Port energy dynamic optimization method and system based on digital twinning

Abstract

The invention provides a port energy dynamic optimization method and system based on digital twinning, and relates to the technical field of port energy optimization. According to the method, the full-link closed-loop control of the port energy data is realized by constructing a physical-edge-virtual three-layer digital twin architecture, and the utilization rate of renewable energy sources is improved and the carbon emission is reduced through real-time data synchronization and multi-time scale optimization strategies. Through combination of edge layer preprocessing and a virtual layer physical constraint model, the problem that dynamic response is delayed and a physical rule is disjointed in a traditional method is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of port energy optimization, and in particular to a port energy dynamic optimization method and system based on digital twins. Background Art

[0002] As the core carrier of international trade, the global maritime industry undertakes global trade transportation, but its energy system is highly dependent on fossil fuels. As a key logistics node, ports have strong dynamic fluctuations in energy demand and need to coordinate multiple types of clean energy such as wind energy, solar energy, and hydrogen energy. Traditional energy management methods have significant defects: First, static models are disconnected from dynamic demand: the existing system relies on offline optimization and fixed models, which makes it difficult to respond to fluctuations in renewable energy and sudden load changes in ship operations in real time, resulting in energy waste; second, the lack of carbon tracking mechanism: traditional methods lack refined monitoring of carbon emissions and cannot link ship operations, energy procurement and carbon quotas; third, multi-time scale control is split: millisecond-level power regulation (such as crane start and stop) and hour-level energy scheduling (such as hydrogen energy storage) lack coordination, resulting in grid frequency instability and reduced energy storage life; fourth, data heterogeneity hinders optimization: port energy equipment (shore power, energy storage, photovoltaic) data formats and sampling rates vary greatly, and edge computing capabilities are insufficient, resulting in low prediction accuracy of cloud models.

[0003] Although digital twin technology has been applied in the industrial field, it has not yet solved core problems such as multi-energy flow coupling optimization, real-time tracking of carbon flow, and cross-level control synchronization in port scenarios. Summary of the invention

[0004] In view of the above problems existing in the prior art, the first aspect of the present invention proposes a port energy dynamic optimization method based on digital twin, comprising: Step S1, constructing a multi-level digital twin model of the port energy system, the multi-level digital twin model including a physical layer, an edge layer and a virtual layer; Step S2, collecting real-time operation data of the port energy equipment through the sensor network in the physical layer, preprocessing the real-time operation data in the edge layer, generating a compressed dynamic data stream, and synchronizing the dynamic data stream to the virtual layer; Step S3, in the virtual layer, the physical information neural network is trained based on the dynamic data flow and the preset physical constraint equations to generate a multi-energy flow prediction model, and a multi-time scale energy scheduling strategy is generated by combining the real-time carbon signal data through a quantum-inspired optimization algorithm; Step S4, decomposing the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmitting them back to the energy router of the physical layer through the edge layer; Step S5, executing dynamic adjustment instructions through the energy router to adjust the operating parameters of the port energy equipment, and allocating renewable energy and energy storage resources based on periodic scheduling instructions.

[0005] In combination with the first aspect, in some implementations of the first aspect, in step S3, the specific steps of training the physical information neural network include: Step S31, extracting energy equipment state parameters from the dynamic data stream as input; Step S32, converting the thermodynamic partial differential equation and the power network conservation equation in the physical constraint equation into residual terms, and performing weighted fusion with the output of the physical information neural network to generate a constraint optimization target; Step S33, updating the weight parameters of the physical information neural network through the distributed nodes in the joint learning algorithm to minimize the constraint optimization target and generate a multi-energy flow prediction model.

[0006] In combination with the first aspect, in some implementations of the first aspect, in step S3, the specific steps of generating the energy scheduling strategy include: Step S34, inputting the energy demand forecast data and real-time carbon signal data output by the multi-energy flow forecasting model into the quantum-inspired optimization algorithm; Step S35, decomposing the energy scheduling problem into a discrete variable optimization sub-problem and a continuous variable optimization sub-problem through a hybrid computing unit in a quantum-inspired optimization algorithm; Step S36, using a quantum annealing algorithm to generate discrete decision parameters for the discrete variable optimization subproblem, and using a gradient descent algorithm to generate continuous decision parameters for the continuous variable optimization subproblem; Step S37, combining discrete decision parameters and continuous decision parameters to generate a multi-time scale energy scheduling strategy.

[0007] In combination with the first aspect, in some implementations of the first aspect, the discrete variable optimization subproblem includes the start and stop status of the energy storage device and the switching instructions of the energy router, and the continuous variable optimization subproblem includes the energy storage charging and discharging rate and the energy purchase amount.

[0008] In combination with the first aspect, in some implementations of the first aspect, step S4 includes: Step S41, extracting voltage frequency deviation and instantaneous power change from the energy dispatch strategy and generating a dynamic adjustment instruction; Step S42, extracting the carbon quota allocation plan and energy procurement plan from the energy scheduling strategy, and generating a periodic scheduling instruction; Step S43, aligning the time bases of the dynamic adjustment instructions and the periodic scheduling instructions through the timestamp synchronization module, generating a time-consistent scheduling instruction set, and transmitting it back to the energy router of the physical layer through the edge layer.

[0009] In combination with the first aspect, in some implementations of the first aspect, step S5 includes: Step S51, inputting the dynamic adjustment instruction into the power electronic interface of the energy router to generate a voltage regulation signal and an energy storage charging and discharging control signal; Step S52, inputting the periodic scheduling instruction into the blockchain verification module of the energy router to generate energy transaction records and carbon emission audit data; Step S53, using smart contracts to compare the energy transaction records with the carbon emission audit data for consistency, generate energy allocation instructions, and allocate renewable energy and energy storage resources based on the energy allocation instructions.

[0010] In combination with the first aspect, in some implementations of the first aspect, the blockchain verification module connects the port's internal energy ledger and the external carbon trading platform through a cross-chain protocol, and synchronizes energy trading records and carbon emission audit data to the external carbon trading platform.

[0011] In combination with the first aspect, in some implementations of the first aspect, in step S2, the preprocessing includes: Step S21, filtering the real-time operation data through the abnormality detection unit to generate a filtered data sequence; Step S22, synchronizing the filtered data sequence with a preset clock source through a timestamp alignment module to generate a data stream with a unified time; Step S23, performing feature extraction and dimensionality reduction on the time-uniform data stream through a lightweight convolutional network to generate a compressed dynamic data stream.

[0012] In combination with the first aspect, in some implementations of the first aspect, the real-time carbon signal data includes power grid carbon intensity data and ship emission characteristic data, the power grid carbon intensity data is acquired in real time through an external energy database, and the ship emission characteristic data is generated by analyzing the ship automatic identification system.

[0013] In a second aspect, the present invention provides a port energy dynamic optimization system based on digital twins, the system adopts the method provided in any of the above embodiments, and the system includes: A modeling module is used to build a multi-level digital twin model of the port energy system, which includes a physical layer, an edge layer, and a virtual layer; The data collection module is deployed at the physical layer and collects real-time operating data of port energy equipment through a sensor network; The preprocessing module is deployed at the edge layer to filter, align timestamps, and reduce the dimension of features of real-time running data, generate compressed dynamic data streams, and synchronize the dynamic data streams to the virtual layer; The model training module is deployed in the virtual layer and is used to train the physical information neural network based on dynamic data flow and preset physical constraint equations to generate a multi-energy flow prediction model; The strategy generation module is deployed in the virtual layer and is used to combine multiple energy flow prediction models and real-time carbon signal data to generate energy scheduling strategies at multiple time scales through quantum-inspired optimization algorithms; The instruction decomposition module is deployed at the edge layer to decompose the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmit them back to the energy router at the physical layer; The execution module, deployed on the energy router, is used to execute dynamic adjustment instructions to adjust the operating parameters of the port energy equipment and allocate renewable energy and energy storage resources based on periodic scheduling instructions.

[0014] In a third aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program is used to execute the method provided in any of the above embodiments.

[0015] In a fourth aspect, the present invention provides an electronic device, the electronic device comprising: processor; a memory for storing processor-executable instructions; The processor is used to execute the method provided in any of the above embodiments.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: Step S1 constructs a physical-edge-virtual three-layer digital twin model to provide a unified mapping framework for real-time data streams (such as millisecond-level voltage transients and hour-level energy storage temperatures), breaking the limitations of isolated modeling of traditional subsystems; Step S2 completes data filtering, timestamp alignment and lightweight compression at the edge layer, which not only reduces transmission delays, but also retains key dynamic features (such as crane motor harmonics), providing high-quality input for the virtual layer; Step S3 embeds physical laws such as thermodynamic partial differential equations into model training through a physical information neural network (PINN) Practice, constrain the rationality of data-driven prediction, and combine the quantum-inspired optimization algorithm to simultaneously solve millisecond-level frequency regulation and daily carbon quota allocation to avoid policy conflicts caused by hierarchical optimization; Step S4 decomposes the strategy into dynamic adjustment instructions (voltage / power transient response) and periodic scheduling instructions (carbon quota, procurement plan), and realizes the "prediction-decision-execution" closed loop through edge layer reverse control; Step S5 relies on the power electronic interface and blockchain module of the energy router to correct the frequency deviation at the millisecond level, allocate clean energy according to the periodic scheduling instructions, and verify the consistency of carbon data based on smart contracts.

[0017] The synergistic effect of steps S1 to S5 has the following beneficial effects: First, a data-model closed loop is formed: edge layer preprocessing ensures the quality of virtual layer model input, and the model output strategy reversely controls the physical equipment through the edge layer to form a two-way optimization link; second, cross-time scale coupling is achieved: the quantum optimization algorithm integrates transient response and long-term planning, and the timestamp synchronization module ensures the timing consistency of instructions to avoid overcharging / discharging of energy storage; third, carbon-energy synergy is achieved: real-time carbon signals (grid strength, ship emissions) dynamically weighted optimization targets so that energy distribution meets both economic efficiency and compliance. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 Shown is a flow chart of a method for dynamic optimization of port energy based on digital twins provided in one embodiment of the present invention.

[0020] Figure 2 Shown is a flow chart of a port energy dynamic optimization method based on digital twins provided in another embodiment of the present invention.

[0021] Figure 3 Shown is a structural schematic diagram of a port energy dynamic optimization system based on digital twins provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0023] The specific implementation modes of the present invention are described below.

[0024] Example 1

[0025] like Figure 1 and Figure 2 As shown, the present invention proposes a port energy dynamic optimization method based on digital twin, including: Step S1, constructing a multi-level digital twin model of the port energy system, the multi-level digital twin model including a physical layer, an edge layer and a virtual layer; Step S2, collecting real-time operation data of the port energy equipment through the sensor network in the physical layer, preprocessing the real-time operation data in the edge layer, generating a compressed dynamic data stream, and synchronizing the dynamic data stream to the virtual layer; Step S3, in the virtual layer, the physical information neural network is trained based on the dynamic data flow and the preset physical constraint equations to generate a multi-energy flow prediction model, and a multi-time scale energy scheduling strategy is generated by combining the real-time carbon signal data through a quantum-inspired optimization algorithm; Step S4, decomposing the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmitting them back to the energy router of the physical layer through the edge layer; Step S5, executing dynamic adjustment instructions through the energy router to adjust the operating parameters of the port energy equipment, and allocating renewable energy and energy storage resources based on periodic scheduling instructions.

[0026] This method realizes dynamic optimization of port energy by constructing a multi-level digital twin model of physical layer, edge layer and virtual layer. In the physical layer, the sensor network is deployed at the key nodes of the port energy equipment (such as shore power system, photovoltaic array, energy storage device) to collect real-time operation data, including voltage, current, temperature and equipment status signals. The edge layer is composed of distributed computing nodes. After receiving the physical layer data, it performs preprocessing operations: first, the noise interference (such as abnormal fluctuations caused by electromagnetic interference) is removed by filtering algorithm, and then the data streams with different sampling frequencies (millisecond-level power parameters and minute-level temperature data) are synchronized to a unified clock source through the timestamp alignment module. Finally, the core features (such as voltage harmonics and energy storage charging and discharging curves) are extracted using a lightweight convolutional network, and compressed dynamic data streams are generated and uploaded to the virtual layer. The virtual layer is deployed in the cloud. After receiving the dynamic data stream, multi-energy flow modeling is performed based on the physical information neural network: the network embeds the thermodynamic partial differential equation (describing the heat diffusion of the energy storage system) and the power network conservation equation (such as Kirchhoff's law) into the training process in the form of residual terms to ensure that the prediction results conform to the physical laws.

[0027] At the same time, combined with real-time carbon signal data (carbon emission intensity of power grid, emission characteristics of ships), a multi-time scale scheduling strategy covering milliseconds to days is generated through a quantum-inspired optimization algorithm. The generated strategy is decomposed into dynamic adjustment instructions (such as voltage and frequency regulation signals) and periodic scheduling instructions (such as carbon quota allocation scheme) by the edge layer, and transmitted back to the energy router at the physical layer. The energy router executes dynamic adjustment instructions (such as adjusting the switching frequency of the inverter) through the power electronic interface, and allocates renewable energy and energy storage resources (such as giving priority to photovoltaic power) based on the periodic scheduling instructions, ultimately achieving dynamic optimization of the energy system.

[0028] This method achieves full-link closed-loop control of port energy data by building a physical-edge-virtual three-layer digital twin architecture, and improves the utilization rate of renewable energy and reduces carbon emissions through real-time data synchronization and multi-time scale optimization strategies. The combination of edge layer preprocessing and virtual layer physical constraint model solves the problem of dynamic response delay being out of touch with physical laws in traditional methods.

[0029] In combination with the first aspect, in some implementations of the first aspect, in step S3, the specific steps of training the physical information neural network include: Step S31, extracting energy equipment state parameters from the dynamic data stream as input; Step S32, converting the thermodynamic partial differential equation and the power network conservation equation in the physical constraint equation into residual terms, and performing weighted fusion with the output of the physical information neural network to generate a constraint optimization target; Step S33, updating the weight parameters of the physical information neural network through the distributed nodes in the joint learning algorithm to minimize the constraint optimization target and generate a multi-energy flow prediction model.

[0030] In some embodiments, the training process of the physical information neural network is as follows: first, the energy equipment state parameters, including the state of charge of the energy storage system, the output power of the photovoltaic array and the shore power load curve, are extracted from the dynamic data stream as network input. Subsequently, the physical constraint equations are converted into residual terms: for example, the partial differential equations describing the thermal management of the battery (such as the Fourier heat conduction equation) are discretized into residual form and weighted fused with the neural network output (predicted temperature field) to form a constrained optimization objective function. The joint learning algorithm collaboratively updates the network weights through distributed nodes (such as edge computing units): each node calculates the gradient based on local data and aggregates it to the global model in the cloud through a secure aggregation protocol to ensure data privacy. During the training process, the constraint optimization objective function drives the network to simultaneously meet the data fitting accuracy and physical laws. For example, when predicting the energy storage temperature, it is forced to meet the conservation of energy to avoid physical contradictions that may occur in pure data-driven models.

[0031] Alternatively, the residual term fusion can adopt an adaptive weighting strategy (such as dynamically adjusting the physical constraint weights according to the prediction error) instead of fixed-ratio weighting.

[0032] The physical information neural network constrains the training process through the residual term of the physical equation to ensure that the prediction results meet both data fitting and thermodynamic / electrical laws, avoiding the physical contradictions of pure data-driven models. Distributed joint learning ensures data privacy and model generalization capabilities, improves prediction accuracy, and supports cross-port scenario migration.

[0033] In combination with the first aspect, in some implementations of the first aspect, in step S3, the specific steps of generating the energy scheduling strategy include: Step S34, inputting the energy demand forecast data and real-time carbon signal data output by the multi-energy flow forecasting model into the quantum-inspired optimization algorithm; Step S35, decomposing the energy scheduling problem into a discrete variable optimization sub-problem and a continuous variable optimization sub-problem through a hybrid computing unit in a quantum-inspired optimization algorithm; Step S36, using a quantum annealing algorithm to generate discrete decision parameters for the discrete variable optimization subproblem, and using a gradient descent algorithm to generate continuous decision parameters for the continuous variable optimization subproblem; Step S37, combining discrete decision parameters and continuous decision parameters to generate a multi-time scale energy scheduling strategy.

[0034] In some embodiments, the generation process of the energy dispatch strategy includes the following steps: the demand forecast data (such as photovoltaic power generation in the next 24 hours, ship berthing energy consumption) output by the multi-energy flow prediction model and the real-time carbon signal data (such as the current grid carbon intensity, ship emission factor) are input into the quantum-inspired optimization algorithm. The algorithm decomposes the problem into discrete variable optimization sub-problems (such as energy storage start-stop status, energy router switch instructions) and continuous variable optimization sub-problems (such as energy storage charging and discharging rate, hydrogen energy purchase volume) through a hybrid computing unit. The discrete sub-problem is solved by the quantum annealing algorithm: the problem is mapped to the Ising model, and the optimal solution is searched through the quantum annealing machine, such as determining the start-stop status of the energy storage device during the peak electricity price period. The continuous sub-problem is iteratively solved by the gradient descent algorithm: for example, optimizing the energy storage charging and discharging rate to minimize battery loss. Finally, the two types of parameters are merged to generate a multi-time scale strategy containing millisecond-level adjustment instructions (such as frequency deviation compensation) and daily-level procurement plans.

[0035] Alternatively, discrete variable optimization may employ a genetic algorithm, and continuous variable optimization may employ a sequential quadratic programming method.

[0036] The quantum-inspired optimization algorithm decomposes the complex scheduling problem into discrete and continuous sub-problems, and uses quantum annealing and gradient descent to efficiently solve them, taking into account both millisecond-level control and daily-level planning requirements. The discrete-continuous variable separation strategy reduces the computational dimension.

[0037] In combination with the first aspect, in some implementations of the first aspect, the discrete variable optimization subproblem includes the start and stop status of the energy storage device and the switching instructions of the energy router, and the continuous variable optimization subproblem includes the energy storage charging and discharging rate and the energy purchase amount.

[0038] In some embodiments, the discrete variable optimization subproblem focuses on device state decisions: the start and stop states of energy storage devices need to balance instantaneous power demand and device life (e.g., frequent start and stop accelerates battery aging), and the switch instructions of energy routers need to coordinate multiple energy flow paths (e.g., switching to hydrogen power supply during typhoons). The continuous variable optimization subproblem involves refined control: the energy storage charge and discharge rates need to match the fluctuations of renewable energy (e.g., rapidly increase the discharge rate when photovoltaic power drops sharply), and the energy purchase volume needs to be combined with market prices and carbon quotas (e.g., increase local hydrogen production when carbon prices are high). By separating discrete and continuous variables, the problem dimension is reduced and the solution efficiency is improved, for example, avoiding the computational complexity of mixed integer nonlinear programming.

[0039] This method clarifies the optimization boundaries of discrete variables (equipment start and stop / switch instructions) and continuous variables (charging and discharging rates / purchase volume) to avoid strategy conflicts caused by mixed optimization.

[0040] In combination with the first aspect, in some implementations of the first aspect, step S4 includes: Step S41, extracting voltage frequency deviation and instantaneous power change from the energy dispatch strategy and generating a dynamic adjustment instruction; Step S42, extracting the carbon quota allocation plan and energy procurement plan from the energy scheduling strategy, and generating a periodic scheduling instruction; Step S43, aligning the time bases of the dynamic adjustment instructions and the periodic scheduling instructions through the timestamp synchronization module, generating a time-consistent scheduling instruction set, and transmitting it back to the energy router of the physical layer through the edge layer.

[0041] In some embodiments, the decomposition and synchronization process of the energy scheduling strategy is as follows: extract the voltage frequency deviation (such as fluctuations within the range of ±0.5Hz) and the instantaneous change in power (such as the megawatt-level demand jump caused by the sudden loading of the crane) from the strategy to generate dynamic adjustment instructions; at the same time, extract the carbon quota allocation plan (such as the daily carbon emission cap) and the energy procurement plan (such as the hydrogen energy procurement volume in the next week) to generate periodic scheduling instructions. The timestamp synchronization module uses a precise clock protocol to align the time base of the two types of instructions. For example, the effective time of the millisecond-level control instruction is bound to the start time of the periodic scheduling instruction to ensure the timing consistency of the instruction execution. The synchronized instruction set is transmitted to the energy router through the edge layer to avoid control failure due to transmission delays (such as overcharging of energy storage).

[0042] Dynamic adjustment instructions (voltage / power transient response) and periodic scheduling instructions (carbon quota / purchase plan) are aligned in timing through the timestamp synchronization module, solving the strategy lag problem in traditional hierarchical control.

[0043] In combination with the first aspect, in some implementations of the first aspect, step S5 includes: Step S51, inputting the dynamic adjustment instruction into the power electronic interface of the energy router to generate a voltage regulation signal and an energy storage charging and discharging control signal; Step S52, inputting the periodic scheduling instruction into the blockchain verification module of the energy router to generate energy transaction records and carbon emission audit data; Step S53, using smart contracts to compare the energy transaction records with the carbon emission audit data for consistency, generate energy allocation instructions, and allocate renewable energy and energy storage resources based on the energy allocation instructions.

[0044] In some embodiments, the execution process of the energy router is divided into two steps: the dynamic adjustment instruction input power electronic interface generates voltage regulation signals (such as stabilizing bus voltage through PWM modulation) and energy storage charging and discharging control signals (such as adjusting the duty cycle of the bidirectional converter). The periodic scheduling instruction input blockchain verification module generates energy transaction records (such as power purchase contract hash value) and carbon emission audit data (such as ship operation carbon emission log). The smart contract generates energy allocation instructions by comparing the consistency of energy transaction records with audit data (such as verifying the matching degree of purchased electricity and actual consumption). For example, if the audit data shows that a ship has excessive emissions, the smart contract can automatically freeze its carbon quota application.

[0045] The energy router executes dynamic adjustment instructions (such as ±0.5Hz frequency compensation) through the power interface, the blockchain module records the carbon trading data of the periodic scheduling instructions, and the smart contract automatically verifies the data consistency. This design reduces the cost of manual audits and enables real-time tracking of ship carbon emissions.

[0046] In combination with the first aspect, in some implementations of the first aspect, the blockchain verification module connects the port's internal energy ledger and the external carbon trading platform through a cross-chain protocol, and synchronizes energy trading records and carbon emission audit data to the external carbon trading platform.

[0047] In some embodiments, the blockchain verification module connects the port's internal energy ledger with the external carbon trading platform through a cross-chain protocol. The internal ledger records local energy transactions in the port (such as energy storage lease contracts), and the external platform accesses regional carbon market data (such as EU carbon emission quota prices). The cross-chain protocol ensures atomic synchronization of the two types of data: for example, the carbon quota purchased by the port is updated to the external platform in real time to prevent duplicate calculations. Alternatively, relay chain technology or hash time lock protocol can be used to achieve cross-chain interaction.

[0048] The cross-chain protocol connects the port's internal energy ledger with the external carbon trading platform to ensure the global consistency of carbon quota allocation and transaction records. For example, in the EU carbon tariff scenario, port carbon data can be automatically synchronized to the regulatory system to avoid compliance fines.

[0049] In combination with the first aspect, in some implementations of the first aspect, in step S2, the preprocessing includes: Step S21, filtering the real-time operation data through the abnormality detection unit to generate a filtered data sequence; Step S22, synchronizing the filtered data sequence with a preset clock source through a timestamp alignment module to generate a data stream with a unified time; Step S23, performing feature extraction and dimensionality reduction on the time-uniform data stream through a lightweight convolutional network to generate a compressed dynamic data stream.

[0050] In some embodiments, the specific operations of data preprocessing include: the anomaly detection unit uses a sliding window variance analysis method to identify and filter out abnormal data (such as zero value drift caused by sensor failure); the timestamp alignment module synchronizes data from different devices to a unified time axis through a reference clock source (such as GPS timing); the lightweight convolutional network uses a deep separable convolution layer to extract key features (such as voltage transient waveforms) and reduces the dimension through a pooling layer. Alternatively, anomaly detection can use an isolation forest algorithm, and feature extraction can use an autoencoder structure.

[0051] In the preprocessing stage, anomaly detection, timestamp alignment and lightweight compression are used to compress the original data stream and retain key dynamic features (such as voltage harmonics), reducing the computing load at the edge layer and improving the efficiency of cloud model training.

[0052] In combination with the first aspect, in some implementations of the first aspect, the real-time carbon signal data includes power grid carbon intensity data and ship emission characteristic data, the power grid carbon intensity data is acquired in real time through an external energy database, and the ship emission characteristic data is generated by analyzing the ship automatic identification system.

[0053] In the real-time carbon signal data, the power grid carbon intensity data is obtained in real time from the external energy database through the API interface, reflecting the carbon emissions of the current power grid per megawatt-hour of electricity; ship emission characteristic data is calculated and generated by parsing the navigation log of the ship's automatic identification system (AIS) (such as engine type, fuel consumption rate) combined with the International Maritime Organization (IMO) emission factor library. For example, when a container ship is berthed, its carbon emission weight is dynamically adjusted according to the fuel type (heavy oil / liquefied natural gas) reported by the AIS.

[0054] The integration of power grid carbon intensity data and ship AIS emission data has built a refined real-time carbon signal system. For example, the carbon quota weight is dynamically adjusted according to the ship fuel type, reducing the error in emission accounting.

[0055] Example 2

[0056] like Figure 3As shown, in a second aspect, the present invention provides a port energy dynamic optimization system based on digital twins, the system adopts the method provided by any of the above embodiments, and the system includes: A modeling module 10 is used to construct a multi-level digital twin model of the port energy system, wherein the multi-level digital twin model includes a physical layer, an edge layer, and a virtual layer; The data collection module 20 is deployed at the physical layer and collects real-time operation data of the port energy equipment through the sensor network; The preprocessing module 30 is deployed at the edge layer and is used to filter, align timestamps and reduce features of real-time operation data, generate compressed dynamic data streams, and synchronize the dynamic data streams to the virtual layer; The model training module 40 is deployed in the virtual layer and is used to train the physical information neural network based on the dynamic data flow and the preset physical constraint equations to generate a multi-energy flow prediction model; The strategy generation module 50 is deployed in the virtual layer and is used to combine the multi-energy flow prediction model and the real-time carbon signal data to generate a multi-time scale energy scheduling strategy through a quantum-inspired optimization algorithm; The instruction decomposition module 60 is deployed at the edge layer and is used to decompose the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmit them back to the energy router at the physical layer; The execution module 70 is deployed in the energy router and is used to execute dynamic adjustment instructions to adjust the operating parameters of the port energy equipment and allocate renewable energy and energy storage resources based on periodic scheduling instructions.

[0057] This system corresponds to the method provided in Example 1, and has the following beneficial effects: First, a data-model closed loop is formed: edge layer preprocessing ensures the quality of virtual layer model input, and the model output strategy reversely controls the physical equipment through the edge layer to form a two-way optimization link; second, cross-time scale coupling is achieved: the quantum optimization algorithm integrates transient response and long-term planning, and the timestamp synchronization module ensures the timing consistency of instructions to avoid overcharging / discharging of energy storage; third, carbon-energy synergy is achieved: real-time carbon signals (grid strength, ship emissions) dynamically weighted optimization targets so that energy distribution meets both economic efficiency and compliance.

[0058] The present invention further provides an electronic device, the electronic device comprising: processor; a memory for storing processor-executable instructions; The processor is used to execute the method provided in any of the above embodiments.

[0059] The present invention also provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program is used to execute the method provided in any of the above embodiments.

[0060] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A port energy dynamic optimization method based on digital twin, characterized in that: include: Step S1, constructing a multi-level digital twin model of the port energy system, wherein the multi-level digital twin model includes a physical layer, an edge layer, and a virtual layer; Step S2, collecting real-time operation data of port energy equipment through the sensor network in the physical layer, preprocessing the real-time operation data in the edge layer to generate a compressed dynamic data stream, and synchronizing the dynamic data stream to the virtual layer; Step S3, in the virtual layer, training a physical information neural network based on the dynamic data flow and preset physical constraint equations to generate a multi-energy flow prediction model, and generating a multi-time scale energy scheduling strategy through a quantum-inspired optimization algorithm in combination with real-time carbon signal data; Step S4, decomposing the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmitting them back to the energy router of the physical layer through the edge layer; Step S5, executing the dynamic adjustment instruction through the energy router to adjust the operating parameters of the port energy equipment, and allocating renewable energy and energy storage resources based on the periodic scheduling instruction.

2. The method according to claim 1, characterized in that In step S3, the specific steps of training the physical information neural network include: Step S31, extracting energy device state parameters from the dynamic data stream as input; Step S32, converting the thermodynamic partial differential equation and the power network conservation equation in the physical constraint equation into residual terms, and performing weighted fusion with the output of the physical information neural network to generate a constraint optimization target; Step S33, updating the weight parameters of the physical information neural network through the distributed nodes in the joint learning algorithm so as to minimize the constrained optimization objective and generate the multi-energy flow prediction model.

3. The method according to claim 1, characterized in that In step S3, the specific steps of generating the energy scheduling strategy include: Step S34, inputting the energy demand forecast data output by the multi-energy flow forecasting model and the real-time carbon signal data into the quantum-inspired optimization algorithm; Step S35, decomposing the energy scheduling problem into discrete variable optimization sub-problems and continuous variable optimization sub-problems through the hybrid computing unit in the quantum-inspired optimization algorithm; Step S36, using a quantum annealing algorithm to generate discrete decision parameters for the discrete variable optimization subproblem, and using a gradient descent algorithm to generate continuous decision parameters for the continuous variable optimization subproblem; Step S37, combining the discrete decision parameters and the continuous decision parameters to generate the multi-time-scale energy scheduling strategy.

4. The method according to claim 3, characterized in that The discrete variable optimization sub-problem includes the start and stop status of the energy storage device and the switch instructions of the energy router, and the continuous variable optimization sub-problem includes the energy storage charging and discharging rate and the energy purchase amount.

5. The method according to claim 1, characterized in that: The step S4 comprises: Step S41, extracting the voltage frequency deviation and the instantaneous power change from the energy scheduling strategy to generate the dynamic adjustment instruction; Step S42, extracting the carbon quota allocation plan and energy procurement plan from the energy scheduling strategy to generate the periodic scheduling instruction; Step S43, aligning the time bases of the dynamic adjustment instructions and the periodic scheduling instructions through a timestamp synchronization module, generating a time-consistent scheduling instruction set, and transmitting it back to the energy router of the physical layer through the edge layer.

6. The method according to claim 1, characterized in that The step S5 comprises: Step S51, inputting the dynamic adjustment instruction into the power electronic interface of the energy router to generate a voltage regulation signal and an energy storage charging and discharging control signal; Step S52, inputting the periodic scheduling instruction into the blockchain verification module of the energy router to generate energy transaction records and carbon emission audit data; Step S53, using a smart contract to compare the energy transaction record with the carbon emission audit data for consistency, generate an energy allocation instruction, and allocate renewable energy and energy storage resources based on the energy allocation instruction.

7. The method according to claim 6, characterized in that The blockchain verification module connects the port's internal energy ledger and the external carbon trading platform through a cross-chain protocol, and synchronizes the energy trading records and carbon emission audit data to the external carbon trading platform.

8. The method according to claim 1, characterized in that In the step S2, the preprocessing includes: Step S21, filtering the real-time operation data through an abnormality detection unit to generate a filtered data sequence; Step S22, synchronizing the filtered data sequence with a preset clock source through a timestamp alignment module to generate a data stream with a unified time; Step S23, performing feature extraction and dimensionality reduction on the time-uniform data stream through a lightweight convolutional network to generate the compressed dynamic data stream.

9. The method according to claim 1, characterized in that: The real-time carbon signal data includes power grid carbon intensity data and ship emission characteristic data. The power grid carbon intensity data is acquired in real time through an external energy database, and the ship emission characteristic data is generated by analyzing the ship automatic identification system.

10. A port energy dynamic optimization system based on digital twin, characterized in that: The system adopts the method according to any one of claims 1 to 9, and the system comprises: A modeling module, used to construct a multi-level digital twin model of the port energy system, wherein the multi-level digital twin model includes a physical layer, an edge layer, and a virtual layer; A data collection module is deployed in the physical layer and collects real-time operation data of port energy equipment through a sensor network; A preprocessing module, deployed in the edge layer, for filtering, aligning timestamps and reducing features of the real-time operation data, generating a compressed dynamic data stream, and synchronizing the dynamic data stream to the virtual layer; A model training module, deployed in the virtual layer, for training a physical information neural network based on the dynamic data flow and preset physical constraint equations to generate a multi-energy flow prediction model; A strategy generation module, deployed in the virtual layer, for combining the multi-energy flow prediction model and the real-time carbon signal data to generate a multi-time scale energy scheduling strategy through a quantum-inspired optimization algorithm; An instruction decomposition module, deployed at the edge layer, for decomposing the energy scheduling strategy into dynamic adjustment instructions and periodic scheduling instructions, and transmitting them back to the energy router at the physical layer; An execution module is deployed on the energy router, and is used to execute the dynamic adjustment instruction to adjust the operating parameters of the port energy equipment, and allocate renewable energy and energy storage resources based on the periodic scheduling instruction.

Images