Port-freight corridor collaborative emission reduction method based on data-driven distribution robust optimization

Abstract

This invention belongs to the field of automated port energy management and transportation scheduling, specifically disclosing a port-freight corridor collaborative emission reduction method based on data-driven partial Bourse bar optimization. By constructing a port-corridor integrated energy system, battery swapping stations are integrated into the microgrid as flexible energy storage; simultaneously, truck location trajectory data is mined, and a non-homogeneous Poisson process is used to accurately capture the dynamic arrival rate of traffic flow; a two-stage partial Bourse bar optimization model is constructed, using Wasserstein distance to build a fuzzy set containing dual uncertainties, and reconstructed into a mixed-integer linear programming problem based on strong duality theory. This invention breaks down cross-departmental physical barriers, utilizing low-storage, high-discharge batteries to achieve spatiotemporal energy transfer, maximizing the local consumption of renewable energy while ensuring the system's high risk resistance capability, significantly reducing the total system cost and carbon emissions.

Description

Technical Field

[0001] This invention belongs to the field of automated port energy management and transportation scheduling, specifically involving a port-freight corridor collaborative emission reduction method based on data-driven sub-Brussels optimization. Background Technology

[0002] Currently, building "green freight corridors" connecting ports has become an international consensus, and transportation electrification has become an essential path to achieving deep decarbonization of the transportation industry. With the improvement of port electrification levels, modern ports are gradually transforming from single logistics transshipment nodes into regional energy hubs. However, the operation of port energy systems is currently often isolated, with significant physical and informational barriers between port microgrids and the hinterland freight corridors, leading to severe "source-load mismatch in time and space." For example, during midday at ports, renewable energy (wind and solar) generation is high, but rigid loads are low, easily resulting in wind and solar curtailment; while at night, when truck charging demand surges, the only option is to purchase high-carbon electricity from the main grid.

[0003] To optimize such integrated energy systems, the academic community has conducted extensive cutting-edge research. Specifically, the actual operation of existing port integrated energy systems (PIES) faces the severe challenge of dual uncertainties: on the one hand, as pointed out in existing studies such as "Resiliency-oriented optimal scheduling of microgrids in the presence of demand response programs using a hybrid stochastic-robust optimization approach" and "Integrated energy scheduling under uncertainty for sustainable ports," the supply-side renewable energy generation of wind and solar power has inherent intermittency and volatility; on the other hand, as revealed in studies such as "Multistage dispatch of seaport power systems for incorporating logistical flexibilities in uncertain operational conditions" and "Electricity-hydrogen coupled energy storage bilevel optimization for offshore wind-powered zero-carbon port microgrids considering multiple uncertainties," the demand-side processes of ship berthing, shore power demand, and logistics collection and distribution also exhibit significant randomness.

[0004] To address the aforementioned dual uncertainty problem, existing mainstream optimization algorithms have the following technical limitations:

[0005] (1) When using traditional deterministic optimization or stochastic programming (SP) schemes (such as the strategy discussed in the paper "A risk-averselogistics-energy coordination optimization strategy for port energy system considering demand response program"), the system often relies excessively on precise probability distribution assumptions. This approach has a high risk appetite, and once the actual physical scenario (such as extreme weather or traffic congestion) deviates from the preset prior distribution, the system will face a great risk of supply and demand imbalance and long-tail bias.

[0006] (2) When using traditional robust optimization (RO) schemes (such as the method used in the literature "Robust scheduling of thermal, cooling and electrical hub energy system under market price uncertainty"), the decision logic focuses on the worst-case boundary scenario (such as the extreme superposition of no wind and no light and the load reaching its peak). This method forces the system to meet all boundary conditions within the parameter fluctuation range, resulting in the system configuration maximizing the infrastructure redundancy capacity. This often leads to overly conservative decisions and extremely high initial investment and daily operating costs.

[0007] Although the Distributed Robust Optimization (DRO) method has been proposed in recent years and demonstrated in papers such as "Distributionally Robust Optimal Dispatching Method for Integrated Energy System with Concentrating Solar Power Plant" and "Distributed Robust Optimal Scheduling of Integrated Energy System Considering Demand Response" that it effectively balances economy and robustness by finding the optimal solution under the worst-case probability distribution within fuzzy sets, existing research still lacks fuzzy set construction methods driven by real high-frequency traffic flow data for complex scenarios involving deep cross-sectoral coupling between port microgrids and hinterland freight corridors. Furthermore, it faces the technical bottleneck of model dimensionality explosion under multiple uncertainties, making it difficult to solve. Summary of the Invention

[0008] Purpose of the invention: The purpose of this invention is to address the difficulties in the absorption of new energy sources caused by the physical isolation between port energy systems and freight networks in the existing technology, as well as the technical problems that optimization algorithms in the context of dual uncertainties cannot balance economy and safety. This invention provides a port-freight corridor coordinated emission reduction method based on data-driven sub-Brussels bar optimization.

[0009] Technical solution: This invention is a port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization, comprising:

[0010] Step S1: Construct a port-corridor integrated energy system, connect the port microgrid containing renewable energy power generation equipment with the transportation energy supply node network in the corridor for communication and electrical connection, and configure the backup battery pack in the transportation energy supply node as a flexible load and energy storage unit of the microgrid;

[0011] Step S2: Obtain high-frequency location trajectory data of logistics vehicles, use a non-homogeneous Poisson process to dynamically model the random events of the logistics vehicles arriving at the traffic energy supply node, extract the time-varying features of traffic flow, and generate an empirical distribution of traffic flow uncertainty.

[0012] Step S3: Construct and solve a two-stage sub-Bruker optimization model comprising a first stage and a second stage: The first stage is the capacity planning stage, which, based on historical operational data of ports and corridors, constructs a fuzzy set centered on the empirical distribution and measured by Wasserstein distance to determine the optimal capacity configuration of the renewable energy power generation equipment, transportation energy supply nodes, and backup battery packs; The second stage is the operation scheduling stage, which, based on the optimal capacity configuration, jointly optimizes the charging and discharging strategy of the backup battery packs and the interaction power between the microgrid and the external main grid for the worst-case probability distribution defined by the fuzzy set, in order to minimize the expected total operating cost of the system; wherein, the expected total operating cost of the system includes the environmental cost of system carbon emissions;

[0013] Step S4: Based on the optimal capacity configuration and operation scheduling strategy obtained from the solution, perform source-load-storage coordinated control on the port-corridor integrated energy system. This includes: parsing the continuous decision variables output by the operation scheduling strategy in each scheduling period into time-segmented equipment-level target values; having the energy management controller encapsulate the equipment-level target values ​​into a set of control instructions according to a preset scheduling cycle and send them to the corresponding underlying execution units; and driving the orderly charging, discharging and grid-connection switching actions of the underlying execution units to smooth the load impact caused by dynamic fluctuations in traffic flow on the external main grid and maximize the local consumption of renewable energy in the logistics hub.

[0014] To further improve the above technical solution, in step S1, the corridor is a green freight corridor connecting the port and the hinterland logistics network, the renewable energy power generation equipment includes distributed wind turbines and photovoltaic systems; the transportation energy supply node is a heavy truck battery swapping station; the green electricity generated by the port microgrid prioritizes meeting the rigid operational load inside the port, and the surplus electricity is directed to the power batteries of the heavy truck battery swapping station for storage and supply.

[0015] Furthermore, the process of dynamically modeling the random events of the logistics vehicles arriving at the traffic energy supply node in step S2 includes: performing structured cleaning and spatiotemporal matching on high-frequency location trajectory data; using a screening strategy combining start-end point discrimination and cross-sectional vector method to extract effective vehicle trajectories traveling within the corridor; establishing a non-homogeneous Poisson process model based on a time-dependent intensity function; using the maximum likelihood estimation method to calibrate the parameters of the vehicle arrival rate after approximating the piecewise constant; and extracting the multimodal peak and valley characteristics of intraday fluctuations in vehicle flow to construct an empirical distribution characterizing the uncertainty of the traffic flow.

[0016] Furthermore, the construction of the fuzzy set centered on the empirical distribution and measured by the Wasserstein distance in step S3 includes the following steps:

[0017] The uncertainty of the output of the renewable energy power generation equipment and the uncertainty of the traffic flow are jointly represented as a random vector including wind power output, photovoltaic power output and vehicle arrival flow.

[0018] Based on the historical operational data, a data source is generated containing... A sample set of empirical scenarios is used, and the Dirac point quality function is used to assign equal occurrence weights to each empirical scenario in the sample set to construct the empirical distribution of the random vector.

[0019] The Wasserstein distance between the true distribution and the empirical distribution is defined as the minimum transportation cost required to transfer one probability distribution to another.

[0020] Based on the Wasserstein distance, a set of all possible probability distributions centered on the empirical distribution and whose Wasserstein distance to the empirical distribution is less than or equal to the set robust radius is constructed, and defined as the fuzzy set.

[0021] Furthermore, in step S3, the objective function of the first stage capacity planning stage is to minimize the annualized total cost of system construction, which includes the initial investment cost, system operation and maintenance cost, and battery replacement cost;

[0022] The quantification process of the battery replacement cost is as follows: a nonlinear battery life loss model based on the depth of discharge is introduced, a polynomial function is used to fit the maximum number of cycles of the backup battery pack at different depths of discharge, and the battery damage rate is calculated in combination with the total number of battery charge and discharge cycles, which is then converted into annualized replacement cost.

[0023] Furthermore, in step S3, the objective function of the second stage of operation scheduling is to minimize the expected daily operating cost of the system under the worst-case distribution, and the model in this stage is configured with the following constraints:

[0024] Configure energy balance constraints to force the system to prioritize the consumption of local renewable energy;

[0025] To prevent arbitrage, a mutual exclusion constraint is configured, and Boolean variables are introduced into the mathematical model to strictly limit the integrated energy system from simultaneously purchasing and selling electricity to the external main grid within the same scheduling period, so as to avoid increasing indirect carbon emissions from the power grid.

[0026] Configure reliability constraints for battery swapping services at transportation energy replenishment nodes, and force the number of available backup battery packs within the replenishment node to be greater than or equal to the number of randomly arriving logistics vehicles.

[0027] The system is configured with strong constraints on green electricity consumption, limiting the amount of wind and solar power curtailment from the renewable energy power generation equipment to a preset threshold for the maximum available power generation in their respective time periods.

[0028] Configure underlying physical safety constraints for energy storage, introduce a charge / discharge state mutual exclusion factor in the mathematical model to forcibly restrict the backup battery pack from performing charging and discharging actions simultaneously within the same scheduling period, and update the state of charge of the backup battery pack in real time in conjunction with the charge / discharge energy conversion efficiency, strictly constraining the state of charge within the boundary range formed by the preset lower and upper limits of the state of charge, so as to prevent the battery from being overcharged or deeply over-discharged.

[0029] Furthermore, the model for the second stage of operation and scheduling also includes the environmental cost of system carbon emissions, calculated as follows:

[0030] The cleaned continuous position trajectory data is divided into micro-travel segments, the instantaneous velocity of the trajectory points in each micro-travel segment is extracted, and combined with the vehicle power ratio mapping to a preset two-dimensional threshold range of various standard operating conditions;

[0031] The proportion of time occupied by each standard operating condition in the micro-travel segment is statistically analyzed as the weight of the corresponding operating condition. The direct carbon emissions of the truck are calculated by combining the weight, the corresponding emission factor and the penetration rate of electric heavy trucks.

[0032] The carbon emissions from the sale of surplus renewable energy to the main grid are deducted as an environmental benefit parameter and converted into environmental costs, which are then included in the objective function of the system's expected total operating cost.

[0033] Furthermore, in step S3, when solving the two-stage sub-Bruker optimization model, the strong duality theory is used to reconstruct the three-level nested optimization problem of min-max-min with infinite-dimensional probability distribution variables in the worst probability distribution into an equivalent single-level finite-dimensional mixed integer linear programming problem.

[0034] By introducing dual variables related to the Wasserstein distance constraint and auxiliary variables related to historical sample scenarios, the robustness requirement is transformed into a set of linear constraints, thereby enabling the solver to perform global optimization within a finite-dimensional space.

[0035] Furthermore, the source-load-storage coordinated control executed in step S4 includes interaction logic based on time-of-use pricing:

[0036] Obtain time-of-use electricity price signals from the external main grid;

[0037] During off-peak electricity price periods, by controlling the energy storage converter connected to the backup battery pack within the transportation energy supply node, the backup battery pack can absorb surplus electricity from the external main grid or renewable energy power generation equipment for pre-charging.

[0038] During peak electricity price periods and when renewable energy generation equipment is insufficient, the backup battery pack is discharged by controlling the energy storage converter to meet the transportation energy demand of the corridor, thus replacing the purchase of electricity from the external main grid.

[0039] Furthermore, the process of generating a control instruction set and issuing it to the underlying execution unit in step S4 includes:

[0040] The discharge power variable, power purchase power variable, charging power variable, and power sales power variable output by the second-stage solver are mapped to equipment-level target values ​​and a set of equipment-level control instructions is generated. The set of control instructions includes at least: grid-connected power limit instructions for renewable energy power generation equipment, charging and discharging power instructions for energy storage converters connected to backup battery packs, start / stop and power allocation instructions for charging modules at transportation energy supply nodes, and power purchase / sales switching instructions at grid connection points.

[0041] Specifically, charging power commands, discharging power commands, and start / stop mutual exclusion state commands that restrict simultaneous charging and discharging are generated for the energy storage converter; charging module power allocation commands and backup battery call commands are generated for the transportation energy replenishment node.

[0042] After generating the equipment-level control instruction set, the equipment-level target value is corrected by real-time acquisition of the system's bus power, the state of charge of the backup battery pack, the number of idle backup battery packs at the transportation energy supply node, and the real-time output status of wind and solar power.

[0043] The port energy management system distributes the corrected control command set to each underlying execution device and collects the execution feedback from the underlying devices in real time to form a closed-loop control.

[0044] Beneficial effects: Compared with the prior art, the advantages of the present invention are as follows:

[0045] This invention employs a technology that connects the port microgrid with the corridor battery swapping station network via communication and electrical connections. Leveraging the dual attributes of the swapping stations (replenishment facilities and flexible energy storage), it achieves cross-sectoral energy synergy. Its technical principle lies in breaking down information and physical silos and implementing a "low-storage, high-discharge" strategy, proactively shifting nighttime transportation energy demand to the midday peak photovoltaic power generation period for matching. This feature significantly smooths load fluctuations and, compared to isolated operation modes, effectively avoids wind and solar curtailment, reducing total system costs by approximately 31.5% and carbon emissions by approximately 40.2%.

[0046] Because this invention employs a non-homogeneous Poisson process (NHPP) modeling technique based on location trajectory data, its technical principle lies in dynamically fitting high-frequency arrival events of freight trucks using a time-dependent intensity function. This overcomes the limitation of traditional static probability distributions in failing to capture dramatic intraday peak-valley fluctuations (such as the unique "double-peak" tidal phenomenon in ports). This provides the system with high-fidelity boundary condition inputs, effectively preventing the risk of severely underestimating or overestimating load in power grid planning.

[0047] This invention employs a two-stage sub-Brussels robust optimization model based on Wasserstein distance, and reconstructs it into a mixed-integer linear programming problem using strong duality theory. Its technical principle lies in abandoning the assumption of a single prior distribution and searching for the effective worst-case scenario within a Wasserstein fuzzy set containing real-world empirical distributions. This filters out extremely low-probability pseudo-scenarios while covering statistically significant tail risks. This precise transformation from infinite to finite dimensions makes the model highly computationally feasible, effectively avoiding the failure risk of stochastic programming (SP) and the extreme cost redundancy caused by robust optimization (RO) (the total cost increases by only 3.7% compared to SP, but significantly improves risk mitigation capabilities).

[0048] This invention synergistically combines a nonlinear battery life loss model based on depth of discharge (DOD) with a direct carbon emission calculation technique based on micro-travel segments. Its technical principle lies in accurately mapping the nonlinear degradation physics of electrochemical energy storage and the micro-operating conditions of road segments into economic and environmental parameters. This allows for a true reflection of the implicit depreciation costs under extremely high-frequency charging and discharging during the capacity planning stage. The synergistic effect of these two methods ensures the absolute accuracy of infrastructure investment return calculations and provides a solid scientific foundation for achieving a deep carbon emission reduction of 45.3% in port areas. Attached Figure Description

[0049] Figure 1 A schematic diagram of the physical framework of the Port-Corridor Integrated Energy System (C-PIES) provided for an embodiment of the present invention;

[0050] Figure 2A heat map showing the port scenery and resources in the verification scenario of this invention embodiment;

[0051] Figure 3 This is a diagram showing the distribution of port truck arrival times (arrival rate and cumulative probability) based on historical data in an embodiment of the present invention.

[0052] Figure 4 This is a comparison chart of the predicted truck arrival time and the actual value using the non-homogeneous Poisson process (NHPP) in an embodiment of the present invention.

[0053] Figure 5 This is a Monte Carlo simulation result of a truck arrival scenario in an embodiment of the present invention;

[0054] Figure 6 The diagram shows the 24-hour energy balance results under isolated operation mode (corridor and port optimized separately);

[0055] Figure 7 This is a graph showing the 24-hour energy balance results under the C-PIES collaborative optimization framework of this invention.

[0056] Figure 8 A comparative chart showing the results of collaborative optimization and individual optimization modes across three dimensions: power grid interaction strategy, cost structure, and environmental benefits;

[0057] Figure 9 This is a diagram showing the dynamic response of the energy storage system's state of charge (SOC) and time-of-use (TOU) in an embodiment of the present invention.

[0058] Figure 10 A graph showing the changing trends of system carbon emissions under different levels of electrification;

[0059] Figure 11 A comparison chart of system carbon intensity and emission reduction effects under different levels of electrification;

[0060] Figure 12 This is a comparison chart showing the cost control results of three optimization methods: Directed Reverse Optimization (DRO), Stochastic Programming (SP), and Traditional Robust Optimization (RO). Detailed Implementation

[0061] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0062] Example 1: As Figure 1 As shown, this embodiment constructs a "Port-Corridor Integrated Energy System" (C-PIES) physical architecture. This system breaks through the geographical boundaries of traditional port energy management and consists of two core components: a port-side energy production and supply center and a corridor-side green logistics and transportation network.

[0063] In this architecture, the port is defined as the core physical node and regional energy hub of the system, integrating multiple functions of energy production, storage, and consumption. Distributed wind turbines (WT) and photovoltaic (PV) systems are deployed within the port microgrid. Clean electricity is generated directly from abundant renewable resources, providing the main energy supply for the system. Heavy-duty truck battery swapping stations (BSS) are built within the port. The BSS has a dual function: on the one hand, it serves as a replenishment infrastructure, providing rapid battery swapping services for electric heavy-duty trucks traveling between the port and its hinterland freight corridors, ensuring logistics efficiency; on the other hand, the backup battery packs within the station act as flexible energy storage units for the port microgrid, absorbing fluctuating green electricity through an orderly charging and discharging strategy, thus smoothing system power fluctuations. The port microgrid is connected to the external main grid. The system operates according to the principle of "self-generation and self-consumption, surplus electricity fed into the grid": when green electricity is insufficient, power is purchased from the main grid to ensure the port's rigid operational load and battery swapping needs; when green electricity is surplus, electricity is sold to the main grid to generate economic benefits.

[0064] Given the uncertainties in C-PIES and the lack of precise prior knowledge of probability distributions, traditional deterministic optimization or stochastic programming relying on specific distributions is difficult to apply. Therefore, this embodiment proposes a port-freight corridor coordinated emission reduction method based on data-driven sub-Brussels bar optimization.

[0065] Step 1: Construct the physical architecture of the Port-Corridor Integrated Energy System (C-PIES): Integrate the port microgrid (including distributed photovoltaic, wind power, and energy storage systems) with the battery swapping station (BSS) network in the freight corridor.

[0066] Step 2, Traffic Flow Dynamic Modeling: The random arrival of heavy trucks is modeled using a non-homogeneous Poisson process (NHPP), and the time-varying characteristics of traffic flow are characterized by high-frequency GPS trajectory data.

[0067] Step 3: Construct a two-stage Distributed Bar Optimization (DRO) model:

[0068] Phase 1 (Capacity Planning): Construct Wasserstein fuzzy sets based on historical operating data, formulate robust capacity planning schemes that are immune to probability distribution perturbations, and determine the optimal configuration of wind, solar and energy storage.

[0069] Phase Two (Operation Scheduling): Based on the configuration of Phase One, rescheduling is carried out for the worst-case distribution of uncertainty. By optimizing the charging and discharging strategies of batteries in the battery swapping station and the interaction power with the main grid, the supply and demand differences are balanced in real time, and the expected operating cost of the system is minimized.

[0070] Step 4: Design of Source-Load-Storage Coordination Mechanism: Based on the optimal capacity configuration and operation scheduling strategy obtained from the solution, source-load-storage coordinated control is implemented for the port-corridor integrated energy system. The continuous decision variables output by the operation scheduling strategy in each scheduling period are parsed into time-segmented equipment-level target values. The energy management controller encapsulates the equipment-level target values ​​into a set of control instructions according to a preset scheduling cycle and sends them to the corresponding underlying execution units. By driving the orderly charging, discharging and grid-connection switching actions of the underlying execution units, the load impact caused by the dynamic fluctuations of traffic flow on the external main grid is mitigated, and the local consumption of renewable energy in the logistics hub is maximized.

[0071] Example 2: Due to the intermittency of wind and solar power output and the randomness of truck arrivals in C-PIES, this invention proposes a data-driven two-stage distributed bar optimization (DRO) decision framework.

[0072] 2.1 First Phase: Infrastructure Collaborative Planning Sub-model

[0073] The first phase aims to determine the optimal capacity configuration of wind, solar, and energy storage equipment before the uncertainty parameters are revealed. The objective function is to minimize the annualized total cost of system construction (TAC), including investment costs. Operation and maintenance costs and battery replacement costs :

[0074]

[0075] Introducing the capital recovery factor ( For interest rates, (for the facility's lifespan), fixed investment costs The calculation is as follows:

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] In the formula, , These represent the initial investment cost per unit capacity for wind turbines and photovoltaic systems, respectively. , , These represent the annual operating and maintenance costs per unit capacity of wind turbines, photovoltaic systems, and energy storage systems, respectively. , These represent the installed capacity of the wind turbine and the photovoltaic system, respectively. , These represent the maximum rated power and rated capacity of the energy storage system, respectively. , These represent the investment cost per unit power and per unit capacity of the energy storage system, respectively. , , These represent the capital recovery coefficients for the corresponding wind turbines, photovoltaic systems, and energy storage equipment, respectively. Represents the fixed costs of infrastructure construction for wind turbines, photovoltaic systems, energy storage systems, and battery swapping stations. This indicates the cost of a single heavy-duty truck battery pack. Indicates the first The number of spare batteries configured in each battery swapping station.

[0082] Given the high-frequency characteristics of battery swapping for electric heavy-duty trucks and the need for energy storage to participate in grid ancillary services, battery cycle life has become a key sensitive parameter affecting the long-term economic viability of the system. Simply using a fixed depreciation period cannot accurately reflect the nonlinear degradation characteristics of batteries under actual operating conditions. Therefore, a method based on depth of discharge (D&D) is introduced. A battery life degradation model is used to quantify battery replacement costs, based on the battery's lifespan at a specific depth of discharge. Maximum number of loops The result obtained by polynomial fitting is:

[0083]

[0084] Battery damage rate and battery life The calculation is as follows:

[0085]

[0086]

[0087] In the formula, This represents the total number of charge-discharge cycles within the evaluation period (or typical day). Represents the total number of scheduling days or annualized time coefficient for system operation. Annualized battery replacement cost. The calculation is as follows:

[0088]

[0089]

[0090]

[0091] In the formula, This represents the ratio of battery replacement cost to initial investment cost. Representing the battery number The year of the change. It refers to the number of times the equipment is replaced. It indicates the overall planned operational lifespan of the integrated energy system.

[0092] Given the limited land resources at the port, the construction of photovoltaic panels must meet the constraints of the port's total available area.

[0093]

[0094] In the formula, This indicates the area occupied by a unit capacity photovoltaic system. The total area of ​​the port that can be used for photovoltaic construction.

[0095]

[0096] Limited by port shoreline resources, wind turbine construction is also constrained; the port's geographical resources determine the maximum number of turbine locations that can be constructed. This indicates the number of wind turbines being constructed. Maximum number of racks allowed to be built:

[0097]

[0098] In the formula, This indicates the rated power output of a single unit.

[0099] 2.2 Second Stage: Typical Daily Operational Optimization Sub-model

[0100] The second phase objective is to minimize the expected daily operating costs under the worst-case probability distribution, including electricity purchase costs. Environmental costs :

[0101]

[0102] Taking into account both the offsetting effect of electricity sales to the grid and time-of-use pricing (peak watt-hour pricing) ,flat ,valley Electricity purchase and sales costs:

[0103]

[0104]

[0105]

[0106]

[0107]

[0108] In the formula, This indicates the cumulative electricity purchase cost during peak, normal, and off-peak hours. These respectively represent the system in Active power purchased from and sold to the external main grid during a given period.

[0109] When the output of renewable energy sources such as wind and solar power at a port can meet its needs, the port is permitted to engage in electricity sales. The sold green electricity is considered an effective offset to the carbon emissions from the power grid and is included in the electricity purchase cost calculation.

[0110]

[0111]

[0112]

[0113] In the formula, This indicates the price of electricity sold. This indicates the maximum allowable power limit at the grid connection point.

[0114] To prevent carbon emissions from electricity arbitrage, regulations stipulate that ports must prioritize the use of green electricity generated at their ports and cannot simultaneously purchase and sell electricity.

[0115]

[0116]

[0117]

[0118] In the formula, The tables represent respectively The system is in both power purchase and power sales states during certain time periods.

[0119] Environmental costs include the costs of port power curtailment and grid carbon emission costs. Carbon emission costs do not consider the carbon emission costs generated during construction, but only the carbon emissions from the grid and trucks.

[0120]

[0121]

[0122]

[0123] In the formula, This indicates the penalty cost for abandoning scenic spots; The penalty factor (yuan / kWh) represents the amount of wind and solar power curtailed. Indicates in The wind and solar power generation that was forced to be abandoned during certain periods; These represent the indirect carbon emission costs arising from grid interaction and the direct carbon emission costs from truck fuel consumption, respectively. The grid carbon emission calculation includes the carbon emissions offset by the sale of green electricity.

[0124]

[0125] In the formula, This indicates the carbon tax price (RMB / ton). These represent the carbon emission factor for purchasing electricity from the main grid and the carbon offset factor for selling green electricity to the main grid, respectively. The calculation method for direct carbon emissions from trucks uses a trajectory-based microscopic vehicle emission model. (micro-model), combined with working condition weights and emission factors and penetration rate ,

[0126] By processing massive amounts of GPS data, the cleaned continuous GPS trajectory is divided into segments with a duration of [duration missing]. Micro-travel segments Reference: "Reducing the road freight emissions through integrated strategy in the port cities," extracting micro-trip segments. The instantaneous velocity and acceleration of each high-frequency trajectory point are calculated, and the vehicle specific power (VSP) of the corresponding trajectory point is calculated in conjunction with vehicle physical parameters. Subsequently, the micro-travel segments are classified using instantaneous velocity and vehicle specific power as the criteria. Within the speed fluctuation range The system compares and matches all trajectory points with the two-dimensional threshold ranges of 23 preset standard operating conditions, determining the standard operating condition category to which each trajectory point belongs; and statistically analyzes the data. Standard operating conditions in micro-stroke segments The number of trajectory points contained in the segment is calculated, and this number accounts for a certain percentage of the segment. The time proportion of the total number of trajectory points is used as the first time proportion. Various working conditions in trajectory segments Weights in The emissions of each truck are calculated based on emission factors under different operating conditions. The main calculation steps are as follows:

[0127]

[0128]

[0129] In the formula, Representing the trajectory Middle emissions average emission rate, Indicates the emission factors under different operating conditions. Representing the trajectory emissions Total emissions Indicates micro-travel segments Duration, This indicates the penetration rate of electric heavy-duty trucks.

[0130] Energy balance and energy storage operation constraints: The system follows the energy management strategy of "self-generation and self-consumption, surplus electricity storage, and orderly grid connection." The electricity generated by wind and solar power has three destinations: direct connection to the green power grid, injection into the public bus, and storage / waste. Firstly, the port generates green power to meet its own needs, storing the surplus in the energy storage system or selling it. When the energy storage is full, the green power is sold. If the maximum power sales limit is exceeded, the remaining green power will be wasted. Green power allocation:

[0131]

[0132] In the formula, This refers to the portion of green electricity directly used by the port. This indicates that the remaining green electricity has been sold. This represents the amount of wind power curtailed. Global energy conservation:

[0133]

[0134] In the formula, This indicates the rigid basic load power required to maintain basic operations within the port.

[0135] Energy storage state of charge (SOC) update equation (with percentage limits of 0.2-0.85):

[0136]

[0137] In the formula, These represent the backup battery packs during the current time period. and the previous period The percentage of state of charge; These represent the combined charging efficiency and discharging efficiency of the energy storage converter and the battery, respectively. This indicates the time step size for each scheduling period; This indicates the maximum total energy available in the battery swapping station at this time.

[0138] After arriving at the port, electric heavy-duty trucks undergo rapid battery swapping at battery swapping stations (BSS) to replenish their energy. This process transforms uncertain traffic flow into a flexible charging and discharging load with time-shifting characteristics on the power grid side.

[0139]

[0140]

[0141] In the formula, This indicates the power requirement for battery swapping. Indicates the number of trucks arriving. Indicates the rated energy of a single battery; Indicates that the transportation energy supply node is The number of backup battery packs that are fully charged or available during a given period.

[0142] Energy storage capacity constraints:

[0143]

[0144] In the formula, These represent the minimum safe energy limit allowed by the energy storage system and the current actual stored energy, respectively.

[0145] Energy storage system output power constraints:

[0146]

[0147]

[0148]

[0149] In the formula, This represents a Boolean control variable consisting of 0s and 1s, representing the energy storage system's... The activation status of charging and discharging commands during a given time period.

[0150] Wind power output is affected by real-time wind speed Influence:

[0151]

[0152] In the formula, , , and These represent the real-time wind speed, cut-in wind speed, cut-out wind speed, and rated wind speed, respectively. This is the rated power of the WT system.

[0153] Photovoltaic output is affected by irradiance and temperature Influence:

[0154]

[0155] In the formula, and These represent solar irradiance under real and standard test conditions (1000 W / m2, 25℃), respectively. and These represent the temperatures of the photovoltaic panel under real and standard test conditions, respectively. This is the maximum power output of the photovoltaic panel under standard test conditions; It is the temperature coefficient.

[0156] To maximize the utilization efficiency of wind power, a strict constraint requires that the wind curtailment rate be below 5%.

[0157]

[0158]

[0159] 2.3 Reconstruction of the Two-stage DRO Model Based on Strong Duality Theory

[0160] The two-stage Distributed Robust Optimization (DRO) model for integrated port-freight corridors aims to minimize the expected operating cost of the system under the worst-case probability distribution. Mathematically, this model exhibits a typical "Min-Max-Min" three-layer nested structure. The inner maximization problem requires finding the worst-case probability distribution within a fuzzy set metric using Wasserstein distance. Since fuzzy sets contain infinite-dimensional probability distribution variables, they cannot be directly solved. This embodiment, based on strong duality theory and the geometric properties of the Wasserstein metric, reconstructs the original problem into an equivalent single-layer mixed-integer linear programming problem. This transforms the robustness requirement for the "worst-case probability distribution" into a set of linear constraints, allowing direct global optimization using a solver, ensuring the accuracy and feasibility of the solution.

[0161] To address the triple uncertainties of wind power output, solar power output, and truck traffic flow in the port-corridor system, a random vector is defined. To represent uncertainties, in the two-stage distributed robust optimization model, represents the available output of wind and solar power at each moment within the entire planning period T. Based on historical data, it generates... An experience scenario Constructing empirical distributions :

[0162]

[0163] In the formula, It is the i-th sample Dirac point quality.

[0164] To cover the discrepancy between the true and empirical distributions, a fuzzy set based on Wasserstein distance is constructed. (Empirical distribution) and true distribution The Wasserstein distance between two distributions is defined as the minimum cost required to "transport" one distribution to another:

[0165]

[0166] In the formula, It is the support set of random variables. For uncertain parameters, obey , obey ; yes and The joint probability distribution.

[0167] Based on the Wasserstein distance, construct a... Centered on, with radius Fuzzy set :

[0168]

[0169] In the formula, Indicates the type of uncertainty, Indicating uncertainty The robust radius.

[0170] The original two-stage DRO model (nested Min-Max-Min) is in the following form:

[0171]

[0172] In the formula, Indicates the configuration of a given capacity. and random scenarios Next, the second phase minimizes the runtime scheduling cost function.

[0173] Based on strong duality theory, it is precisely reconstructed into an equivalent finite-dimensional single-level mixed integer linear programming (MILP) problem:

[0174]

[0175]

[0176] in, These are the dual variables related to the Wasserstein distance constraint. Is with the first A historical sample scenario Related auxiliary variables. For the sample , This represents the upper bound of the worst-case cost for this sample.

[0177] Through model reconstruction, the original infinite-dimensional optimization problem is transformed into a large-scale MILP problem with a finite number of variables and linear constraints, which is solved directly using the commercial solver Gurobi.

[0178] The energy storage discharge power variable output by the solver in the second stage ( ), Microgrid power purchase variables ( ), Energy storage charging power variable ( ) and microgrid electricity sales power variables ( Continuous decision variables such as these are mapped to the target values ​​of the underlying hardware device layer to generate a device-level control instruction set. The control instruction set includes at least the grid-connected power limit instruction for renewable energy power generation equipment, the charging and discharging power instruction for energy storage systems, the start / stop and power allocation instruction for the charging module of the battery swapping station, and the power purchase and sale switching instruction for the grid connection point.

[0179] For energy storage systems, charging power commands, discharging power commands, and commands based on mutual exclusion constraints are generated. The system generates start / stop mutual exclusion state commands; generates charging module power allocation commands and backup battery call commands for battery swapping stations; during execution, it performs closed-loop correction on the target values ​​of the equipment layer by collecting physical state quantities such as microgrid bus power, state of charge (SOC) of backup battery packs, number of idle and available batteries in the battery swapping station, and real-time wind and solar power output; finally, the port energy management system sends the corrected control command set to each underlying execution device and collects the execution feedback of the underlying devices in real time.

[0180] Example 3: This example uses Quanzhou Port in Fujian Province and its hinterland freight corridor as the verification object. As an important collection and distribution hub on the southeast coast, Quanzhou Port faces increasingly severe pressure to reduce emissions from landside freight transport. The solar power output of Quanzhou Port is as follows: Figure 2 As shown in the heat map.

[0181] The two-stage sub-Bruker optimization framework proposed in this invention is implemented in Python 3.10.19 environment. The reconstructed MILP model is solved by Gurobi 12.0.2 solver. All numerical simulations are executed on a computer configured with an Intel Core i7-14650HX processor with a main frequency of 2.2GHz and 16GB RAM.

[0182] To verify the feasibility and effectiveness of the proposed two-stage sub-Bruker optimization framework, a numerical case study was conducted using a port scenario. The project planning period was set at 20 years, with an annual interest rate of 6%. Relevant parameters are summarized in Table 1.

[0183] Table 1: Relevant Parameters for Port Scenarios

[0184]

[0185] This embodiment incorporates a time-of-use pricing mechanism for grid interaction to reflect demand-side response potential. Grid electricity prices are set as follows: RMB 993.6 / MWh during peak hours, RMB 613.1 / MWh during normal hours, and RMB 320.4 / MWh during off-peak hours. Furthermore, considering a carbon tax of RMB 288.0 / tCO2, the grid emission factor is 0.581 tCO2 / MWh.

[0186] This embodiment uses high-frequency GPS trajectory data of heavy trucks in Fujian Province as the basis for analysis. The data collection period is from August 16, 2022 to September 19, 2022, a total of 35 days. The original dataset contains fields such as vehicle unique identifier, timestamp, latitude and longitude coordinates, instantaneous speed, and direction angle. First, the massive amount of data is structured by setting a speed threshold to remove drift points and invalid records that have been stationary for a long time. Then, the original CSV file is reconstructed into a columnar storage format using the Polars data processing library, and the timestamps collected by different terminals are standardized to UTC time to achieve spatiotemporal consistency of multi-source data.

[0187] To accurately quantify the carbon emission characteristics of port hinterland channels, this study collected high-frequency GPS trajectory data of port cargo trucks, with the initial dataset covering the trajectories of 1,895 trucks. To ensure measurement accuracy, the study employed a micro-emission calculation framework to clean and match the raw data, removing drift points, stationary points, and invalid samples that could not match emission factors due to missing data, resulting in a total of 1,853 valid high-precision emission data samples from trucks.

[0188] By performing emission calculations and spatial cluster analysis on truck GPS trajectory data, high-emission intensity areas were identified. Therefore, the freight corridor in the Quanzhou-Xiamen direction was selected as a green corridor for construction, thereby forming a coupling between the port-side and roadside energy systems. This site selection was mainly based on the following two considerations: First, the freight corridor passes through densely populated areas and is distributed along the coastline, so reducing the carbon emissions of heavy trucks has good social and environmental benefits; second, the freight corridor is an important section connecting the port, railway hubs, and various stations, and the emission calculation results show that the heavy truck traffic is large and the emission intensity is high.

[0189] By combining OD discrimination and cross-sectional vector analysis, truck traffic data for the Quanzhou-Xiamen green corridor is extracted. Geofencing is used to delineate the Quanzhou Port area (including Shihu, Xiaocuo, Douwei, and Houzhu operating areas) to filter out all port vehicles. Simultaneously, the cross product of vehicle trajectory vectors and cross-sectional vectors is calculated to determine whether a vehicle is traveling within the selected corridor.

[0190] After filtering 34 days of GPS data, 713 of the 1853 vehicles with valid emissions data were trucks traveling from Quanzhou to Xiamen. The emissions data for these 713 trucks are shown in Table 2.

[0191] Table 2: Pollutant Emissions from the Quanzhou-Xiamen Freight Corridor

[0192] Emissions (t) 9.1844 0.6238 39.3678 0.3344 3320.7121

[0193] The port's freight transport system is coupled with operational windows, shipping schedules, and urban traffic tides, resulting in significant non-stationarity and time-varying dynamics in its truck arrival flow. Traditional static probability distributions cannot capture these dramatic intraday peak-and-valley fluctuations. Therefore, this system employs a non-homogeneous Poisson process (NHPP) to model 11,201 valid arrival events. By utilizing maximum likelihood estimation to extract the "bimodal" characteristics of the traffic flow, the NHPP model can extremely accurately capture the dynamic flow changes brought about by port operational windows and urban tides, providing a high-fidelity nominal distribution for the DRO model. It accurately depicts traffic flow trends at the macro level while preserving the independent incremental characteristics of individual arrivals at the micro level.

[0194] Considering that the scheduling of integrated energy systems is typically done on an hourly time granularity, a piecewise constant approximation strategy is adopted. Assume that in the h-th scheduling period, the truck arrival rate... To maintain constancy, based on the processed high-frequency GPS trajectory dataset, the time-division intensity vector is calibrated using the maximum likelihood estimation method. This intensity vector not only describes the baseline shape of traffic flow, but also directly constitutes the nominal distribution of the uncertainty fuzzy set in the subsequent DRO model.

[0195] Based on high-frequency GPS data from Quanzhou Port over 35 consecutive days, covering 11,201 valid arrival events, a truck arrival intensity curve for the port was fitted, as shown below. Figure 3 As shown, the freight traffic flow at Quanzhou Port exhibits a strong "double-peak" characteristic, with trucks mainly arriving at the port between 11:00-12:00 in the morning and 16:00-17:00 in the afternoon.

[0196] Freight corridor traffic flow is significantly non-stationary and time-varying due to the coupling effects of port operation windows, shipping schedules, and urban traffic tides. Traditional static probability distributions struggle to capture its dramatic intraday peak-valley fluctuations, potentially leading to load underestimation or wind / solar curtailment in power grid planning. Therefore, statistical inference was performed on 11,201 extracted valid arrival events, employing a non-homogeneous Poisson process to analyze freight car arrival events. Modeling is performed, and its intensity function is... The cumulative intensity function can describe the instantaneous arrival rate at time t. .

[0197] Figure 4 This paper presents a comparison of the fitting effects of non-homogeneous Poisson process models in predicting truck arrival traffic. Seven consecutive days of historical data from Quanzhou Port were selected as the validation set. In the figure, the blue solid line represents the actual truck arrival traffic based on GPS data, and the red dashed line represents the traffic based on GPS data. The calculated predicted traffic flow. From a time perspective, truck arrival rates exhibit significant periodic fluctuations, with daily peaks and troughs alternating. Comparative results show that the NHPP predicted curve and the actual observed curve maintain a high degree of consistency in both trend and amplitude, accurately capturing dynamic traffic changes influenced by port operation windows and urban traffic tides.

[0198] Figure 5 Showing based on The results of a Monte Carlo simulation of a typical daily truck arrival scenario are shown. The red dashed line in the figure represents the theoretical average arrival rate fitted based on historical data, the blue solid line is the sample mean of 1000 simulation runs, and the light blue shaded area represents the 90% confidence interval. The results show that truck traffic exhibits a typical "bimodal" distribution, occurring between 11:00-12:00 AM and 4:00-5:00 PM, consistent with actual truck arrival patterns.

[0199] Example 4: In order to verify the effectiveness of the port-corridor coordination mechanism of the present invention in solving the problem of spatiotemporal resource mismatch, this example compares the typical daily scheduling strategies of the system under the two modes of collaborative optimization and isolated operation. Figure 6 and Figure 7 The 24-hour energy balance under the two modes is shown respectively, which intuitively reveals the change in operating mode brought about by the collaborative mechanism.

[0200] exist Figure 6In the isolated model, the port microgrid and the freight corridor are treated as two physically isolated islands. The port side possesses abundant renewable energy generation capacity, but its own rigid load is relatively stable and low, resulting in a large surplus of electricity during midday. Due to the lack of inter-system collaborative optimization strategies, this renewable energy cannot be used to supply freight trucks in the corridor and can only be sold as surplus electricity. Simultaneously, the corridor side faces battery swapping loads, and due to the lack of internal system power support, it must purchase high-carbon electricity from the main grid in real time. This coexistence of electricity sales and purchases reflects a typical spatiotemporal mismatch between source and load in the isolated model.

[0201] In contrast, Figure 7 The C-PIES collaborative mode of this invention breaks down physical and information barriers, coupling the port and corridor into a unified energy system. The port's surplus green electricity is no longer sent out, but is instead transported to the battery swapping station for storage on-site, thus completely realizing the spatial and temporal transfer of energy.

[0202] The system demonstrated excellent buffering and regulation capabilities in response to traffic surges identified by the DRO (Driving Response Equipment). Port battery swapping stations acted as energy buffers, responding immediately by discharging electricity when demand suddenly spiked. This mechanism effectively isolated direct, rigid shocks from the supply and demand sides, smoothing load fluctuations while ensuring the safety and stability of the main power grid.

[0203] By analyzing the evolution of power grid interaction strategies, cost structure, and environmental benefits, this study reveals the advantages of the C-PIES architecture in improving system collaborative optimization. Figure 8 This difference was quantified from three dimensions: grid interaction strategy, cost structure, and environmental benefits.

[0204] (a) In terms of grid interaction strategy: In the isolated mode, the grid interaction of the system presents a passive power purchase scheme that follows the load. The power purchase behavior is completely subject to the fluctuation of real-time logistics demand and lacks the ability to actively respond to time-of-use electricity price signals. Especially during the peak operation period at night, due to the lack of flexible adjustment resources, the system is forced to purchase a large amount of electricity during the peak electricity price period. In contrast, the collaborative mode establishes an active scheduling strategy with source-load collaboration as the core. Based on the global optimization perspective, the system implements an aggressive pre-charging strategy during the off-peak electricity price period to fully explore the time-shifting value of port energy storage. During the peak electricity price period, through the complementarity of energy storage discharge and local renewable energy, it almost completely replaces the grid power input, effectively alleviating the grid power supply pressure.

[0205] (b) In terms of economic costs: the isolated model is hampered by high operating expenses, mainly due to peak-hour electricity purchases; although the collaborative model increases the initial CAPEX investment, it significantly reduces OPEX operating costs during the planning period, which not only reduces the total cost but also triggers a fundamental change in the cost structure. This peak-shaving and valley-filling cross-time scheduling strategy significantly improves the return on investment of system operation.

[0206] (c) In terms of environmental benefits, the "load follows source" mechanism based on the DRO model actively shifts the corridor's battery swapping load to the midday period of peak photovoltaic power generation, enabling the system to operate in a "quasi-off-grid" manner during this period. This strategy maximizes the local absorption of renewable energy, significantly replaces grid-purchased electricity, and achieves a carbon emission reduction benefit of 45.3%, verifying the enormous potential of port-corridor integrated collaboration in balancing economic efficiency and decarbonization.

[0207] Figure 9 This study further reveals the dynamic coupling relationship between the state of charge (SOC) of the port energy storage system and the time-of-use (TOU) price of the grid within the C-PIES optimization framework. Results show that the model possesses significant price response characteristics, intelligently formulating charging and discharging trajectories based on electricity price fluctuations, and significantly reducing system operating costs through a cross-time-period energy management mechanism. During off-peak hours, the system does not idle energy storage due to lower loads but instead implements an aggressive charging strategy, causing the SOC to rapidly climb and approach the safe upper limit of the energy storage capacity. This strategy aims to complete energy storage using the lowest marginal electricity purchase cost to cope with subsequent peak loads and high electricity prices. When electricity prices reach peak levels, the SOC begins to decline, and the energy storage system performs discharging operations during this period, releasing previously stored low-cost electricity to replace expensive grid electricity. This low-storage, high-discharge strategy significantly reduces electricity purchase expenditures, achieving peak shaving and valley filling, and alleviating grid supply pressure. During normal electricity price periods, the system exhibits high flexibility. Through a source-load-storage coordination mechanism, it prioritizes the use of renewable energy output during this period to meet real-time loads, supplementing with energy storage regulation when renewable energy output is insufficient. This sophisticated energy management strategy ensures full utilization of renewable energy while maximizing both economic and environmental benefits of the system.

[0208] Figure 10 This study demonstrates the change in system carbon emissions as the electrification rate of trucks increases from 10% to 90% under the C-PIES framework. The results show that this framework effectively reduces carbon emissions through a synergistic "source-load-storage" mechanism. Although the electrification transition leads to an increase in indirect emissions from grid-purchased electricity, this relatively small environmental marginal cost successfully offsets the higher carbon emissions from direct diesel combustion. This confirms the significant advantages of port-side high-proportion renewable energy consumption and energy storage peak-shaving and valley-filling strategies in reducing equivalent carbon intensity.

[0209] Through the first phase of capacity optimization, the system effectively mitigates the risk of grid impact under high load conditions. Even large-scale charging demand does not lead to excessive reliance on high-carbon grid power, thus ensuring that the emission reduction benefits outweigh the negative impacts of indirect emissions. Figure 11 As shown, despite increased electricity demand, the system's average carbon intensity remained consistently low, significantly below the grid benchmark. Notably, the system's carbon intensity decreased when electrification reached 80%. This is because, at 80% electrification, the model opted to build one new wind turbine, offsetting some of the carbon emissions from grid-purchased electricity.

[0210] As electrification rates approach 90%, the system's carbon emission intensity shows a slight marginal increase. This phenomenon is due to the fact that port photovoltaic installations have reached a saturation threshold due to limited space resources. Under these conditions, further increasing the penetration rate of green electricity requires further expansion of the incremental configuration of "wind power + large-scale energy storage".

[0211] However, the high cost of building new wind turbines and energy storage facilities, coupled with the substantial capital expenditure required for energy storage to mitigate the intermittency of wind power, means the system cannot effectively recoup its costs by replacing the operating expenses of cheap off-peak grid electricity within the project cycle. The marginal emission reduction cost of infrastructure has already exceeded the social cost of purchasing electricity from the grid. Therefore, although the model aims for deep decarbonization, the optimization results do not favor 100% renewable energy coverage. Purchasing cheap off-peak electricity as an option for nighttime and extreme weather scenarios has proven to be the globally optimal strategy that balances decarbonization goals with investment economics.

[0212] Figure 12 This paper presents a visual comparison of the investment and operating costs of the Distributed Robust Optimization (DRO) method of this invention with stochastic programming (SP) and traditional robust optimization (RO).

[0213] The Reverse Optimization (RO) method, due to its extreme defense mechanism and highly redundant configuration, is built upon a box-like uncertainty set. Its decision-making logic forces the system to satisfy all boundary conditions within the parameter fluctuation range, including extremely severe scenarios with very low physical probability of occurrence (such as the superposition of zero renewable energy output and peak load). This overly conservative strategy forces the system to configure maximum infrastructure redundancy capacity and limits the flexibility of real-time scheduling under extreme constraints, resulting in severe resource mismatch and high operating costs, reaching 25.03, far exceeding those of SP and DRO.

[0214] The SP method achieves the lowest annualized total cost (9.94), but this is often based on a strong assumption: that the pre-defined probability distribution can perfectly describe the actual physical scenario. However, due to the significant non-Gaussian, multimodal, and long-tailed distribution characteristics of port cargo handling heavy truck traffic flow and renewable energy output, the SP model faces potential distribution bias risks, and its low cost inherently carries a certain degree of risk.

[0215] In contrast, the DRO method adopted in this invention abandons the assumption of a single prior distribution. By constructing a Wasserstein fuzzy set containing empirical distribution characteristics, it can identify and avoid statistically significant worst-case scenarios. Although this results in a slightly higher investment cost (10.03) than SP, it gains extremely strong robust immunity to dual uncertainties, effectively avoiding the risk of system failure caused by improper distribution assumptions in SP, and achieving the optimal trade-off between economy and safety.

[0216] The results show that the Wasserstein distance-based DRO model used in this embodiment achieves strong robustness to uncertainty with minimal economic cost when dealing with the dual uncertainties of port and corridor, demonstrating good overall benefits and confirming the applicability of the DRO model in solving comprehensive energy problems in ports.

[0217] As described above, although this embodiment uses Quanzhou Port in Fujian Province and wind turbines and photovoltaic systems of a specific capacity as examples, those skilled in the art should understand that these specific geographical locations, time spans (e.g., 35 days), equipment rated parameters (e.g., 0.5kW for photovoltaic systems, 2000kW for wind turbines), and empirical data (e.g., 23 operating conditions, 11201 events) are merely specific implementation scenarios to verify the feasibility of this method. In actual industrial applications, the above parameters can be adaptively adjusted and recalibrated according to the installed capacity of different port microgrids, the actual traffic flow of different freight corridors, and the parameters of different models of electric heavy trucks. Such equivalent substitutions do not depart from the protection scope of this invention.

Claims

1. A port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization, characterized in that, Includes the following steps: Step S1: Construct a port-corridor integrated energy system, connect the port microgrid containing renewable energy power generation equipment with the transportation energy supply node network in the corridor through communication and electrical connection, and configure the backup battery pack in the transportation energy supply node as a flexible load and energy storage unit of the microgrid; Step S2: Obtain high-frequency location trajectory data of logistics vehicles, use a non-homogeneous Poisson process to dynamically model the random events of the logistics vehicles arriving at the traffic energy supply node, extract the time-varying features of traffic flow, and generate an empirical distribution of traffic flow uncertainty. Step S3: Construct and solve a two-stage sub-Bruker optimization model comprising a first stage and a second stage: The first stage is the capacity planning stage, which, based on historical operational data of ports and corridors, constructs a fuzzy set centered on the empirical distribution and measured by Wasserstein distance to determine the optimal capacity configuration of the renewable energy power generation equipment, transportation energy supply nodes, and backup battery packs; The second stage is the operation scheduling stage, which, based on the optimal capacity configuration, jointly optimizes the charging and discharging strategy of the backup battery packs and the interaction power between the microgrid and the external main grid for the worst-case probability distribution defined by the fuzzy set, in order to minimize the expected total operating cost of the system; wherein, the expected total operating cost of the system includes the environmental cost of system carbon emissions; Step S4: Based on the optimal capacity configuration and operation scheduling strategy obtained from the solution, perform source-load-storage coordinated control on the port-corridor integrated energy system, including: parsing the continuous decision variables output by the operation scheduling strategy in each scheduling period into time-segmented equipment-level target values, and having the energy management controller encapsulate the equipment-level target values ​​into a set of control instructions according to a preset scheduling cycle and send them to the corresponding underlying execution units, thereby driving the orderly charging and discharging and grid connection switching actions of the underlying execution units.

2. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 1, characterized in that, In step S1, the corridor is a green freight corridor connecting the port and the hinterland logistics network, the renewable energy power generation equipment includes distributed wind turbines and photovoltaic systems, and the transportation energy supply node is a heavy truck battery swapping station. The green electricity generated by the port microgrid is prioritized to meet the rigid operational loads within the port, and any surplus electricity is directed to the power batteries of the heavy truck battery swapping station for storage and supply.

3. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 2, characterized in that, The process of dynamically modeling the random event of the logistics vehicle arriving at the transportation energy supply node in step S2 includes: The high-frequency location trajectory data is structured and spatiotemporally matched. A screening strategy combining start-end point discrimination and cross-sectional vector method is adopted to extract the effective vehicle trajectories traveling in the corridor. A non-homogeneous Poisson process model based on a time-dependent intensity function is established. The maximum likelihood estimation method is used to calibrate the parameters of the vehicle arrival rate after approximation of the piecewise constant. Multimodal peak and valley features of intraday fluctuations in vehicle flow are extracted to construct an empirical distribution characterizing the uncertainty of the traffic flow.

4. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 3, characterized in that, The construction of the fuzzy set centered on the empirical distribution and measured by the Wasserstein distance in step S3 includes the following steps: The uncertainty of the output of the renewable energy power generation equipment and the uncertainty of the traffic flow are jointly represented as a random vector including wind power output, photovoltaic power output and vehicle arrival flow. Based on the historical operational data, a data source is generated containing... A sample set of empirical scenarios is used, and the Dirac point quality function is used to assign equal occurrence weights to each empirical scenario in the sample set to construct the empirical distribution of the random vector. The Wasserstein distance between the true distribution and the empirical distribution is defined as the minimum transportation cost required to transfer one probability distribution to another. Based on the Wasserstein distance, a set of all possible probability distributions centered on the empirical distribution and whose Wasserstein distance to the empirical distribution is less than or equal to the set robust radius is constructed, and defined as the fuzzy set.

5. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 4, characterized in that, In step S3, the objective function of the first stage of capacity planning is to minimize the annualized total cost of system construction, which includes the initial investment cost, system operation and maintenance cost, and battery replacement cost. The quantification process of the battery replacement cost is as follows: a nonlinear battery life loss model based on the depth of discharge is introduced, a polynomial function is used to fit the maximum number of cycles of the backup battery pack at different depths of discharge, and the battery damage rate is calculated in combination with the total number of battery charge and discharge cycles, which is then converted into an annualized replacement cost.

6. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 5, characterized in that, In step S3, the objective function of the second stage of operation scheduling is to minimize the expected daily operating cost of the system under the worst-case distribution, and the model in this stage is configured with the following constraints: Configure energy balance constraints to force the system to prioritize the consumption of local renewable energy; To prevent arbitrage, a mutual exclusion constraint is configured, and Boolean variables are introduced into the mathematical model to strictly limit the integrated energy system from simultaneously purchasing and selling electricity to the external main grid within the same scheduling period, so as to avoid increasing indirect carbon emissions from the power grid. Configure reliability constraints for battery swapping services at transportation energy replenishment nodes, and force the number of available backup battery packs within the replenishment node to be greater than or equal to the number of randomly arriving logistics vehicles. Configure strong constraints on green electricity consumption to limit the amount of wind and solar power curtailment of the renewable energy power generation equipment to not exceed a preset threshold of the maximum available power generation in their respective time periods; Configure underlying physical safety constraints for energy storage, introduce a charge / discharge state mutual exclusion factor in the mathematical model to forcibly restrict the backup battery pack from performing charging and discharging actions simultaneously within the same scheduling period, and update the state of charge of the backup battery pack in real time in conjunction with the charge / discharge energy conversion efficiency, strictly constraining the state of charge within the boundary range formed by the preset lower and upper limits of the state of charge, so as to prevent the battery from being overcharged or deeply over-discharged.

7. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 6, characterized in that, The model for the second phase of operation and scheduling also includes the environmental cost of system carbon emissions, calculated as follows: The cleaned continuous position trajectory data is divided into micro-travel segments, the instantaneous velocity of the trajectory points in each micro-travel segment is extracted, and combined with the vehicle power ratio mapping to a preset two-dimensional threshold range of various standard operating conditions; The proportion of time occupied by each standard operating condition in the micro-travel segment is statistically analyzed as the weight of the corresponding operating condition. The direct carbon emissions of the truck are calculated by combining the weight, the corresponding emission factor and the penetration rate of electric heavy trucks. The carbon emissions from the sale of surplus renewable energy to the main grid are deducted as an environmental benefit parameter and converted into environmental costs, which are then included in the objective function of the system's expected total operating cost.

8. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 7, characterized in that, In step S3, when solving the two-stage sub-Bruker optimization model, the strong duality theory is used to reconstruct the three-level nested optimization problem of min-max-min with infinite-dimensional probability distribution variables in the worst probability distribution into an equivalent single-level finite-dimensional mixed integer linear programming problem. By introducing dual variables related to the Wasserstein distance constraint and auxiliary variables related to historical sample scenarios, the robustness requirement is transformed into a set of linear constraints, thereby enabling the solver to perform global optimization within a finite-dimensional space.

9. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 1 or 8, characterized in that, The source-load-storage coordinated control executed in step S4 includes interactive logic based on time-of-use pricing: Obtain time-of-use electricity price signals from the external main grid; During off-peak electricity price periods, by controlling the energy storage converter connected to the backup battery pack within the transportation energy supply node, the backup battery pack can absorb surplus electricity from the external main grid or renewable energy power generation equipment for pre-charging. During peak electricity price periods and when renewable energy generation equipment is insufficient, the backup battery pack is discharged by controlling the energy storage converter to meet the transportation energy demand of the corridor, thus replacing the purchase of electricity from the external main grid.

10. The port-freight corridor coordinated emission reduction method based on data-driven sub-bar optimization according to claim 9, characterized in that, The process of generating a control instruction set and sending it to the underlying execution unit in step S4 includes: The discharge power variable, power purchase power variable, charging power variable, and power sales power variable output by the second-stage solver are mapped to equipment-level target values ​​and a set of equipment-level control instructions is generated. The set of control instructions includes at least: grid-connected power limit instructions for renewable energy power generation equipment, charging and discharging power instructions for energy storage converters connected to backup battery packs, start / stop and power allocation instructions for charging modules at transportation energy supply nodes, and power purchase / sales switching instructions at grid connection points. Specifically, charging power commands, discharging power commands, and start / stop mutual exclusion state commands that restrict simultaneous charging and discharging are generated for the energy storage converter; charging module power allocation commands and backup battery call commands are generated for the transportation energy replenishment node. After generating the equipment-level control instruction set, the equipment-level target value is corrected by real-time acquisition of the system's bus power, the state of charge of the backup battery pack, the number of idle backup battery packs at the transportation energy supply node, and the real-time output status of wind and solar power. The port energy management system sends the corrected control command set to each underlying execution device and collects the execution feedback from the underlying devices in real time to form a closed-loop control.

Images