port production management intelligent agent.

[0028] like Figure 5 As shown, the specific steps by which the multi-energy supply and demand coordination agent generates scheduling instructions based on production demand forecasts and energy supply forecasts through an optimization algorithm are as follows: S501. To meet the needs of coordinated regulation of multi-energy supply and demand in ports, construct an optimization target system oriented towards ensuring uninterrupted production operations, minimizing energy costs, and maximizing the proportion of green energy. S502. In conjunction with the actual operational constraints of the port, construct a production-side, energy-side, and collaborative constraint system. Production-side constraints include operation sequence constraints, task priority constraints, and equipment capacity constraints. Energy-side constraints include supply capacity constraints, remaining energy storage capacity constraints, and energy transmission constraints. Collaborative constraints include supply and demand balance constraints and green energy consumption constraints. S503. In order to generate collaborative scheduling instructions across the energy side and the production side while satisfying the optimization objectives and constraints, a hyperheuristic scheduling method based on reinforcement learning and large language model (LLM) is adopted. Through adaptive selection and dynamic adjustment of different heuristic operators, efficient search and quality improvement of multi-objective scheduling solutions are achieved. S504. Design corresponding encoding and decoding methods for the scheduling characteristics of the port's energy and production sides to reflect the problem characteristics of the actual engineering environment, and use a preset initialization method to initialize the population. S505. By discretizing the system's operational status, a finite state set is designed. Furthermore, it combines a large language model (LLM) to intelligently generate several scheduling strategies as heuristic operators to form an action set. Specifically: S505a defines the state of reinforcement learning as a triplet. Production load status Reflects the current operational pressure level, with 0 indicating low load, 1 indicating medium load, and 2 indicating high load; energy supply status. This reflects the current relative tension of available energy compared to demand; 0 indicates energy shortage, 1 indicates energy balance, and 2 indicates energy surplus; equipment utilization status. This reflects the utilization level of critical equipment, with 0 indicating low utilization, 1 indicating moderate utilization, and 2 indicating high utilization; the state set is represented as... ; S505b, In a finite set of actions In this context, each action represents a scheduling heuristic operator. These heuristic operators can be autonomously generated and appropriately adjusted based on prior knowledge by a large language model (LLM), including but not limited to: High-priority tasks are moved forward, while low-priority tasks are postponed; Equipment is started off during off-peak hours; The order of tasks has been partially rearranged; Green electricity should be prioritized for critical operations; Reduce non-critical workloads.

[0029] S506. Under the condition of meeting the preset triggering conditions, call the Large Language Model (LLM) to comprehensively analyze the stage performance of the set of heuristic operators in the reinforcement learning process, and adaptively adjust the weight of each heuristic operator based on the analysis results. S506a. In the reinforcement learning process, at the... After the next iteration, the most recent information needs to be provided to the large language model LLM. The statistical information of each operator's call frequency, average reward, and reward variance during each step is analyzed and inferred by the Large Language Model (LLM) to output suggestions for adjusting operator weights. , This indicates that heuristic operators are disabled in this phase. Except for fixed periods In addition, if recently If the average reward change in the next iteration is lower than the threshold, or if the same heuristic operator is selected multiple times consecutively but the reward is almost 0 or lower than 0, the large language model LLM can be called to adjust the operator weights. The S506b Large Language Model (LLM) hinting project is as follows: You are an intelligent analysis module used for scheduling algorithms; your responsibility is to provide weight adjustment suggestions for each heuristic operator based on the stage statistics provided during reinforcement learning; you need to follow these principles: encourage heuristic operators with high average rewards and good stability; suppress or disable heuristic operators with low average rewards, large reward variance, or continuous ineffectiveness; avoid excessive weight concentration to prevent premature convergence in the search; the weight value range is... You can set the weight of operators with extremely poor performance to 0; your output is only used to guide the subsequent operator selection probability adjustment; recently The heuristic operator statistics for this iteration are as follows: Total number of operators: Each heuristic operator call frequency Each heuristic operator average reward Each heuristic operator reward variance Please output a string of length [length missing]. Operator weight vector: The output needs to meet the following constraints: ; This indicates that heuristic operators are disabled at the current stage. The weights should reflect the relative performance of each operator; only output the weight vector, without any additional explanation or text description.

[0030] S507. During the scheduling optimization process, the reinforcement learning model is based on the current system state. and the established state-action value function Guided by a periodic large language model The greedy strategy selects a heuristic operator, and all individuals in the population locally adjust the scheduling scheme based on the selected operator; this is guided by a periodic large language model. In the -greedy strategy, when the interval random values ​​in At that time, based on the weight vector output by the large language model After sampling, a heuristic operator is selected, when the random value When selecting the current state Heuristic operator maximizing the state-action value function; guided by a periodic large language model. The -greedy strategy is expressed by the following formula, where Indicates the state Select action Decision-making rules Indicates the probability of exploration. This represents the weight vector output by the large language model. Choose actions based on probability. Indicates the state Choose the action that maximizes the state-action value function from all possible actions. : .

[0031] S508. After applying the selected heuristic operator to all individuals in the current population, calculate the immediate reward value based on the performance changes of individuals before and after adjustment across multiple optimization objectives; and update the state-action value function of the reinforcement learning model according to the immediate reward, while simultaneously completing the transition from the current decision state to the next state; specifically: S508a, the reward function is designed following a goal-oriented approach, employing a linearly weighted immediate reward, wherein... This indicates the percentage change in task delays. This indicates the rate of change in energy costs. This indicates an increase in the rate of green energy utilization. They represent , , The proportion in the reward function, and Population Individual Rewards Based solely on a comparison before and after a single operator execution, without relying on history:

[0032] S508b, Considering that the selected heuristic operator acts on all individuals in the population in the current iteration, where i is the individual index, i=1, 2, ..., NP, the state-action value function... The overall reward value used for the update Defined as all individuals in the current population Sum of instant rewards:

[0033] S508c, upon obtaining the total reward value Then, the current state-action value function is updated using the following formula, where Indicates the learning rate. As a discount factor, Indicates the next state The maximum value among all possible actions in the state-action value function. This means assigning the value on the right to the value on the left: .

[0034] S509. Using non-dominated sorting and related crowding criteria, individuals with better Pareto performance and diversity are selected from all current individuals to form the next generation population. S5010. Repeat steps S506 to S509 until the preset termination condition is met. After the iteration is completed, output the final scheduling scheme. The termination condition can be based on a fixed number of iterations, that is, stop after reaching the preset maximum number of loops, or it can be based on a time limit, stop when the running time exceeds the preset duration.

[0035] S6. Under the constraints of the collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent coordinates the control of various energy units in the port microgrid and hydrogen production system, thereby achieving dynamic optimization and flexible switching of the energy supply structure.

[0036] Under the constraints of the collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management intelligent agent performs the following specific operations to coordinate and control various energy units within the port microgrid and hydrogen production system: S601. When green electricity is abundant, priority should be given to directly supplying green electricity to the port's sliding contact line, charging piles, shore power system and other production loads. The remaining part should be allocated to port energy storage charging, and the other part should be sent to the port's hydrogen production station to produce green hydrogen. At the same time, the purchase of gray electricity should be suspended to reduce costs. S602. When green electricity is insufficient but energy storage is sufficient, green electricity is given priority to supply high-priority production loads. At the same time, energy storage discharge is activated to supplement the green electricity gap and meet other production loads. If energy storage discharge is still insufficient, a small amount of grey electricity is purchased to supplement the energy. S603. Under peak load conditions, green electricity, energy storage discharge and gray electricity are used simultaneously to ensure peak load, while temporary power restrictions or staggered power use are implemented for non-core equipment. S604. Under off-peak load conditions, green electricity is prioritized to charge energy storage, and the remaining green electricity is used entirely for hydrogen production. At the same time, the purchase of gray electricity is suspended to avoid energy waste. S605. In an emergency, immediately switch to full gray electricity supply, simultaneously trigger fault alarms of energy storage or green energy equipment, and notify maintenance personnel for emergency repairs. S606. In the hydrogen energy supply chain, green hydrogen should be used first, and gray hydrogen should be used as a supplement when green hydrogen is insufficient.

[0037] S7. Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent dynamically optimizes production tasks and equipment timing, thereby improving the port production system's adaptability to energy fluctuations and overall operational efficiency while ensuring key production tasks and safe operation.

[0038] Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production management and control intelligent agent dynamically optimizes production tasks and equipment timing as follows: S701. When energy supply is tight, prioritize tasks, prioritize high-efficiency tasks, suspend or postpone low-efficiency tasks, and schedule equipment to operate in staggered shifts to avoid starting multiple high-energy-consuming devices at the same time. Start them in batches at fixed time intervals to smooth out peak electricity or hydrogen consumption. S702. When energy supply is sufficient, scheduling operations are carried out in parallel, multiple high-energy-consuming equipment are started in one go, and loading, unloading and transportation operations are carried out simultaneously to reduce the overall operation time. At the same time, preparatory tasks can be executed in advance to improve the efficiency of subsequent task connection. S703. When the supply of green hydrogen is insufficient, some hydrogen-powered transport vehicles will be temporarily switched to electric vehicles, and hydrogen-powered vehicles will be refueled in different time periods to avoid queuing and congestion at hydrogen refueling stations. S704. When emergency power is interrupted, suspend non-safety-essential equipment, keep only safety-assurance equipment running, and rearrange task priorities. Once power is restored, prioritize the tasks with the highest risk of delay.

[0039] S8. The port energy management and control intelligent agent and the port production management and control intelligent agent will feed back the execution results of the scheduling instructions, the actual energy supply and consumption data and the production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent will correct the prediction model input and prediction results based on the execution deviation, and adjust the subsequent scheduling decisions in a synchronous manner to realize the closed-loop collaborative control of the port energy system and production system of rolling prediction - dynamic decision-making - feedback correction.

[0040] This invention also provides a port multi-energy supply and demand intelligent collaborative management and control system with central decision-making and dual-end execution, comprising: Operating environment module: Based on the actual operation scenario of the port, it is used to build an operating environment that includes port microgrid, green electricity and gray electricity supply, hydrogen production and hydrogen refueling system, and multiple types of production energy loads; Intelligent Collaborative Management and Control Architecture Module: Used to construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends. Energy and production side data collection modules: The port energy management and control intelligent agent and the port production management and control intelligent agent are used to collect energy and production side data respectively, and simultaneously provide them to the multi-energy supply and demand coordination intelligent agent; Predicting future production demand and energy supply module: The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool, predict production demand and energy supply in a specific future period, and output the results according to the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions. The scheduling instruction generation module: Based on the production demand forecast and energy supply forecast, the multi-energy supply and demand coordination intelligent agent generates scheduling instructions through optimization algorithms, aiming to achieve no delays in production operations, the lowest energy consumption cost, and the highest proportion of green energy, combined with the actual operational constraints of the port production system. The instructions are then transmitted to the port energy management intelligent agent and the port production management intelligent agent. Collaborative Control Module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent is used to coordinate the control of various energy units in the port microgrid and hydrogen production system, so as to realize the dynamic optimization and flexible switching of the energy supply structure. Dynamic optimization module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent is used to dynamically optimize production tasks and equipment timing. Under the premise of ensuring key production tasks and safe operation, it improves the port production system's adaptability to energy fluctuations and overall operating efficiency. Correction Module: The port energy management and control intelligent agent and the port production management and control intelligent agent feed back the execution results of scheduling instructions, actual energy supply and consumption data and production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent corrects the prediction model input and prediction results based on the execution deviation, and adjusts subsequent scheduling decisions in a synchronous manner, so as to realize the closed-loop collaborative control of rolling prediction, dynamic decision-making and feedback correction of the port energy system and production system.

[0041] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for intelligent collaborative management and control of multi-energy supply and demand in ports, characterized by: (1) Central decision-making and dual-end execution Includes the following steps: S1. Based on the actual operation scenario of the port, construct an operating environment that includes port microgrid, coordinated supply of green electricity and gray electricity, hydrogen production and hydrogen refueling system, and multiple types of production energy loads; S2. Construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends. S3, the port energy management and control intelligent agent and the port production management and control intelligent agent collect data from the energy side and the production side respectively, and simultaneously provide the data to the multi-energy supply and demand coordination intelligent agent; S4. The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool, predicts production demand and energy supply in a specific future period, and outputs the results in the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions. S5. Based on the production demand forecast and energy supply forecast obtained in step S4, the multi-energy supply and demand coordination intelligent agent generates scheduling instructions through optimization algorithms and transmits the instructions to the port energy management intelligent agent and the port production management intelligent agent. S6. Under the constraints of the collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent performs collaborative control of various energy units in the port microgrid and hydrogen production system. S7. Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent dynamically optimizes production tasks and equipment timing. S8, the port energy management and control intelligent agent and the port production management and control intelligent agent will feed back the execution results of the scheduling instructions, the actual energy supply and consumption data and the production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent will correct the prediction results based on the execution deviation and adjust the subsequent scheduling decisions in a synchronous manner.

2. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S1, constructing the actual port operation scenario includes the following steps: S101. The port photovoltaic power generation device and wind power generation device will connect the generated green electricity to the microgrid. The port energy storage device is connected to the microgrid and will charge when there is a surplus of green electricity and discharge when there is a shortage of green electricity or during peak load. At the same time, the port microgrid will purchase gray electricity as supplementary energy as needed when local green electricity and energy storage cannot meet the energy demand through connection with the external power grid. S102. The electrical energy output by the microgrid is distributed according to three application scenarios, specifically: the first application scenario provides direct power support for the electric drive equipment of shore cranes, yard cranes, port mobile machinery, and some horizontal transport vehicles through the port sliding contact line power supply system and port charging piles; the second application scenario provides shore power services to docked ships through the shore power system; and the third application scenario supplies power to the port hydrogen production station, uses green electricity to electrolyze hydrogen, and stores and uses the resulting hydrogen as green hydrogen. S103. In the hydrogen energy supply chain, the port hydrogen production station converts electrical energy into hydrogen energy, and the hydrogen refueling station provides hydrogen refueling services to hydrogen-powered vehicles in the port as needed, using the green hydrogen produced or the gray hydrogen purchased from chemical plants around the port.

3. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S2, the method for constructing the port's multi-energy supply and demand intelligent collaborative management and control architecture includes the following steps: S201, the multi-energy supply and demand coordination intelligent agent, as the central decision-making unit of the control architecture, is responsible for aggregating and analyzing real-time data from the port energy control intelligent agent and the port production control intelligent agent, completing energy supply and demand forecasting and collaborative decision-making, and dynamically correcting the forecast results and decision-making strategies. S202, the port energy management and control intelligent agent and the port production management and control intelligent agent are located at the two ends of the execution respectively. Through bidirectional information interaction with the multi-energy supply and demand coordination intelligent agent, on the one hand, it synchronizes the real-time status data of the energy side and the production side to the central hub, and on the other hand, it receives the scheduling instructions issued by the central hub and executes the corresponding control operations. The execution results are then fed back to the multi-energy supply and demand coordination intelligent agent in real time.

4. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S3 Energy-side data includes: microgrid capacity, photovoltaic and wind turbine installed capacity, rated energy storage capacity, and upper limit of hydrogen production and refueling station capacity; real-time output of photovoltaic and wind turbines, remaining energy storage capacity, real-time hydrogen production and refueling output, and external purchase prices of grey electricity and grey hydrogen; meteorological forecast data, peak and off-peak electricity price periods, and coordinated data on hydrogen supply stability early warning. Production-side data includes: ship arrival schedules, cargo types and throughput, list of operating equipment, rated energy consumption or hydrogen consumption of equipment; current operation progress, real-time operating status of equipment, and energy demand for sudden operations; operation sequence, peak energy consumption, and operation efficiency fluctuation data within a custom time period.

5. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S4, the method for predicting production demand and energy supply within a specific future period includes the following steps: S401, a multi-energy supply and demand coordination intelligent agent integrates energy-side and production-side data to build a full-scale predictive data pool; S402. Determine whether a fixed time interval has been reached or whether an emergency has occurred. Emergency conditions include energy supply failure, sudden operational demand, and sudden changes in weather conditions. If yes, execute steps S403 to S406. If no, execute only steps S404 and S406. S403. A hybrid forecasting model consisting of a time-series forecasting sub-model and a large language model is used to forecast basic energy demand on the production side. S404. An online incremental learning algorithm is adopted, which takes real-time port production operation data and actual energy consumption data for the corresponding period as input. Through continuous monitoring and learning updates of the deviation between predicted energy demand and actual energy consumption, the output of the production-side energy demand prediction model is dynamically corrected, and the output energy demand prediction result that matches the current operating status is given. S405. An energy supply prediction model composed of a machine learning prediction model and a large language model is used to predict the basic energy supply capacity. S406. Combining real-time port energy system operation data and actual energy supply data for the corresponding period, through continuous monitoring and evaluation of the deviation between the predicted supply capacity and the actual energy supply status, an online correction mechanism is adopted to dynamically correct the energy supply capacity prediction results and output supply capacity prediction results that match the current energy system operation status. S407. The production demand forecast and energy supply forecast results are integrated at the same spatiotemporal granularity, and the final output is the supply and demand balance prediction curve, supply and demand gap or surplus warning, and forecast confidence level for future periods. S408. Execute steps S401 to S407 repeatedly within the preset prediction time domain, and end the prediction loop when the prediction window ends.

6. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S5, the method for generating scheduling instructions through optimization algorithms includes the following steps: S501. To meet the needs of coordinated regulation of multi-energy supply and demand in ports, construct an optimization target system oriented towards ensuring uninterrupted production operations, minimizing energy costs, and maximizing the proportion of green energy. S502. In conjunction with the actual operational constraints of the port, construct a production-side, energy-side, and collaborative constraint system. Production-side constraints include operation sequence constraints, task priority constraints, and equipment capacity constraints. Energy-side constraints include supply capacity constraints, remaining energy storage capacity constraints, and energy transmission constraints. Collaborative constraints include supply-demand balance constraints and green energy consumption constraints. S503. In order to generate collaborative scheduling instructions across the energy side and the production side while satisfying the optimization objectives and constraints, a hyperheuristic scheduling method based on reinforcement learning and large language model is adopted. Through adaptive selection and dynamic adjustment of different heuristic operators, multi-objective scheduling solution search and quality improvement are achieved. S504. Design corresponding encoding and decoding methods based on the scheduling characteristics of the port's energy and production sides, reflect the problem characteristics of the actual engineering environment, and use a preset initialization method to initialize the population. S505. By discretizing the system's operational status, a finite state set S is designed, and several scheduling strategies are intelligently generated using a large language model as heuristic operators to form an action set A. S506. Under the condition of meeting the preset triggering conditions, call the large language model to conduct a comprehensive analysis of the phased performance of the set of heuristic operators in the reinforcement learning process, and adaptively adjust the weight of each heuristic operator based on the analysis results. S507. During the scheduling optimization process, the reinforcement learning model is based on the current system state. and the established state-action value function Guided by a periodic large language model The greedy strategy selects a heuristic operator, and all individuals in the population locally adjust the scheduling scheme based on the selected operator; this is guided by a periodic large language model. In the -greedy strategy, when the interval random values ​​in At that time, based on the weight vector output by the large language model After sampling, a heuristic operator is selected, when the random value When selecting the current state Heuristic operator maximizing the state-action value function; guided by a periodic large language model. The -greedy strategy is expressed by the following formula, where Indicates the state Select action Decision-making rules Indicates the probability of exploration. This represents the weight vector output by the large language model. Choose actions based on probability. Indicates the state Choose the action that maximizes the state-action value function from all possible actions. : ； S508. After applying the selected heuristic operator to all individuals in the current population, calculate the immediate reward value based on the performance changes of the individuals before and after adjustment on multiple optimization objectives; and update the state-action value function of the reinforcement learning model according to the immediate reward, while completing the transition from the current decision state to the next state. S509. Using non-dominated sorting and related crowding criteria, individuals with better Pareto performance and diversity are selected from all current individuals to form the next generation population. S5010. Repeat steps S506 to S509 until the preset termination condition is met. After the iteration is completed, output the final scheduling scheme. The termination condition is based on a fixed number of iterations, that is, stopping after reaching the preset maximum number of loops, or it can be based on a time limit, stopping when the running time exceeds the preset duration.

7. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 6, characterized in that, In step S508, specifically: S508a, the reward function is designed following a goal-oriented approach, employing a linearly weighted immediate reward, wherein... This indicates the percentage change in task delays. This indicates the rate of change in energy costs. This indicates an increase in the rate of green energy utilization. They represent , , The proportion in the reward function, and Population Individual Rewards Based solely on a comparison before and after a single operator execution, without relying on history: S508b, Considering that the selected heuristic operator acts on all individuals in the population in the current iteration, where i is the individual index, i=1, 2, ..., NP, the state-action value function... The overall reward value used for the update Defined as all individuals in the current population Sum of instant rewards: S508c, upon obtaining the total reward value Then, the current state-action value function is updated using the following formula, where Indicates the learning rate. As a discount factor, Indicates the next state The maximum value among all possible actions in the state-action value function. This means assigning the value on the right to the value on the left: 。 8. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S6, the method for coordinated control of various energy units within the port microgrid and hydrogen production system includes the following steps: S601. When green electricity is abundant, priority should be given to directly supplying green electricity to the port's sliding contact line, charging piles and shore power system production load. The remaining part should be allocated to port energy storage charging, and the other part should be sent to the port's hydrogen production station to produce green hydrogen. At the same time, the purchase of gray electricity should be suspended to reduce costs. S602. When green electricity is insufficient but energy storage is sufficient, green electricity is given priority to supply high-priority production loads. At the same time, energy storage discharge is activated to supplement the green electricity gap and meet other production loads. If energy storage discharge is still insufficient, a small amount of grey electricity is purchased to supplement the energy. S603. Under peak load conditions, green electricity, energy storage discharge and gray electricity are used simultaneously to ensure peak load, while temporary power restrictions or staggered power use are implemented for non-core equipment. S604. Under off-peak load conditions, green electricity is prioritized to charge energy storage, and the remaining green electricity is used entirely for hydrogen production. At the same time, the purchase of gray electricity is suspended to avoid energy waste. S605. In an emergency, immediately switch to full gray electricity supply, simultaneously trigger fault alarms of energy storage or green energy equipment, and notify maintenance personnel for emergency repairs. S606. In the hydrogen energy supply chain, green hydrogen should be used first, and gray hydrogen should be used as a supplement when green hydrogen is insufficient.

9. The intelligent collaborative management and control method for multi-energy supply and demand in ports with central decision-making and dual-end execution as described in claim 1, characterized in that, In step S7, the method for dynamically optimizing production tasks and equipment timing includes the following steps: S701. When energy supply is tight, prioritize tasks, prioritize high-efficiency tasks, suspend or postpone low-efficiency tasks, and schedule equipment to operate in staggered shifts to avoid starting multiple high-energy-consuming devices at the same time. Start them in batches at fixed time intervals to smooth out peak electricity or hydrogen consumption. S702. When energy supply is sufficient, scheduling operations are carried out in parallel, multiple high-energy-consuming equipment are started in one go, and loading, unloading and transportation operations are carried out simultaneously to reduce the overall operation time, while preparatory tasks are executed in advance. S703. When the supply of green hydrogen is insufficient, some hydrogen-powered transport vehicles will be temporarily switched to electric vehicles, and hydrogen-powered vehicles will be refueled in different time periods to avoid queuing and congestion at hydrogen refueling stations. S704. When emergency power is interrupted, suspend non-safety-essential equipment, keep only safety-assurance equipment running, and rearrange task priorities. Once power is restored, prioritize the tasks with the highest risk of delay.

10. A port multi-energy supply and demand intelligent collaborative management and control system with central decision-making and dual-end execution, characterized in that, include: Operating environment module: Based on the actual operation scenario of the port, it is used to build an operating environment that includes port microgrid, green electricity and gray electricity supply, hydrogen production and hydrogen refueling system, and multiple types of production energy loads; Intelligent Collaborative Management and Control Architecture Module: Used to construct a port multi-energy supply and demand intelligent collaborative management and control architecture with the multi-energy supply and demand coordination intelligent agent as the core decision-making center and the port energy management and control intelligent agent and the port production management and control intelligent agent as the execution ends. Data collection modules for the energy and production sides: The port energy management and control intelligent agent and the port production management and control intelligent agent are used to collect data from the energy side and the production side respectively, and simultaneously provide the data to the multi-energy supply and demand coordination intelligent agent. Predicting future production demand and energy supply module: The multi-energy supply and demand coordination intelligent agent integrates real-time data from multiple sources from the port energy management intelligent agent and the port production management intelligent agent to build a predictive data pool, predict production demand and energy supply in a specific future period, and output the results according to the same spatiotemporal granularity to provide a basis for subsequent scheduling decisions. The scheduling instruction generation module: Based on the production demand forecast and energy supply forecast, the multi-energy supply and demand coordination intelligent agent generates scheduling instructions through optimization algorithms, aiming to achieve no delays in production operations, the lowest energy consumption cost, and the highest proportion of green energy, combined with the actual operational constraints of the port production system. The instructions are then transmitted to the port energy management intelligent agent and the port production management intelligent agent. Collaborative Control Module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port energy management and control intelligent agent is used to coordinate the control of various energy units in the port microgrid and hydrogen production system, so as to realize the dynamic optimization and flexible switching of the energy supply structure. Dynamic optimization module: Under the constraints of collaborative scheduling instructions issued by the multi-energy supply and demand coordination intelligent agent, the port production control intelligent agent is used to dynamically optimize production tasks and equipment timing. Under the premise of ensuring key production tasks and safe operation, it improves the port production system's adaptability to energy fluctuations and overall operating efficiency. Correction Module: The port energy management and control intelligent agent and the port production management and control intelligent agent feed back the execution results of scheduling instructions, actual energy supply and consumption data and production status to the multi-energy supply and demand coordination intelligent agent. The multi-energy supply and demand coordination intelligent agent corrects the prediction model input and prediction results based on the execution deviation and adjusts subsequent scheduling decisions in a synchronous manner.

Images