Harbor district power distribution network sequential energy storage configuration optimization method based on traffic-electric energy fusion

Abstract

The invention discloses a port area power distribution network sequential energy storage configuration optimization method and system based on traffic-electric energy fusion. The method comprises the following steps: acquiring multi-source data of a traffic system and a power system, and carrying out multi-source data collection and baseline modeling to obtain port baseline data; based on the port area baseline data, carrying out micro-amplitude peak shifting on the traffic side operation phase to execute power spectrum shaping, and obtaining traffic side shaped data; based on the traffic side reshaped data and harbor area baseline data, traffic side executable flexibility capability, cost and risk are packaged into a standardized three-signal interface; and based on the standardized three-signal interface, the energy storage equipment model and the power grid constraint set, performing power side closed-loop solution to obtain a power side feedback signal and power side execution scheduling. Through a sequential framework, active noise reduction on the traffic side and flexible quantification cooperation on the power side are achieved, and the energy storage configuration cost and the system operation cost are reduced on the premise that the operation throughput is not affected.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of port power distribution network, and particularly relates to a port power distribution network sequence type energy storage configuration optimization method based on traffic-electricity fusion. BACKGROUND

[0002] Under the guidance of the double carbon target, the greenization and electrification transformation of the port as a key node of global logistics and energy consumption is imperative. The all-electrified port power distribution network (for example, a large number of electrically driven shore cranes, rail-mounted cranes and electric trucks) not only improves the operation efficiency, but also becomes a heavy load center with high power, high concurrency and strong randomness. Especially the instantaneous superposition of large power operations such as shore crane lifting poses a serious challenge to the power quality and safety and stability of the port power grid. Therefore, how to coordinate the flexible resources such as energy storage system with the port operation to suppress power fluctuation and reduce operation cost has become the research focus in the field of modern port energy management.

[0003] Currently, the existing technology mainly focuses on passive management on the power system side for the configuration and dispatching of port energy storage. For example, the energy management system (EMS) uses the energy storage system to perform peak clipping, valley filling, demand control or price arbitrage based on short-term prediction of the total load of the port. These strategies usually regard the traffic operation system (TOS) of the port as a given and uncontrollable load source. In some schemes, traditional rigid demand response is also used, such as uniformly delaying or reducing the charging power of electric trucks during the peak period of the power grid to obtain price concessions. These methods achieve preliminary interaction between the power grid and the port load to some extent.

[0004] However, the above existing technology has limitations in dealing with the volatility of the highly coupled traffic-electricity system in the port, such as passive response of energy storage to high-frequency impact, inability to finely quantify the value of traffic flexibility, and lack of cross-system collaborative optimization. First, the energy storage system is forced to passively and fully suppress the instantaneous and high-frequency power fluctuations of shore crane lifting and other operations. This high-rate and high-frequency cyclic charging and discharging does not eliminate fluctuations from the source, but rather leads to rapid degradation of energy storage (such as batteries) and excessive wear of the converter (PCS), significantly increasing the life cycle cost of the system. Second, the power side (such as EMS) cannot perceive the real and executable flexible boundaries of the traffic side (such as TOS). The power side regards the operation load as a rigid black box and cannot quantify the marginal cost (such as operation delay cost and congestion risk) and realization risk (such as uncertainty) of traffic flexibility (for example, moving a shore crane operation by a few seconds). Finally, due to the lack of the above quantitative interface, the system cannot establish a sequence and collaborative optimization model between the energy storage degradation cost and the traffic operation cost, resulting in separate management of the two and failure to achieve the optimal overall operation cost of the port. SUMMARY

[0005] The purpose of this invention is to provide a sequential energy storage configuration optimization method for port area distribution networks based on the integration of transportation and electricity. In order to solve one of the aforementioned technical problems existing in the current technology.

[0006] Technical solution: A sequential energy storage configuration optimization method for port area distribution networks based on transportation-electricity integration, including: Acquire multi-source data from the transportation and power systems, perform multi-source data aggregation and baseline modeling, and obtain port area baseline data; Based on the port area baseline data, the traffic-side operation phases are slightly staggered to perform power spectrum shaping, resulting in traffic-side shaped data. Based on the traffic-side sculpted data and port area baseline data, the traffic-side executable flexibility, costs, and risks are encapsulated into a standardized three-signal interface. Based on the standardized three-signal interface, energy storage device model and power grid constraint set, a closed-loop solution is performed on the power side to obtain the power side feedback signal and power side execution scheduling.

[0007] Preferably, rolling verification and data reinjection are performed based on the power-side feedback signal and the power-side execution scheduling to generate a reinjection dataset; the reinjection dataset is then used to update the multi-source data aggregation and baseline modeling.

[0008] Optionally, a slight peak shift is performed on the traffic-side operation phase to perform power spectrum shaping, resulting in traffic-side shaped data, including: Based on the prototype of adjustable equipment constraints contained in the port area baseline data, micro-phase variables and feasible domain constraints for traffic-side operations are established. Analyze the physical limits of the energy storage device model to determine the target cutoff frequency corresponding to the power fluctuations that the energy storage can withstand; A spectral shaping optimization objective is constructed, and the micro-phase variables are solved to minimize the high-frequency energy above the target cutoff frequency in the total power sequence of the port area. Based on the phase shift scheme obtained from the solution and the port area baseline data, traffic-side shaped data is generated. The traffic-side data after shaping includes the port area power sequence after shaping, the adjustable slope upper limit sequence, and the adjustable power boundary sequence.

[0009] Optionally, a spectral shaping optimization objective is constructed, including: Based on the equipment adjustable constraint prototype and rolling prediction set in the port area baseline data, the operation cost and risk function of traffic-side operations are evaluated. By integrating high-frequency energy with operational costs and risk functions, a combined optimization objective is formed; The cost and risk function should include at least: cost of operation delay, risk of work-in-process spillover, and penalty for pile congestion.

[0010] Optionally, minimizing high frequency energy comprises: constructing a linearized power model of the port total power based on the micro-phase variable and the event power templates in the port baseline data; designing a discrete high-pass filter according to the target cut-off frequency and constructing a corresponding convolution matrix; applying the convolution matrix to the linearized power model to express the high frequency energy as a time-domain quadratic form with respect to the micro-phase variable.

[0011] Optionally, solving the micro-phase variable to obtain the phase micro-shift scheme as shown in Figure 5 constructing a feasible region projection operator based on the feasible region constraints; solving a relaxation problem of the spectral shaping optimization objective in continuous domain, which combines the time-domain quadratic form and the operation cost and risk function, to obtain a continuous solution; applying the feasible region projection operator to constraint repair the continuous solution and performing discretization to generate the phase micro-shift scheme.

[0012] Optionally, the standardized three-signal interface comprises: a flexible envelope based on the traffic-side reshaped data, for quantifying the traffic-side up and down power limits, up and down slope limits, and minimum sustainable duration; a piecewise convex cost function mapped based on the port baseline data and the operation cost log and shaping quality indicators derived in the power spectrum shaping step, for quantifying the marginal cost of invoking flexible resources; and a reliability coefficient sequence estimated based on the uncertainty prediction and historical realization records in the port baseline data, for quantifying the realizable confidence of flexible commitment.

[0013] Optionally, the piecewise convex cost function mapped comprises: aggregating traffic-side operation costs and estimating marginal cost slopes based on the operation cost log and the prediction set in the port baseline data; applying spectral shaping revenue discount to the marginal cost slopes using the shaping quality indicators; applying risk amplification to the discounted slopes based on the uncertainty evaluation in the port baseline data; performing convexity tuning and integration on the slopes to obtain a base piecewise convex cost; applying the reliability coefficient sequence to conservatize the base piecewise convex cost to generate the piecewise convex cost function.

[0014] Optionally, the reliability coefficient sequence estimated comprises: ​extracting uncertainty factors from the port baseline data, the uncertainty factors at least including: weather and photovoltaic prediction bias, uncertainty of pile position occupation and queuing, and operation arrival and completion time fluctuation; fusing the uncertainty factors to generate a preliminary confidence time series; performing calibration on the preliminary confidence time series based on historical realization records in the port baseline data, minimizing consistency residual of the preliminary confidence and the historical realization probability to obtain a reliability coefficient sequence.

[0015] Optionally, the power side closed-loop solution is performed, including: analyzing a standardized three-signal interface to obtain a flexible envelope and the reliability coefficient sequence; applying the reliability coefficient sequence to conservatively reduce the upper and lower power limits and the uplink and downlink slope limits defined in the flexible envelope, to generate a conservative flexible envelope; the conservative flexible envelope is used as a constraint set for the power side closed-loop solution.

[0016] Optionally, the power side closed-loop solution is performed, including: constructing a joint objective function, which aims to minimize the total cost in the rolling time domain; the total cost at least aggregates: grid purchase cost derived based on the grid constraint set; a piecewise convex cost function; energy storage degradation cost derived based on the energy storage device model; and grid side ramp rate penalty derived based on the grid constraint set.

[0017] Optionally, the power side closed-loop solution is performed to obtain a power side feedback signal, including: extracting a power shadow price from the dual solution of the system power balance constraint after solving the joint objective function; extracting a slope shadow price from the dual solution of the grid side ramp rate constraint; generating a platformized power request based on the optimal scheduling trajectory obtained by solving, combined with the power shadow price and the slope shadow price; the power side feedback signal includes the power shadow price, the slope shadow price and the platformized power request.

[0018] Beneficial effects, the present application realizes traffic side active noise reduction and power side flexible quantization cooperation through the sequential architecture, reduces the energy storage configuration cost and system operation cost without affecting the operation throughput. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A step flowchart of a port distribution network sequential energy storage configuration optimization method based on traffic-electricity fusion in the embodiments of the present application.

[0020] Figure 2 A step flow chart for obtaining traffic side shaping data in embodiments of the present application.

[0021] Figure 3 A step flow chart for constructing a spectrum shaping optimization objective in embodiments of the present application.

[0022] Figure 4 A step flow chart for minimizing high frequency energy above a target cutoff frequency in a port-wide power sequence in embodiments of the present application.

[0023] Figure 5 A step flow chart for solving a micro-phase variable to obtain a phase micro-shift scheme in embodiments of the present application. DETAILED DESCRIPTION

[0024] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the scope of protection of the present application.

[0025] Embodiment one, in this embodiment, first, the multi-source data of the traffic system and the power system is obtained. The multi-source data refers to the basic information required for implementing the present application.

[0026] Specifically, these data are collected and divided into at least three categories: Traffic system data: including but not limited to: detailed timing of shore crane operation (records the operation plan and actual operation start and end time of each shore crane), operation section phase (for example, distinguishing the stages of shore crane operation with different power characteristics such as lifting, rotating, trolley, and descending), yard work-in-process curve (reflecting the busy degree and buffer capacity of yard operation), and electric truck operation log. The electric truck operation log preferably contains high-precision position trajectory, cumulative running mileage, estimated arrival and departure time window, and connection and disconnection log of charging pile.

[0027] Power system data: including but not limited to: active power sequence of key busbars inside the port (for example, within the station) (high frequency sampling, for example, second level), branch current distribution, key node voltage, and historical grid input power ramp rate (RampRate) data.

[0028] Environmental and prediction data: including but not limited to: future short-term (for example, future 4 hours or 24 hours) weather forecast, photovoltaic output prediction, and tidal period (tide affects ship berthing, and then affects shore crane operation plan).

[0029] After the above data is acquired, multi-source data collection and baseline modeling are performed to obtain the port baseline data.

[0030] Baseline modeling refers to processing, analyzing and calibrating the above multi-source raw data to generate a set of structured prior models and prediction data that can be used for subsequent optimization solving. The modeling process is based on the physical executable characteristics of equipment (such as shore cranes, trucks) and processes (such as charging, loading and unloading), and is completed through statistical analysis, machine learning training or physical parameter calibration.

[0031] The final port baseline data obtained is a structured data set, which preferably includes the following three core components: 1. Baseline port power sequence (P port_base ): This is a time series representing the total active power consumption curve of the port in the future period predicted according to historical patterns and current operation plans without applying any optimization of the present invention (i.e. without phase peak shifting).

[0032] 2. Equipment adjustable constraint prototype (ProtoConstraints): is a set of key rules used to define the physical and process boundaries of the traffic side equipment that can be scheduled. Exemplarily, the prototype includes: For shore cranes: the allowable micro-movement window (e.g. ±30 seconds) for its specific operation phase (such as lifting), the maximum power slope upper limit of lifting and slewing, the must-be-continuous operation segment (lock _flag =1) that cannot be interrupted.

[0033] For electric truck charging: the rated capacity of charging piles, the start-up delay from standby to full power, the minimum switching delay between charging piles, and other prior knowledge.

[0034] 3. Rolling forecast set (Forecast bundle ): is a dynamically updated prediction data package used to support short-term rolling optimization. Specifically, it includes: short-term port total power prediction (as a supplement or correction to P port_base ), short-term photovoltaic output prediction, forecast of in-process inventory quantity of the yard (for assessing congestion risk), and future idle time prediction of charging pile parking spaces.

[0035] In this embodiment, the above three components (P port_base , ProtoConstraints, Forecast bundle ) generated by baseline modeling collectively constitute the port baseline data, providing essential data input and constraint boundaries for subsequent power spectrum shaping (Example III) and three-signal packaging (Example IV).

[0036] Embodiment II describes the overall framework of the sequential energy storage configuration optimization method provided by the present application. Through the novel sequential architecture of traffic-side first shaping and power-side later solving, information closed-loop backfilling, efficient coordination of the port traffic system and the power system is achieved. According to one aspect of the present application, in some embodiments, the cost of energy storage configuration also needs to be taken into account.

[0037] Specifically, as shown in Figure 1 the method of the present embodiment includes the following steps: Step 1: Obtain multi-source data of the traffic system and the power system, perform multi-source data collection and baseline modeling, and obtain port baseline data.

[0038] The specific implementation of this step, the data types involved (traffic, power, environment), and the port baseline data generated after modeling (preferably including baseline port power sequence P port_base , device adjustable constraint prototype ProtoConstraints, and rolling prediction set Forecast bundle ) have been described in detail in Embodiment I.

[0039] Step 2: Based on the port baseline data, perform slight peak shifting on the traffic-side operation phases to perform power spectrum shaping, and obtain traffic-side shaped data.

[0040] Operation phase refers to a specific working link of the main energy-using equipment in the port (such as the quay crane, electric truck), such as the lifting, rotation, and lowering of the quay crane. Slight peak shifting refers to adjusting the starting time of these operation phases by a small amount (for example, several seconds to tens of seconds) without violating the physical and process constraints defined by the device adjustable constraint prototype (for example, the maximum allowed micro-shift time window, the sequence of operations).

[0041] Power spectrum shaping is the technical goal to be achieved in this step. In the present embodiment, the total power (P port_base ) of the port often contains a large number of high-frequency power fluctuations with short duration but high amplitude caused by the overlapping of multiple device start-stop (especially quay crane lifting). These high-frequency fluctuations are extremely detrimental to the cycle life of energy storage devices and the stability of the power grid.

[0042] In order to perform high-frequency noise reduction, low-frequency is left to energy storage. By actively and intelligently shifting the operation phases (for example, shifting two simultaneously started quay crane lifting operations by a few seconds), the total power curve can be significantly smoothed, i.e., the high-frequency energy in the total power sequence spectrum above a certain target cutoff frequency (determined by the capacity of the energy storage device) is reduced, without affecting the total operation throughput.

[0043] Using P port_baseProtoConstraints and Forecast bundle As input, it is realized by solving an optimization problem aiming at minimizing high-frequency energy and minimizing operation cost and risk (see embodiment 3 for details).

[0044] This step gets traffic-side reshaped data, which specifically includes: reshaped port power sequence (P port_shaped , i.e. smoothed total power curve), adjustable slope upper limit sequence (R max_series , i.e. maximum allowed ramp rate at each time), and adjustable power boundary sequence (DeltaP port_shaped , i.e. allowed up / down adjustment range at each time). max_series port_shaped

[0045] Step 3: Based on traffic-side reshaped data and port baseline data, encapsulate traffic-side executable flexibility, cost and risk into a standardized three-signal interface.

[0046] Power-side dispatch system (e.g. EMS) is difficult and unnecessary to handle traffic-side massive device states and complex operation logic (ProtoConstraints). Therefore, this step translates and encapsulates traffic-side flexibility potential from power-side perspective.

[0047] The standardized three-signal interface refers to a structured data packet (also called RG-FLEX interface in the scheme), which contains three orthogonal signal components: Flexibility capacity (quantity and shape): based on traffic-side reshaped data (P port_shaped , R max_series , DeltaP max_series ), quantify the up / down power limit, up / down slope limit and minimum sustainable duration that traffic-side can provide at each time slice in the future.

[0048] Flexibility cost (price): based on port baseline data (especially congestion prediction in ProtoConstraints and Forecast bundle ), map the internal cost (e.g. operation delay, yard congestion risk) of traffic-side caused by invoking the above flexibility capacity into a piecewise convex cost function.

[0049] Flexibility risk (reliability): based on port baseline data (especially uncertainty prediction in Forecast bundle , such as weather, photovoltaic prediction deviation) and historical realization record (backfill data from step 5), estimate a reliability coefficient sequence to quantify the realizable confidence of the above flexibility capacity commitment (for example, a value between 0 and 1). ​​

[0050] Step 4: Based on the standardized three-signal interface, the energy storage device model, and the grid constraint set, perform power-side closed-loop solving to obtain power-side feedback signals and power-side execution scheduling.

[0051] In this step, the power-side system receives the standardized three-signal interface. The power-side no longer cares about the specific operations of the traffic side, but only cares about the available flexibility, calling cost, and available probability defined by the three signals.

[0052] The energy storage device model defines the physical boundaries of the energy storage, such as charging and discharging efficiency, maximum charging and discharging power, energy limit (SOC upper and lower limits), and cycle life and degradation cost model. The grid constraint set defines the safety boundary of the grid, such as the power limit for interaction with the upper grid, the power flow and node voltage constraints within the port area, and the grid-side power ramp rate limit.

[0053] The closed-loop solving of the power side refers to running a joint optimization model (see Example Five for details), which takes the three signals as input, takes the energy storage model and grid constraint set as constraints, and aims to minimize the total cost (e.g., the sum of the purchase power cost + energy storage degradation cost + the cost of calling traffic flexibility) in a rolling time domain.

[0054] The output of this step is divided into two parts: Power-side execution scheduling: This is the instruction that needs to be executed immediately in this rolling period, specifically the energy storage scheduling scheme (ESS schedule ), which is the charging and discharging power timing of the energy storage system in the future period (e.g., 15 minutes or 1 hour).

[0055] Power-side feedback signals: These are the price and shape signals that need to be fed back to the traffic side to guide the optimization of the next rolling period. Specifically, it preferably includes: power shadow price (μ P ), slope shadow price (μ R ), and plateau power request (Plateau request ).

[0056] Step 5: Based on the power-side feedback signals and power-side execution scheduling, perform rolling verification and data backfilling to generate a backfilling data set.

[0057] First, execute the power-side execution scheduling, i.e., issue the energy storage scheduling scheme ESS schedule to the energy storage controller PCS for execution.

[0058] At the same time, the traffic side system receives the power-side feedback signals. For example, Plateau request (the plateau power request) will be used as one of the optimization objectives (i.e., to minimize P port_shaped(Approaching the platform within a specified time period); while shadow electricity prices are used to guide the update of the cost function on the transportation side.

[0059] Rolling verification refers to the system monitoring actual operating indicators in real time during the execution process, such as: actual port area power high-frequency energy, actual energy storage cycle depth and number of cycles, whether voltage over-limit or default occurs, and whether traffic throughput remains at the target level.

[0060] Data backfeeding refers to packaging the results of the above verification and monitoring into a backfeed dataset. _bundle This dataset details the execution results, deviations, and actual fulfillment rates of this round of optimization (e.g., a promise of 500kW flexibility was made, but only 400kW was actually fulfilled).

[0061] Step 6: Update the multi-source data aggregation and baseline modeling using the refeedback dataset. This step will update the output of Step 5 (Feedback). bundle Feedback is sent back to step 1.

[0062] This update is multi-layered: Update prior data: Recycle and update the P data in Example 1 using the actual power curves and operation logs from this round of execution. port_base (Baseline port area power sequence).

[0063] Update model parameters: Achieve deep loop closure.

[0064] Calibrate the reliability model: Use historical fulfillment records from the recharge dataset to calibrate the reliability coefficient (Conf) used in step 3 (Example 4). series The model is as follows: If the system finds that the realization rate is always low when the photovoltaic prediction deviation is large, it will automatically lower the reliability coefficient in that scenario.

[0065] Calibrate the cost model: Use the actual congestion and delay statistics in the recharged dataset to re-estimate the piecewise convex cost function (φ) in step 3 (Example 4). _segments The marginal cost slope of ).

[0066] In summary, this embodiment constructs a sequential closed-loop optimization system through the above six steps. Steps 2 (spectral shaping) and 3 (three-signal encapsulation) realize flexible supply-side reform on the transportation side; step 4 realizes efficient demand-side solution on the power side; while steps 5 and 6 construct a data-driven feedback closed loop, enabling the system to continuously self-calibrate and evolve.

[0067] Embodiment 3 describes the traffic-side power spectrum shaping and micro-phase optimization method, i.e. the refinement of Step 2 in Embodiment 2. The purpose is to proactively eliminate the high-frequency power fluctuations harmful to the grid and energy storage at the traffic side (source), rather than passively erasing them by energy storage.

[0068] In this embodiment, as shown in Figure 2 , the traffic-side operation phase is slightly staggered to perform power spectrum shaping, obtaining traffic-side shaped data, including: Step 301: According to the equipment adjustable constraint prototype contained in the port baseline data, the micro-phase variable of the traffic-side operation and the feasible region constraint are established.

[0069] In this step, the micro-phase variable refers to a decision variable δ i defined for each schedulable operation event i (e.g. the kth lifting operation of a shore crane) in the port baseline data (from Embodiment 1). The physical meaning of this variable δ i is the time micro-shift of the operation event i relative to its baseline start time t0 i , usually in seconds. For example, δ i = +10 seconds means that the operation is delayed by 10 seconds, and δ i = -5 seconds means that the operation is performed 5 seconds earlier.

[0070] The feasible region constraint refers to a set of mathematical constraints that must be met to ensure that any combination of values of the micro-phase variable δ i is physically and technologically executable. This constraint set is constructed based on the equipment adjustable constraint prototype (ProtoConstraints) in Embodiment 1. Specifically, this feasible region constraint preferably includes the following categories: Phase window constraint (boundary constraint): Each micro-phase variable δ i itself has an allowed offset range, for example -window±(i)≤δ i ≤+window±(i). For example, the micro-movable window of a shore crane operation is ±30 seconds, so its δ i must be taken within the interval [-30, +30] seconds.

[0071] Order and interval constraint (linear inequality constraint): For two adjacent operation events i and j on the same device (e.g. the same shore crane or the same charging pile), the execution order and minimum safety interval must be guaranteed. This is modeled as: (t0 i +δ i )+duration i +sep _min (i,j) ≤ (t0 j +δj ). Where duration i is the duration of event i, sep _min (i,j) is the minimum separation (e.g., safety time for a shore crane trolley turn) from the end of event i to the start of event j.

[0072] Non-interruptible constraint (logical constraint): For certain critical segments of work (e.g., an ongoing lift main hoist segment) that are marked as lock _flag = 1, their interior is not allowed to be split and must be moved as a whole by δ i .

[0073] Freeze window constraint (external constraint): The system determines some freeze windows Freeze bundle based on the rolling forecast set (Forecast _windows from Example 1) (e.g., the last 5 minutes before a ship is about to depart). If the planned execution time (t0 i + δ i ) of any work event i falls into a Freeze _windows , then δ i must be forced to 0 or adjusted to outside the window.

[0074] Pile position and queuing constraint (resource constraint): For a charging work event i of an electric truck, its start charging time (t0 i + δ i ) must match the free time of the charging pile position predicted in the rolling forecast set to avoid causing charging pile queuing or conflict.

[0075] Step 302: Analyze the physical limits of the energy storage device model to determine the target cutoff frequency corresponding to the power fluctuation that the energy storage can handle.

[0076] Since the energy storage device (ESS) has limited response capability for high-frequency (i.e., rapidly changing) power fluctuations, and such high-frequency charging and discharging will seriously damage the service life of the energy storage. The purpose of the present invention is to actively filter out these high-frequency components that the energy storage does not want to handle through the micro-phase adjustment of step 301.

[0077] The energy storage device model (ESS_model) defines the physical limits of the energy storage, including key parameters such as maximum power ramp rate (e.g., kW / s), maximum charging and discharging power (kW), and energy capacity (kWh).

[0078] Based on the above limits, this step determines a target cutoff frequency f c (or its corresponding discrete frequency index k c ). This f cThe physical meaning is the boundary line between the responsibilities of the transportation side and the energy storage side.

[0079] Higher than f c Fluctuations (e.g., f>0.02Hz, i.e., fluctuations with a period of less than 50 seconds): are considered high-frequency noise and are undesirable for energy storage. These fluctuations must be minimized by this embodiment (traffic side spectrum shaping).

[0080] Below f c Fluctuations (e.g., f < 0.02 Hz, i.e., fluctuations with a period greater than 50 seconds): are considered low-frequency fluctuations that energy storage systems can safely and economically smooth out.

[0081] Step 303: Construct a spectral shaping optimization objective and solve for the micro-phase variables to minimize the high-frequency energy above the target cutoff frequency in the total power sequence of the port area.

[0082] In this step, firstly, the high-frequency energy J HF In the frequency domain, it is defined as J HF =∑ k>kc |FFT{P port (t)}[k]| 2 Where k is the discrete frequency index, k c It is the cutoff frequency index, FFT is the Fast Fourier Transform, P port (t) is the total power timing of the port area, while P port (t) is a function of all micro-phase variables δ.

[0083] Directly optimize the above J HF This is extremely difficult computationally. Therefore, the present invention preferably employs a combined approach to transform it into a time-domain optimization problem that can be solved efficiently.

[0084] Specifically, the construction and solution process includes constructing a linearized power model of the total power of the port area, such as... Figure 4 As shown: Based on the micro-phase variable δ and the event power template p in the port area baseline data i (t) (from Example 1, P) event_templates A linearized power model of the total power of the port area is constructed.

[0085] P port (t;δ) (i.e., total power P) port As a function of time t and all minor shifts δ, P can be approximated by a first-order Taylor expansion as: port (t;δ)≈P port_base (t)+∑ i δ i ·p' i (t); where P port_base(t) is the baseline harbor power sequence (from Example 1); δ i is the micro-phase variable of the job event i; p' i (t) is the pre-computed micro-shift δ i of event i. The linearized model is valid when δ i is small (e.g., within ±90 seconds).

[0086] Convert the high-frequency energy into a time-domain quadratic form: According to the target cutoff frequency f c , design a discrete high-pass filter h HP (e.g., FIR filter) and construct its corresponding convolution matrix H HP .

[0087] In the time domain, the high-frequency energy J HF can be approximated as the energy (i.e., square of the two-norm) of the total power P port after passing through the high-pass filter h HP : J HF ≈∑ t (h HP P port (t; δ)) 2 .

[0088] Apply the convolution matrix to the linearized power model, i.e., J HF (δ) ≈ ||H HP ·P vec (δ)||2 2 , where P vec (δ) is the vector form of the linearized power model above.

[0089] Combining the two, the high-frequency energy J HF is finally expressed as a time-domain quadratic form with respect to the micro-phase variable δ: J HF (δ) = (Aδ - b) T (Aδ - b) + d; where δ is the decision vector composed of all δ i ; A and b are constant matrix and vector pre-computed from H HP , P port_base , and p' i . This quadratic form is a convex function, which is extremely easy to solve.

[0090] Construct the combined optimization objective: Minimizing J HF may cause severe job delays. Therefore, the present invention constructs a combined optimization objective, such as Figure 3shown.

[0091] First, based on the equipment adjustable constraint prototypes in the port baseline data and the rolling prediction set, the operation cost and risk function J ops (δ) of the traffic side operation is evaluated

[0092] The operation cost and risk function at least includes: Operation delay cost: for example, for the operation i on the critical path, the cost cost _delay_i = c _ delay_i ·max(0, δ i ), where c _delay_i is the cost of unit time delay.

[0093] In-process overflow risk: if δ i causes the predicted yard in-process (from Forecast bundle ) to exceed the safety threshold, a penalty term is applied.

[0094] Pile position congestion penalty: if δ i causes the electric truck to request charging when the charging pile is congested (from Forecast bundle ), a penalty term is applied.

[0095] Preferably, J ops may also include a phase synchronization penalty (avoid multiple quays simultaneously lifting) or a platform compliance term (for compliance with the platform request issued in the previous round of Example Five).

[0096] Finally, the high-frequency energy J HF (δ) and the operation cost and risk function J ops (δ) are fused to form a combined optimization objective J: min δ J = J HF (δ) + α·J ops (δ); Wherein α is a weighting coefficient for balancing power quality (J HF ) and operation efficiency (J ops ).

[0097] The two-stage solver is used to solve the process as follows: In order to solve the above combined optimization objective efficiently and robustly, the present application preferably adopts a two-stage solving strategy.

[0098] Based on the feasible region constraints in step 301 (which have been represented as Gδ ≤ h,l ≤ δ ≤ u), a feasible region projection operator Proj c(·). This operator can project any infeasible solution z back into the feasible region C.

[0099] Solve the relaxation of the spectral shaping optimization objective on the continuous domain. That is, solve min J subject to Gδ ≤ h, l ≤ δ ≤ u. This is a convex quadratic program (QP) problem that can be solved quickly using standard solvers (e.g. OSQP, Gurobi) to obtain a continuous solution δ cont (e.g. δ i = -10.73 seconds).

[0100] Apply the feasible region projection operator to the continuous solution to repair constraints (i.e. δ proj = Proj c (δ cont )), and perform discretization (e.g. round to the nearest second-level granularity δ star = -11 seconds) to generate the final executable phase shift scheme Phase dither_plan .

[0101] Further, as a robustness design, after obtaining δ star , the system will check whether its J HF has already fallen below the high frequency energy budget HF budget set in step 302. If not, the system can automatically increase the weights of α or γ (J HF ) and return to solve again, realizing an automatic parameter tuning closed loop within S2.

[0102] Step 304: Generate traffic-side reshaped data according to the phase shift scheme obtained by solving and the port baseline data.

[0103] After obtaining the optimal phase shift scheme δ star in step 303, this step uses the scheme to generate traffic-side reshaped data for subsequent steps (Embodiment Four) to use.

[0104] Traffic-side reshaped data includes: Reshaped port power sequence (P port_shaped ): Substitute δ star into the original, nonlinear event power template to perform accurate superposition P port _ shaped (t) = ∑ i P event (t - (t0 i + δ star _ i )), to obtain the final smoothed power curve.

[0105] Adjustable slope upper limit sequence (Rmax_series ): Compute P port_shaped The maximum up / down slope in each time slice as a constraint for power side scheduling.

[0106] DeltaP max_series ): Evaluate how much power space is left in each time slice for up / down regulation (originated from DeltaP star Based on the scheme, how much power space is left in each time slice for up / down regulation (originated from DeltaP i Not fully utilized its ±window ± Boundary).

[0107] These three sets of data constitute the traffic side executable flexible amount and shape base, which will be used as the input of example four (three signal packaging).

[0108] Example four, three signal standardized packaging of traffic side flexible capacity.

[0109] Since the power side system (such as the scheduling center) should not and cannot understand the complex traffic operation logic (such as the phase of the shore crane, the queuing of the truck) in example three, this example serves as a translator to standardize the packaging of the flexible potential of the traffic side from three dimensions that the power system can understand: amount (FlexEnvelope), price (φ_segments), and risk (Conf_series).

[0110] In this example, the standardized three signal interface includes: Signal one (quantitative capacity): based on the flexible envelope generated after the traffic side shaping data, used to quantify the up / down regulation power limit, up / down slope limit, and minimum sustainable duration of the traffic side.

[0111] This flexible envelope FlexEnvelope is a quantification of the traffic side physical adjustable boundary. It is constructed based on the output of example three (P port_shaped , R max_series , DeltaP max_series ).

[0112] Specifically, for each time slice t in the future rolling time domain: Up / down regulation power limit (P up_max (t), P down_max (t)): based on DeltaP _max_series .

[0113] Up / down slope limit (R up_max (t), R down_max (t)): based on R _max_series .

[0114] Minimum sustainable duration (T dur(t) is derived based on the remaining movable margin and the non-interruptible segment constraint.

[0115] Preferably, in generating this envelope, the present invention also introduces a robustness treatment. The system reads the rolling Forecast bundle set in Embodiment One, and actively removes the non-fulfillable part from the FlexEnvelope, by removing the flexibility corresponding to the unavailable equipment (e.g. equipment under maintenance, occupied berth) from the Forecast max_series set. For example, even though the DeltaP bundle set shows that a certain shore crane has 500kW regulation capacity, if the Forecast up_max set shows that it is under planned maintenance at time t, the corresponding 500kW in P _t (t) will be removed, thus ensuring the truthfulness of signal one (capacity).

[0116] Signal two (quantified cost): Based on the port baseline data and the job cost log and shaping quality metrics derived in the power spectrum shaping step, the mapped piecewise convex cost function, is used to quantify the marginal cost of invoking flexible resources.

[0117] The piecewise convex cost function φ _t (δP) is the pricing of signal one (capacity). It tells the power side: at time t, how much internal cost (e.g. delay, congestion) of the traffic side needs to be paid for invoking δP power of flexibility.

[0118] The generation of this function is a multi-layer mapping process: Step 401: Aggregate the traffic side job cost and estimate the marginal cost slope based on the job cost log and the Forecast

[0119] set in the port baseline data. _cost_log This step reads the job cost log (Ops i ) generated in Embodiment Three. This log records the marginal cost corresponding to each δ j . This step aggregates it from event level to time slice level, and estimates the marginal cost slope s 0 .

[0120] Step 402: Apply spectrum shaping revenue discount to the marginal cost slope using the shaping quality metrics.

[0121] If the spectrum shaping effect in Embodiment Three is good (i.e. the shaping quality metrics Shaping_quality_metrics shows that the high frequency energy has been greatly reduced), then the remaining power fluctuation has been reduced in toxicity to the grid, and the marginal cost of invoking flexibility should also be reduced accordingly.

[0122] Specifically, s j 0 is multiplied by a spectrum shaping revenue discount factor η spec (η spec ∈(0, 1], e.g. 0.9), resulting in a discounted slope s j 1 = η spec · s j 0 .

[0123] Step 403: Apply risk amplification to the discounted slope based on uncertainty assessment in the port baseline data.

[0124] This step reads the rolling forecast set (Forecast bundle ) of Example One. If the forecast shows high uncertainty in the future (e.g. bad weather, large fluctuation in job arrivals), it means that there is high risk in honoring the flexible commitment.

[0125] Specifically, s j 1 is multiplied by a risk amplification factor η risk (η risk ≥ 1, e.g. 1.2), resulting in s j 2 = η risk · s j 1 .

[0126] Step 404: Perform convexity alignment on the slope and integrate to get the base piecewise convex cost.

[0127] Due to the computation in the above steps, the slope sequence s j 2 may not satisfy convexity (i.e. s j+1 <s j ). This step enforces non-decreasing slope by a sorting regression or cumulative maximization algorithm (i.e. convexity alignment), ensuring s j+1 ≥ s j . This is crucial for guaranteeing global optimality of the power side (Example Five) optimization solution.

[0128] Step 405: Apply reliability coefficient sequence to the base piecewise convex cost for conservative treatment, generating the piecewise convex cost function.

[0129] This step couples signal two (cost) with signal three (reliability) below. If the reliability Conf(t) is low (e.g. 0.8), it means that the flexibility is likely to be unfulfilled, in which case its offer should be raised to penalize such unreliability.

[0130] Specifically, the final cost function φ_teff (δP) is conservatively scaled by dividing by the reliability factor: φ _teff (δP) = φ t (δP) / max(Conf(t),υ). Where υ is a small positive number to prevent division by zero.

[0131] Signal Three (Quantified Risk): and the estimated reliability factor sequence based on the uncertainty in the port baseline data and historical fulfillment records, to quantify the confidence of the flexible commitment.

[0132] The reliability factor sequence Conf(t) is a time series between [0,1] to quantify the confidence of signal one (capability). Conf(t)=0.95 means that there is 95% probability that the flexibility of the commitment at time t can be fulfilled.

[0133] The generation process of this sequence is a data-driven closed-loop calibration process: Step 406: Extract uncertainty factors from the port baseline data.

[0134] The system extracts all risk sources that affect the flexibility fulfillment from the rolling forecast set (Forecast bundle ) of embodiment one. The uncertainty factors at least include: Weather and photovoltaic prediction bias: For example, the variance of photovoltaic prediction or the probability of weather sudden changes (such as gusts, heavy fog).

[0135] Uncertainty of pile position occupation and queuing: For example, the volatility of electric truck charging demand prediction.

[0136] Job arrival and completion time fluctuation: For example, the deviation between actual berthing time and planned time of a ship.

[0137] Step 407: Fuse the uncertainty factors to generate a preliminary confidence time series.

[0138] The multiple (e.g. N) uncertainty factors f1,f2,…,f N extracted in step 406 are merged to generate a single confidence score Conf preliminary (t).

[0139] Optionally, the fusion method can be: Optional solution A (weighted fusion): A hierarchical weighting method is used to weighted average each factor.

[0140] Optional solution B (model fusion): A logistic regression model is used, which takes N factors as input and outputs a probability value between [0,1]. The model is preferably trained offline based on historical data.

[0141] Step 408: Based on historical redemption records in the port baseline data, perform calibration on the preliminary confidence timing, minimize the consistency residual of the preliminary confidence and the historical redemption probability, and obtain a reliability coefficient sequence.

[0142] This step is the key closed loop to achieve system self-adaptation. The system will call historical redemption records (which are generated and stored in the port baseline data by steps 5 and 6 (Feedback bundle ) of embodiment two).

[0143] The system compares: In the past, all Conf preliminary (t) is actually 90%.

[0144] If there is a consistency residual (for example, the prediction is 0.9, but the actual historical redemption probability is only 0.8), the system will apply a calibration model (such as order-preserving regression or Platt scaling) to correct Conf preliminary (t) so that the output Conf(t) is consistent with the actual future redemption probability in a statistical sense.

[0145] Finally, this embodiment packages the above generated signal one FlexEnvelope, signal two φ _segments and signal three Conf _series , together with state information such as PhaseSlots (phase freezing window), into a standardized, versioned data interface, which is directly called by the power side closed loop solver of embodiment five.

[0146] Embodiment five provides a power side closed loop solving and shadow price feedback method, which describes how the power system (as a flexible consumer) efficiently, safely and economically utilizes the standardized traffic side flexible resources provided by embodiment four (three-signal interface).

[0147] In this embodiment, the power side closed loop solving includes: Step 501: Analyze the standardized three-signal interface to obtain the flexible envelope and the reliability coefficient sequence; apply the reliability coefficient sequence to conservatively reduce the upper and lower power limits and the uplink and downlink slope limits defined in the flexible envelope, and generate a conservative flexible envelope; the conservative flexible envelope is used as the constraint set for the power side closed loop solving.

[0148] In this step, the power side solver first receives the standardized three-signal interface from embodiment four, which includes the flexible envelope FlexEnvelope (signal one), the piecewise convex cost function φt (signal two), and the reliability coefficient sequence Conf(t) (signal three).

[0149] The conservative reduction is a key action for the power side to hedge risks. Since Conf(t) quantifies the redeemable confidence of the flexible commitment (signal one) (e.g. Conf(t) = 0.9 means there is 90% probability to be redeemed), the power side cannot rely on this commitment 100% when making scheduling plans, but must make a reduction.

[0150] In particular, this reduction is realized by a reliability reduction function β(Conf), which is preferably a non-decreasing function, β: [0, 1]→ [0, 1].

[0151] Option A (linear reduction): β(Conf) = Conf.

[0152] Option B (non-linear reduction): β(Conf) = Conf γ with γ > 1 (e.g. γ = 1.5 or 2), which means the system punishes low reliability more (e.g. Conf = 0.8, β ≈ 0.72).

[0153] Option C (piecewise reduction): e.g. if Conf < 0.7, then β = 0.5; if 0.7 ≤ Conf < 0.9, then β = 0.8; if Conf ≥ 0.9, then β = 0.95.

[0154] The process of generating the conservative flexible envelope FlexEnvelope eff is as follows (for each time slice t): Pup eff (t) = β(Conf(t)) · Pup max Pdown eff (t) = β(Conf(t)) · Pdown max Rup eff (t) = β(Conf(t)) · Rup max Rdown eff (t) = β(Conf(t)) · Rdown max (t). At the same time, as in embodiment four step 405, the piecewise convex cost function has also been made conservative to eff (t, ΔP) = φ t (ΔP) max(Conf(t), ∈).

[0155] This set of conservative FlexEnvelope eff and φ _segments_eff will serve as hard constraints and cost terms for the optimization solution in the subsequent steps.

[0156] Step 502: Construct a joint objective function J aiming to minimize the total cost over the rolling horizon; the total cost at least aggregates: grid purchase cost derived based on the grid constraint set; piecewise convex cost function; storage degradation cost derived based on the storage device model; and grid-side ramp rate penalty derived based on the grid constraint set.

[0157] This step constructs the core optimization problem for the power-side rolling dispatch. The decision variables of this problem preferably include: grid purchase power time series P grid (t) over the future one rolling window (e.g., 1 hour ahead, with one sample point per minute); storage charging power P ess ch (t) over the future one rolling window; storage discharging power P ess_dis (t) over the future one rolling window; storage energy state E ess (t) over the future one rolling window; and flexible occupancy amount ΔP req (t) requested to the transportation side.

[0158] The joint objective function J power aims to minimize the total cost: J power = min∑ t [Cost grid (t)+Cost flex (t)+Cost degradation (t)+Cost ramp (t)+Cost violation (t)] Wherein, the specific implementation of each cost is as follows: Grid purchase cost Cost grid (t): Cost grid (t)=Tariff(t)·P grid (t)·dt. Wherein Tariff(t) is the time-of-use tariff (e.g., peak, flat, valley tariff) obtained from the tariff and carbon price sequence (Tariff bundle ); dt is the time step.

[0159] Piecewise convex cost function Cost flex (t): Cost flex (t)= φ eff (t,ΔPreq(t)). Wherein φ eff is the conservatized piecewise convex cost function provided by step 501 (originated from embodiment four). This term will monetize the internal cost of transportation flexibility and incorporate into the unified optimization of the power side.

[0160] Storage degradation cost Cost degradation (t): Cost degradation ​(t) = c deg · f deg (t) = P ess_ch (t) - P ess (t - 1) = E dis (t) - E ess (t - 1) = E

[0161] where c deg is the equivalent replacement cost per unit degradation (e.g., per standard cycle) of the energy storage (e.g., lithium battery); f deg is a degradation model, optionally, f deg may be a linear model based on equivalent full cycle number, e.g., f deg = (η ch · P ess_ch (t) + P ess_dis (t) / η dis ) · dt / (2 · E capacity ).

[0162] Preferably, f deg may be a more complex piecewise linear or nonlinear function related to depth of discharge (DOD) and charge-discharge rate (C-rate), which is calibrated based on the lifetime curve of the energy storage system model (ESS_model).

[0163] Grid-side ramp rate penalty Cost ramp (t): Cost ramp (t) = c ramp · |P grid (t) - P grid (t - 1)|. Where c ramp is a penalty coefficient. This term is used to incentivize the power side to generate a smoother power purchase curve, reducing the impact on the upper grid, and is the key to achieving grid-friendliness.

[0164] Exceeding limit penalty Cost violation (t) (optional): Cost violation (t) = c cvolt · max(0, U min - U i (t)) + … U n (t)). This term is used to handle soft constraints, such as slight voltage exceeding limits, where c volt is a very high penalty coefficient.

[0165] Step 503: Build a constraint set. To ensure that the optimization problem in step 502 is solvable and physically feasible, a complete constraint set must be built, which is based on the energy storage system model and the grid constraint set.

[0166] System power balance constraint (core constraint): P grid (t)+P pv (t)+P ess_dis (t)=P load_other (t)+P port_shaped (t)+P ess_ch (t)+ΔP req (t). Wherein, P port_shaped (t) is the shaped baseline power output in Example 3; ΔP req (t) represents the incremental flexibility request for this baseline in this round of optimization decisions. pv (t) and P load_other (t) is a predicted value from Example 1.

[0167] Energy storage constraints (from ESS_model): Energy state transition: E ess (t+1)=E ess (t)+η ch ·P ess_ch (t)·dt-(1 / η dis )·P ess_dis (t)·dt-κ self ·E ess (t)·dt. Where η ch and η dis These are the charge / discharge efficiency, κ self It is the self-discharge rate.

[0168] Energy and power boundary: E min ≤E ess (t)≤E max (For example, SOC 20% to 90%); 0 ≤ P ess_ch (t)≤P ch_max ; 0≤P ess_dis (t)≤P dis_max .

[0169] Charge-discharge mutual exclusion constraint: Option A (Mixed Integer P): Introduce a binary variable z(t) ∈ {0, 1}, such that P ess_ch (t)≤M·z(t)andP ess_dis (t)≤M·(1-z(t)). This method is accurate, but the solution speed is slow.

[0170] Option B (Convex Relaxation): Add P to the objective function ess_ch (t)⋅P ess_dis The penalty term of (t) can be used, or its convex relaxation form can be adopted. This method is fast and suitable for real-time rolling scheduling.

[0171] Grid and power flow constraints (from Grid_constraints): Hard constraint on net-side gradient (optional): |P grid (t)-P grid (t-1)∣≤R net_max .

[0172] Power flow and voltage constraints: Option A (Simplified Model): Only constrain the total power P grid (t)≤P grid_max， Ignoring internal trends.

[0173] Option B (Preferred Model): Employ the power flow model of the port area distribution network to ensure the internal node voltage U i (t) and line power flow P ij (t) does not exceed the limit. To ensure solution speed, this power flow model preferably adopts a linearized power flow approximation or a second-order cone programming (SOCP) DistFlow model. That is, U min ≤Ui(t)≤U max And |P ij (t)∣≤P ij_max .

[0174] Flexible constraints (from step 501): P down_eff (t)≤ΔP req (t)≤P up_eff (t).

[0175] Step 504: After solving the joint objective function, extract the power shadow price from the dual solution of the system power balance constraint; extract the slope shadow price from the dual solution of the grid-side ramp rate constraint; and generate a platform-based power request based on the optimal scheduling trajectory obtained by solving the problem, combined with the power shadow price and the slope shadow price.

[0176] This step describes how to extract the power-side feedback signal for feedback from the optimization results of steps 502-503.

[0177] Solving this optimization problem (e.g., a quadratic programming problem (QP) or a second-order cone programming problem (SOCP)) yields a set of optimal scheduling trajectories (i.e., P). grid (t),P ess_ch (t),…,ΔP req (optimal timing of (t)).

[0178] The shadow price, or Lagrangian multiplier, is extracted from the dual solution of the solver.

[0179] Power shadow price μP(t): This is the dual solution extracted from the system power balance constraint in step 503.

[0180] Interpretation: μP(t) means that at time t, if the system (e.g. through traffic-side flexibility) were to provide 1 kW of additional available power, the system total cost J power would decrease by μP(t) yuan.

[0181] At t=10:00, the electricity purchase price Tariff(t)=1.0 yuan / kWh. If the solution yields μP(t)=1.8 yuan / kWh, this means that at this time the system is in a highly congested state (e.g. ramp rate limited or energy storage exhausted), and the system is willing to purchase flexibility at an internal price of 1.8 yuan, much higher than the grid price. This μP(t) signal will strongly direct the traffic-side to provide upward flexibility at the next round t=10:00.

[0182] Ramp rate shadow price μR(t): This is the dual solution extracted from the grid-side ramp rate constraint (hard or soft) in step 503. μR(t) means that at time t, if the grid-side ramp rate upper limit R net_max were relaxed by 1 kW / s, the system total cost J power would decrease by μR(t) yuan. μR(t) is mostly 0, but when P grid (t) is close to the ramp upper limit, μR(t) will spike. This spike signal explicitly tells the traffic-side: at this moment, the grid’s ‘smoothness’ resource is extremely expensive.

[0183] Generate plateau power request Plateau request . This request is issued by the power-side to the traffic-side (embodiment three) for the next rolling period.

[0184] Preferably, the generation algorithm is: based on the optimal dispatch trajectory P grid (t) (i.e. the original solution) solved, perform piecewise constant approximation on it. This is a secondary optimization problem, aiming to find a set of plateaus P target and their intervals [t1, t2] that minimize ∑(P target -P grid (t)) 2 (tracking error) and λTV·(number of plateau switches) (total variation).

[0185] Generation algorithm (optional): it is also possible to generate based on shadow prices. In periods when both μP(t) and μR(t) are high, force a flat P target， lower than the current baseline to be generated to relieve system stress.

[0186] The final output power side feedback signal of this embodiment is {μP(t),μR(t),Plateau request}; the power side execution schedule is {ESS schedule ,Residual _series} (where Residual _series is the residual fluctuation that needs to be compensated by energy storage).

[0187] Embodiment six describes the closed-loop execution, monitoring, feedback, and learning mechanism of the sequential architecture of the present application.

[0188] Specifically, the method of this embodiment includes: Step 601: Perform rolling verification and data backfill according to the power side feedback signal and the power side execution schedule.

[0189] This step describes the execution and monitoring link within a rolling period. First, the system executes the schedule output by Embodiment Five: Power side execution: issue the energy storage scheduling scheme ESS schedule to the energy storage power converter (PCS) and the battery management system (BMS).

[0190] Preferably, the energy storage control adopts a hierarchical control architecture: the outer loop controller is responsible for tracking the grid-side power target or node voltage target implied by ESS schedule ; the inner loop controller is responsible for high-speed response to ensure that the SOC, current, temperature, etc. of the energy storage itself do not exceed the safety boundary.

[0191] Traffic side receives feedback: the traffic side system receives the power side feedback signal {μP,μR,Plateau request}. These signals are not used for immediate execution in this period, but are used as price and pattern guidance, input into Embodiment Three (spectrum shaping optimization) of the next rolling period, as part of its J ops (operation cost function).

[0192] Secondly, the system performs rolling verification, i.e. real-time monitoring and deviation analysis of the execution process.

[0193] Monitoring content: the system collects the following through measurement equipment in real time: actual port total power, actual grid purchase power P grid_actual , actual node voltage U actual、 , and actual energy storage state ESS actual .

[0194] Deviation analysis: Minor deviation: if P grid_actual deviates from the plan, the energy storage system (inner loop control) makes small adjustments within the safety range for compensation.

[0195] Severe deviation (backoff strategy): In the event of a major disturbance (e.g., unexpected tripping of a large quay crane, or U... actual If the system falls below the safety lower limit, it triggers a rollback strategy, immediately suspends the current optimized scheduling, rolls back to the previous known safe operating platform, and sends a warning of flexible request failure to the scheduling center.

[0196] Log recording: All execution data, deviations, alarms, and rollback events are recorded in detail.

[0197] Finally, based on the validation results, a feedback dataset is generated. bundle At the heart of this dataset is an execution monitoring report, Metrics_report, which summarizes the key performance indicators (KPIs) for the current rolling cycle, preferably including: Spectral shaping effect: actual high-frequency energy (kW·h) and decrease rate (%), actual 95th percentile ramp rate (kW / s).

[0198] Energy storage health: Equivalent full cycle (EFC) number of energy storage cycles.

[0199] Power grid safety: number of times node voltage exceeds the limit and cumulative duration, and percentage of line thermal limit exceeds the limit.

[0200] Economic efficiency: actual electricity purchase cost, platform achievement rate (%).

[0201] Traffic performance: throughput retention rate (%).

[0202] Fulfillment rate assessment: Detailed records are kept at time t, for the flexible P promised in Example 4. up_max (t) and the flexible ΔP requested by the power side req (t), and the flexible ΔP ultimately realized on the transportation side. actual (t) is used to calculate the actual realization rate.

[0203] Step 602: Update the multi-source data aggregation and baseline modeling using the re-feedback dataset.

[0204] This step is the outermost closed loop of the sequential architecture of this invention (i.e., feedback from step 5 back to step 1), used for the system to achieve adaptive learning and model calibration.

[0205] Updates are multi-layered, including: Level 1 (Data Update): Feedback bundle The actual measurement timing (e.g., P) port_actual The data was fed back into the database of Example 1 to update the baseline port area power sequence P. _port_base Historical statistics.

[0206] Level two (model parameter self-adaptive calibration, preferred approach): this is the preferred embodiment of the invention, which utilizes the realization rate evaluation and KPIs in Feedback bundle to automatically calibrate the key model parameters in Embodiment four.

[0207] According to one aspect of the present application, the calibration of the reliability model is as follows: The system compares the actual realization rate recorded in Feedback bundle with the preliminary confidence time series Conf preliminary (t) generated at the time in Embodiment four (step 407).

[0208] For example, the system finds that in the past 100 scenarios where the photovoltaic prediction deviation is >15%, although Conf preliminary (t) is predicted to be 0.9, the average value of the actual realization rate is only 0.75.

[0209] The system identifies this consistency residual. The calibration algorithm in Embodiment four (step 408) is executed to adjust the weights of (for example, the logistic regression model) so that the next time a photovoltaic prediction deviation >15% is encountered, the output Conf(t) is automatically reduced to close to 0.75, thereby achieving self-calibration of the model.

[0210] According to one aspect of the present application, the calibration of the reliability model is as follows: The system compares the actual job delay and pile position congestion statistics recorded in Feedback bundle with the marginal cost slope sj estimated at the time in Embodiment four (step 401).

[0211] For example, the system finds that during the 14:00-15:00 period, even if only a small amount of 100kW flexible (ΔP req ) is called, the actual pile position congestion events (from Metrics_report) far exceed expectations.

[0212] The system determines that the current cost slope s j for the 14:00-15:00 period is underestimated. The system automatically adjusts the s j for this period upwards. In the next round of optimization, the power side will find it more expensive to call flexibility during 14:00-15:00, thereby automatically avoiding this congestion period.

[0213] In summary, this embodiment ensures that the method of the invention is not only a static optimization tool, but also an intelligent system that can adapt to changes in port conditions, continuously learn from historical experience, and evolve the model (cost model, reliability model) itself by performing a complete closed loop of execution-verification-recharge-calibration.

[0214] Example Eight: A specific numerical simulation case This example provides a specific, reproducible numerical calculation case to exemplify the synergistic working process of the methods in Example One to Example Six.

[0215] Step 801: Scenario setting and baseline modeling Setting a rolling optimization time window T win = 15 minutes (i.e., 900 seconds), sampling period dt = 1 second, total N = 900 time points.

[0216] Assume that within this 15 minutes, the baseline power is superimposed by three parts: Steady base load: P const = 10 MW.

[0217] Low-frequency operation fluctuation (to be handled by energy storage): P low (t) = 3·sin(2π·t / 120) MW (period 120 seconds).

[0218] High-frequency noise (to be eliminated by the invention): P high (t) = 2·sin(2π·t / 20) MW (period 20 seconds).

[0219] Therefore, P port_base (t) = 10 + P low (t) + P high (t) (MW).

[0220] Constraint prototype ProtoConstraints: Assume that within this time window, M = 5 schedulable quay crane lifting operation events (i = 1 to 5) are identified, with their baseline start times t0 i respectively as [100, 110, 250, 260, 400] seconds. The power template p i (t) of each event is a rectangular wave with a peak of 800 kW and a duration of 15 seconds. The differentiable shifting window of each event is δ i ∈ [-20, +20] seconds.

[0221] Step 802: Spectrum shaping optimization Assume that the energy storage device (ESS_model) is designed to suppress fluctuations with a period greater than 60 seconds (f = 0.0167 Hz). Therefore, the invention sets the target cutoff frequency f c = 0.0167 Hz, aiming to eliminate the influence of P high (t) (period 20 seconds, f = 0.05 Hz).

[0222] Based on f c = 0.0167 Hz, design a discrete high-pass filter hHP and construct the convolution matrix H HP . Based on the event power template p i (t) and P port_base (t) construct the linearized model, finally get the quadratic form J HF (δ) = (Aδ - b) T (Aδ - b).

[0223] Assume only linear delay cost is considered, c_delay i = 10 (unit cost / sec). J ops (δ) = ∑ i 5 =1 10 · max(0, δ i ).

[0224] Solve: Set the trade-off coefficient α = 50. Solve the convex quadratic programming problem min J HF (δ) + 50 · J ops (δ), with constraints δ i ∈ [-20, +20].

[0225] The solver gets δ star = [-15, +10, -12, +15, -5] seconds. Generate P port_shaped (t). After spectral analysis, P port_base has a 2 MW peak at 0.05 Hz, while in P port_shaped , this peak is suppressed by 80%, down to 0.4 MW. The high-frequency energy J HF is significantly reduced. Meanwhile, R _max_series and DeltaP _max_series are generated.

[0226] Step 803: Three-signal packaging Observe the slice at t = 300 seconds (5th minute), based on the δ star scheme (for example, events 3, 4 still have room for fine-tuning), the system calculates that at t = 300 seconds, FlexEnvelope = {P_up _max : 400 kW, P _down_max : 250 kW, R_up _max : 100 kW / s}.

[0227] At t = 300 seconds, Forecast bundle shows that photovoltaic prediction uncertainty is high (factor 1) and the yard in-process prediction is 85% (factor 2). The fusion model (for example, logistic regression) outputs the preliminary confidence Conf preliminary= 0.85. At this time, the system retrieves historical recharge data (from step 805) and finds that the model predicts a historical actual redemption rate of only 0.80 at 0.85. The calibration algorithm kicks in and corrects the final output to Conf(300s) = 0.80.

[0228] At t = 300s, the up-regulation flexibility (δP > 0) is cost-mapped, including the following ways: Aggregation and estimation: estimate the underlying marginal slope s at breakpoints [0, 100, 250]kW j 0 = [0.2, 0.6] (Yuan / kW).

[0229] Discount: good spectral shaping effect (80%), η spec = 0.9. s j 1 = [0.18, 0.54].

[0230] Amplification: high uncertainty (from signal three), η risk = 1.15. s j 2 = [0.207, 0.621].

[0231] Harmonization: 0.621 > 0.207, satisfies convexity, no need to process.

[0232] Conservatization: apply Conf(300s) = 0.80 for conservatization. The slope seen by the power side (step 804) will be s j eff = [0.207 / 0.80, 0.621 / 0.80] = [0.259, 0.776] (Yuan / kW).

[0233] Step 804: power side solving Still at t = 300s. Conservatization discount: the power side receives Conf(300s) = 0.80. Adopt a nonlinear discount strategy β(Conf) = Conf 1.5 ≈ 0.716. Therefore, the conservatization flexibility envelope available to the power side is P up_eff (300s) = 0.716·400kW ≈ 286kW.

[0234] Joint optimization: at this time t = 300s is at the peak of electricity price, Tariff(300s) = 1.2 Yuan / kWh. The energy storage E ess is at a low position. System solving: min∑[1.2·P grid + φ eff + c deg · degradation + …].

[0235] The solver finds that the cost of electricity (1.2 yuan) is higher than the cost of the first flexible resource (0.259 yuan) but lower than the cost of the second flexible resource (0.776 yuan). Meanwhile, the cost of discharging the energy storage (including degradation) is equivalent to 0.9 yuan / kWh.

[0236] Therefore, the optimal decision is to prioritize the 100 kW of traffic flexibility (delta P req = 100 kW) and supplement the remaining insufficient part with the energy storage.

[0237] The feedback signal includes a shadow electricity price: because the system has called all cheap flexible resources and has used expensive energy storage, the dual solution (Lagrange multiplier) of the system power balance constraint is pushed high. The solver outputs P (300 s) = 1.5 yuan / kWh. This P = 1.5 is much higher than the electricity price of 1.2, which strongly indicates to the traffic side (next round) that at t = 300 seconds, the power system is extremely thirsty for power, and its internal value (1.5 yuan) is much higher than the market price (1.2 yuan).

[0238] Based on this high-price period, the solver generates Plateau request : in the next cycle [240 s, 360 s], the traffic side is requested to plateau the power to 9.5 MW.

[0239] Step 805: Perform feedback and recharge the energy storage according to the ESS schedule dispatch. The traffic side receives \{mu P , mu R , Plateau request} for the next round (t = 15 minutes) of optimization.

[0240] At t = 300 seconds, the system monitors that the requested delta P req = 100 kW, while the traffic side actually realizes delta P actual = 90 kW (due to a sudden event). The Feedback bundle is generated.

[0241] The actual power curve is stored in the historical database. This data point, with a prediction Conf = 0.80 and an actual realized delta P actual = 90 kW, is recharged to the reliability calibration submodule (step 408) of Embodiment Four to fine-tune the parameters of the Conf model in the future to make its predictions closer to reality.

[0242] In view of the problem that the service life of energy storage is sharply reduced due to passive response to high-frequency impact, the application no longer relies on energy storage to passively absorb all power fluctuations, but solves high-frequency impact at the source (i.e. traffic side) through a power spectrum shaping method. The method analyzes the physical limit of the energy storage system to determine a target cutoff frequency, and then actively minimizes high-frequency energy above the cutoff frequency by optimizing the micro-phase variable of the equipment such as the quay crane. This changes the role of the energy storage system from passive response in the full frequency range to only responsible for low-frequency active smoothing, thereby fundamentally alleviating the service life loss problem of energy storage.

[0243] In view of the problem that the flexible value of traffic cannot be quantified (black box), the application constructs a standardized three-signal packaging interface. The interface, through a piecewise convex cost function, first monetizes internal traffic side costs such as operation delay and yard congestion into marginal cost (yuan / kW) that can be understood by the power side; at the same time, it quantifies uncertainty such as weather and queuing into a redeemable confidence coefficient through a reliability coefficient. This interface completely opens the black box of the traffic system, allowing the power side to clearly understand the quantity, price and risk of flexibility.

[0244] In view of the problem of lack of cross-system collaborative optimization, the application, based on the above-mentioned quantification interface, constructs a joint optimization objective function. The function first includes the degradation cost of energy storage and the cost of traffic flexibility in the same optimization problem. The system can now make economic decisions: when the cost of moving the quay crane is lower than the cost of energy storage discharge (including degradation), traffic flexibility is preferred. And a rolling verification and data backfill mechanism uses actual execution data to continuously calibrate the cost and reliability models, realizing a sequential closed-loop optimization at the system level.

Claims

1. A sequential energy storage configuration optimization method for port area distribution networks based on transportation-electricity integration, characterized in that, include: Acquire multi-source data from the transportation and power systems, perform multi-source data aggregation and baseline modeling, and obtain port area baseline data; Based on the port area baseline data, the traffic-side operation phases are slightly staggered to perform power spectrum shaping, resulting in traffic-side shaped data. Based on the traffic-side sculpted data and port area baseline data, the traffic-side executable flexibility, costs, and risks are encapsulated into a standardized three-signal interface. Based on a standardized three-signal interface, a pre-configured energy storage device model, and a set of grid constraints, a closed-loop solution is performed on the power side to obtain the power side feedback signal and the power side execution scheduling.

2. The method according to claim 1, characterized in that, A slight peak shift is applied to the traffic-side operation phases to perform power spectrum shaping, resulting in traffic-side shaped data, including: Based on the prototype of adjustable equipment constraints contained in the port area baseline data, micro-phase variables and feasible domain constraints for traffic-side operations are established. Analyze the physical limits of the energy storage device model to determine the target cutoff frequency corresponding to the power fluctuations that the energy storage can withstand; A spectral shaping optimization objective is constructed to minimize the high-frequency energy above the target cutoff frequency in the total power sequence of the port area. The micro-phase variables are solved to obtain the phase micro-shifting scheme. Based on the phase shift scheme and port area baseline data, traffic-side shaped data is generated, including the shaped port area power sequence, the adjustable slope upper limit sequence, and the adjustable power boundary sequence.

3. The method according to claim 2, characterized in that, Construct spectral shaping optimization objectives, including: Based on the equipment adjustable constraint prototype and rolling prediction set in the port area baseline data, the operation cost and risk function of traffic-side operations are evaluated. By integrating high-frequency energy with operational costs and risk functions, a combined optimization objective is formed; The cost and risk function should include at least: cost of operation delay, risk of work-in-process spillover, and penalty for pile congestion.

4. The method according to claim 2 or 3, characterized in that, Minimize the high-frequency energy above the target cutoff frequency in the total power sequence of the port area, including: A linearized power model of the total power of the port area is constructed based on the event power template in the micro-phase variables and the port area baseline data. Based on the target cutoff frequency, design a discrete high-pass filter and construct its corresponding convolution matrix; By applying the convolution matrix to the linearized power model, the high-frequency energy is expressed as a time-domain quadratic form with respect to the micro-phase variable.

5. The method according to any one of claims 2 to 4, characterized in that, Solving for the infinitesimal phase variables yields a phase shift scheme, including: Construct a feasible region projection operator based on feasible region constraints; Solve the relaxation problem of the spectral shaping optimization objective in the continuous domain, which integrates the time-domain quadratic form with the operation cost and risk function, to obtain a continuous solution; The feasible region projection operator is applied to constrain and repair the continuous solution, and then discretization is performed to generate a phase shift scheme.

6. The method according to claim 1, characterized in that, Standardized three-signal interfaces include: A flexible envelope generated from traffic-side shaped data is used to quantify the limits of upward and downward power adjustment, the limits of upward and downward slope, and the minimum sustainable duration on the traffic side. Based on the port area baseline data and the operation cost log and shaping quality indicators derived from the power spectrum shaping process, the mapped piecewise convex cost function is used to quantify the marginal cost of calling flexible resources. And based on the uncertainty predictions and historical fulfillment records in the Hong Kong area baseline data, the estimated reliability coefficient sequence is used to quantify the fulfillment confidence of flexible commitments.

7. The method according to claim 6, characterized in that, The piecewise convex cost function of the mapping includes: Based on the prediction set in the operation cost log and the port area baseline data, the traffic-side operation costs are aggregated and the marginal cost slope is estimated. Apply spectral shaping revenue reduction to the marginal cost slope using shaping quality indicators; Based on the uncertainty assessment in the port area baseline data, a risk amplification is applied to the reduced slope; Perform convexity homology processing on the slope and integrate to obtain the basic piecewise convex cost; The basic piecewise convex cost is conservatively processed by applying a reliability coefficient sequence to generate a piecewise convex cost function.

8. The method according to claim 6 or 7, characterized in that, The estimated reliability coefficient sequence includes: Uncertainty factors were extracted from the port area baseline data. These uncertainties included at least the following: weather and photovoltaic forecast deviations, uncertainties in pile site occupancy and queuing, and fluctuations in work arrival and completion times. By incorporating uncertainty factors, a preliminary confidence time series is generated; Based on historical fulfillment records in the port area baseline data, the initial confidence level time series is calibrated to minimize the consistency residual between the initial confidence level and the historical fulfillment probability, thus obtaining the reliability coefficient sequence.

9. The method according to claim 1 or 6, characterized in that, Perform closed-loop solution on the power side, including: Analyze the standardized three-signal interface to obtain the flexible envelope and reliability coefficient sequence; By applying the reliability coefficient sequence, the up and down power limits and the up and down slope limits defined in the flexible envelope are conservatively reduced to generate a conservative flexible envelope. Conservative flexible envelope is used as the constraint set for solving closed-loop problems on the power side.

10. The method according to claim 1, 6, 7 or 9, characterized in that, Perform closed-loop solution on the power side, including: Construct a joint objective function that aims to minimize the total cost over the rolling time domain; The total cost aggregates at least the following: the grid purchase cost derived from the grid constraint set; Piecewise convex cost function; energy storage degradation cost derived from energy storage device model; And grid-side ramp rate penalty derived from grid constraint set.

11. The method according to claim 1, 9, or 10, characterized in that, The process of performing a closed-loop solution on the power side to obtain the power side feedback signal includes: After solving the joint objective function, the power shadow price is extracted from the dual solution of the system power balance constraint; Extract the slope shadow price from the dual solution of the grid-side ramp rate constraint; Based on the optimal scheduling trajectory obtained from the solution, and combined with the power shadow price and the slope shadow price, a platform-based power request is generated; The power-side feedback signal includes the power shadow price, the slope shadow price, and the platformized power request.

12. A sequential energy storage configuration optimization system for port area distribution networks based on transportation-electricity integration, characterized in that, include: processor; And a memory storing a computer program that, when executed by the processor, implements the method as described in any one of claims 1 to 11.

Images