Port Green Energy Virtual Energy Storage Aggregation and Control Method

Abstract

This invention relates to the field of power grid technology and discloses a method for virtual energy storage aggregation and control of green electricity in ports. The method includes: acquiring aggregation scenario and resource data; generating a power hard constraint mapping and a sub-stake execution reference by calculating multi-band energy indicators and identifying system power transfer characteristics; generating a local fast control configuration set by solving the virtual impedance feasible region and combining it with the power hard constraint amortization compensation budget; fusing the sub-stake execution reference (slow channel) generated based on the power hard constraint with the fast channel execution trajectory obtained after local online execution compensation, and performing priority arbitration under the constraints of the power hard constraint to output the final execution power trajectory. This invention constructs a two-layer control architecture of slow channel planning + fast channel compensation, and through the core link of power hard constraint mapping, ensures the coordinated operation of fast and slow channels and the safety and stability of the entire system, providing the power grid with fast and reliable dynamic support capabilities.

Description

Technical Field

[0001] This invention relates to the field of power grid technology, specifically to a method for virtual energy storage aggregation and control of green electricity in ports. Background Technology

[0002] With increasing global emphasis on environmental protection and carbon neutrality, the green and intelligent transformation of ports, as key hubs of international trade, has become an inevitable trend in the industry. The full electrification of port equipment, such as rubber-tired gantry cranes, rail-mounted gantry cranes, automated guided vehicles (AGVs), and container trucks, is a core measure for achieving zero-carbon ports. This transformation, while significantly reducing port carbon emissions and pollutants, has also spurred the development of a massive number of distributed mobile energy storage units with charging and discharging capabilities. If these dispersed charging loads, which exhibit a degree of temporal and spatial randomness, can be effectively aggregated and coordinated to form a large-scale port green energy virtual energy storage system (VESS), it can not only achieve optimized internal energy management but also provide high-value ancillary services to the power system as a whole, such as frequency regulation, reserve capacity, and demand-side response. This has extremely important research significance and engineering application value for enhancing the stability of regional power grids, promoting the local consumption of renewable energy sources such as wind and solar power, and opening up new value-added business models for port operators.

[0003] Currently, research on the aggregation of charging facilities for grid regulation has made some progress. Existing technical solutions typically employ centralized or decentralized control architectures. In centralized control schemes, an upper-level aggregation and control center calculates the power reference curve for each charging terminal over a future period based on dispatch instructions or internal optimization objectives issued by the grid, and directly distributes this information to each terminal for execution via a communication network. To achieve basic grid support, some solutions introduce droop control strategies, allowing each charging pile to autonomously adjust its active or reactive power output according to a fixed droop coefficient based on locally measured grid frequency or voltage deviations, thereby achieving primary frequency regulation or voltage support functions. Simultaneously, to ensure basic charging needs for vehicles, these solutions typically integrate battery state of charge (SOC) balancing management strategies to prevent some vehicles from affecting normal port operations due to deep discharge. Furthermore, to suppress the impact on the grid caused by the simultaneous start-up, shutdown, or power adjustment of a large number of devices, existing technologies typically set a fixed, conservative limit on the power change rate of charging piles, i.e., using a linear ramp method to control power increases and decreases.

[0004] However, existing technologies still face profound technical challenges in dealing with the complex scenarios of large-scale, high-density power electronic converter (charging pile) aggregation and control. These challenges mainly stem from the inherent contradiction between the speed of system dynamic response and the safety of operational boundaries, as well as the inherent stability risks of interaction between a large number of converters and the power grid. First, existing control schemes generally suffer from the problem of decoupling control objectives from safety boundaries. Second, to achieve high-value grid services, the aggregation system needs to have high-bandwidth response capabilities, but this can induce instability risks in high-bandwidth control and grid interaction. Summary of the Invention

[0005] To address the technical problems of the contradiction between rapid power response and internal electrical safety in the power grid, and the risk of instability in the interaction between a large number of converters and an unknown power grid, this invention proposes a virtual energy storage aggregation and control method for green electricity in ports.

[0006] The technical solution adopted in this invention is as follows:

[0007] This invention proposes a method for the aggregation and control of virtual energy storage in ports, comprising the following steps: acquiring aggregation scenario and resource data, including: a synchronization dataset of virtual energy storage, a baseline power trajectory, a set of equipment constraints, and a set of available capabilities for vehicles and charging facilities; based on the aggregation scenario and resource data, generating a power hard constraint mapping and a branch execution reference by calculating multi-band energy indicators and identifying system power transfer characteristics, wherein the energy indicators include: power quality indicators; based on the aggregation scenario and resource data and the power hard constraint mapping, generating a local fast control configuration set by solving the feasible domain of virtual impedance and allocating the compensation budget, and performing online compensation to obtain the aligned fast channel execution trajectory; fusing the branch execution reference, the local fast control configuration set, and the aligned fast channel execution trajectory, and performing priority arbitration under the constraints of the power hard constraint mapping to output the final execution power trajectory of the virtual energy storage.

[0008] Optionally, generating a power hard constraint mapping and stub execution reference includes: deriving the power peak value and rate of change upper limit of the power domain based on the device DC ripple limit contained in the device constraint set of the aggregated scenario and resource data, thus forming a power hard constraint mapping; and using the power hard constraint mapping to constrain and allocate the baseline power trajectory in the aggregated scenario and resource data to generate a stub execution reference.

[0009] Optionally, the upper limits of the peak power and rate of change in the power domain are derived in reverse, including: calculating the power change energy in three windows—low frequency band, twice the grid frequency band, and switch sideband—based on aggregated scenario and resource data, to obtain the three-frequency band energy index; identifying the amplitude-frequency transfer characteristics of power change to DC voltage ripple based on aggregated scenario and resource data, to obtain the power-to-DC ripple transfer characteristics; and combining the three-frequency band energy index and the power-to-DC ripple transfer characteristics, and solving with the DC ripple limit as the boundary condition to determine the upper limits of the peak power and rate of change.

[0010] Optionally, generating the power hard constraint mapping and sub-staking execution reference also includes: monitoring the number of zero-point crossings of aggregated power to obtain a zero-point crossing count; aggregating low-frequency energy and insulation monitoring alarms reflected by aggregated scenario and resource data to generate a zero-crossing risk index; and dynamically adjusting the polarity holding window duration and buffer bandwidth used to constrain the sub-staking execution reference based on the zero-point crossing count and the zero-crossing risk index.

[0011] Optionally, generating the power hard constraint mapping and stub execution reference also includes: performing a two-stage smooth switching at the boundary of the polarity holding window, wherein the power change rate is slowed down to near zero by using trend feedforward; enabling second-order ramp control to complete the zero-crossing switching of power polarity, and extending the execution duration of the second-order ramp if an increase in the energy index of either frequency band is detected during the switching.

[0012] Optionally, a local fast control configuration set is generated, including: identifying the set of equivalent impedance frequency points of the power grid by injecting power disturbance sequences and collecting responses based on aggregated scenario and resource data; solving the feasible domain of virtual impedance parameters that satisfy the preset frequency domain passivity conditions and the set of equivalent impedance frequency points of the power grid; selecting a set of parameters from the feasible domain of virtual impedance parameters to form a virtual impedance configuration, and using it as a component of the local fast control configuration set.

[0013] Optionally, the online execution of compensation to obtain the aligned fast-path execution trajectory includes: superimposing the virtual impedance configuration onto the power compensation control loop in each local control cycle; evaluating the passivity margin of the current cycle online based on the equivalent impedance frequency point set of the power grid, and generating a passivity compliance flag; and performing safety arbitration on the power compensation result based on the passivity compliance flag to form the final compensation instruction.

[0014] Optionally, generating a local fast control configuration set also includes: determining the total compensation bandwidth budget for the entire station based on the power hard constraint mapping; calculating the compensation participation factor for each device based on the available capabilities of each device in the aggregated scenario and resource data; and allocating the total compensation bandwidth budget to each device according to the compensation participation factor to form a device-level gain upper limit budget, and incorporating it into the local fast control configuration set.

[0015] Optionally, the online execution compensation process for the aligned fast-channel execution trajectory also includes: collecting communication indicators such as communication latency, jitter, and packet loss rate, and combining them into a communication reliability score; determining the state machine state label based on the communication reliability score and a preset threshold, where the state machine state label includes any one of normal, out-of-step, conformal, and derating; and adaptively adjusting the power compensation gain based on the state machine state label before executing the final compensation instruction.

[0016] Optionally, the fusion of the stub execution reference, the local fast control configuration set, and the aligned fast channel execution trajectory, and the priority arbitration under the constraints of the power hard constraint mapping, includes: when fusion of the stub execution reference and the aligned fast channel execution trajectory, real-time detection of whether the synthesized power command exceeds the boundary of the power hard constraint mapping; if it exceeds the boundary, the stub execution reference is trimmed according to the priority rule of frequency support first, voltage support second, and trajectory tracking third, to ensure that the fusion result conforms to the power hard constraint mapping.

[0017] The beneficial effects of this invention are:

[0018] This invention solves the technical problems of the contradiction between fast power response and internal electrical safety in the prior art, as well as the risk of instability when a large number of converters interact with an unknown power grid. It constructs a two-layer control architecture of slow channel planning + fast channel compensation, and through the core link of power hard constraint mapping, it realizes the ability to provide fast and reliable dynamic support to the power grid while actively ensuring the stability of the entire system. Attached Figure Description

[0019] Figure 1 This is a flowchart of a port green electricity virtual energy storage aggregation and control method according to an embodiment of the present invention.

[0020] Figure 2 This is a flowchart of generating power hard constraint mapping and stub execution reference according to an embodiment of the present invention.

[0021] Figure 3 This is a flowchart of generating power hard constraint mapping and stub execution reference according to another embodiment of the present invention.

[0022] Figure 4 This is a flowchart of generating a local fast control configuration set according to an embodiment of the present invention.

[0023] Figure 5 This is a flowchart of the fast-track execution trajectory obtained by online compensation according to an embodiment of the present invention. Detailed Implementation

[0024] In order to solve the problems existing in the current technology, the applicant conducted in-depth research and found that:

[0025] The decoupling problem between control objectives and safety boundaries manifests in the fact that existing power command generation primarily aims to track upper-level dispatch signals or meet droop response requirements, while internal electrical safety constraints of the system, such as voltage ripple on critical DC buses, are not directly and dynamically mapped to the power control domain. Controllers typically rely on a static, conservative power change rate limit set for worst-case conditions. This one-size-fits-all static constraint severely suppresses the rapid response potential of virtual energy storage when the system is in good condition; however, when the system is vulnerable, even if the static limit is followed, rapid power commands issued by the controller may still trigger resonance within the port area microgrid, leading to DC ripple exceeding limits, triggering protection, and causing unexpected interruptions to the grid support services that should be provided. The root cause of this problem lies in the lack of a mechanism that can convert internal, frequency-domain-dependent power quality constraints (such as ripple) into time-varying, directly executable power boundaries (i.e., peak value and rate of change limits) in the power domain in real time and quantitatively. The instability problem arising from high-bandwidth control interaction with the power grid is even more challenging. When hundreds of charging pile control loops (with bandwidths reaching several hertz or even higher to provide high-quality frequency support) interact with an unknown, time-varying equivalent impedance of the power grid, negative damping can easily form in a certain frequency band (usually the mid-to-low frequency band), leading to subsynchronous or supersynchronous oscillations. Existing technologies either cannot effectively address this problem due to insufficient control bandwidth, or, although employing advanced control strategies, typically assume that the power grid impedance is known or in an ideal state, lacking a mechanism to proactively identify power grid characteristics and pre-configure controller parameters accordingly to actively establish system stability margins. This results in the system constantly operating under potential instability risks. These two problems are interconnected; the former limits system performance, while the latter threatens the very survival of the system.

[0026] Example 1 describes the aggregation and control architecture and data foundation of a port green electricity virtual energy storage system, providing an exemplary system environment and data foundation, and elaborates on the preparation stage of data acquisition and preprocessing.

[0027] In this embodiment, a port green electricity virtual energy storage aggregation and control system is deployed in an automated container terminal. The system physically includes a central aggregation and control station, which communicates with multiple distributed energy units within the terminal via industrial Ethernet or a 5G private network. The distributed energy units are primarily clusters of charging piles connected to operating equipment such as container trucks (AGVs), tire-mounted gantry cranes (RTGs), and rail-mounted gantry cranes (RMGs). These charging piles have bidirectional charging and discharging capabilities and can be aggregated into virtual energy storage resources. The central aggregation and control station is used to execute the control method of this invention.

[0028] Specifically, the first stage involved in this embodiment is data aggregation and basic alignment, the process of which includes:

[0029] Step S11, Data Acquisition and Cleaning. The aggregation and control main station reads and aggregates data from various locations at the terminal in real time via the communication network. For example, the data channels collected include at least: the operating power of each charging pile, the active power of the port area at the common access point (PCC) of the power grid, the effective voltage value of the key bus, the bus frequency, the battery status of each vehicle (such as state of charge, temperature), the operating status of the charging facilities (such as standby, charging, discharging), and standard timing pulse signals from the GPS or Beidou system.

[0030] To ensure data synchronization and accuracy, the acquired raw data undergoes a series of preprocessing steps. Time calibration is a crucial step; the master station uses the received timing pulse as a reference to append a uniform timestamp to all uploaded data and resamples it uniformly with a 5ms sampling step size to obtain aligned raw measurements. This aims to eliminate time deviations introduced by differences in sampling clocks between measurement channels, forming the basis for subsequent precise frequency domain analysis and coordinated control. Based on this, missing measurement repair and outlier suppression are performed on the aligned raw measurements. For example, missing measurement repair can be performed using linear interpolation based on neighboring data points or the preserve method, and outliers in the data sequence can be identified and removed using the 3σ criterion (three sigma criterion) or the Local Outlier Factor (LOF) algorithm. After these processes, a high-quality synchronized dataset is obtained for subsequent steps.

[0031] Step S12, Constraint Loading. The aggregation control master station loads and parses preset operational constraints or those received from the power grid dispatching agency. These constraints constitute the safety boundary for control decisions. Specifically, constraints may include: the grid-connected power limit agreed upon between the port area and the main grid (e.g., the maximum grid-connected power shall not exceed 10MW), the insulation monitoring threshold of the charging facility itself (used to determine whether there is a risk of leakage), the trigger voltage condition for the battery pre-charging process, the limit of DC bus ripple voltage (e.g., the peak-to-peak value shall not exceed 5% of the rated voltage), the maximum allowable power change rate of a single charging and discharging device (e.g., not greater than 500kW / s), and the allowable compensation gain range when performing fast frequency or voltage support. These constraints are merged to form a set of equipment constraints.

[0032] Step S14, Resource Capacity Aggregation. The main station periodically (e.g., every minute) collects real-time data on the status of all vehicles and charging facilities in the synchronized dataset to assess the availability of the entire virtual energy storage system. Specifically, this process filters out terminals that are faulty, under maintenance, or fully charged and not allowed to discharge. Then, it accumulates the available charging and discharging power and remaining available energy (in kWh) of all available terminals and records the status of their corresponding charging gun connectors. Finally, an aggregated set of available vehicle and charging facility capabilities is formed. This capability set is dynamically updated, reflecting the real-time energy-power reserve status of the virtual energy storage pool.

[0033] Example 2: This example describes the overall flow of the aggregation control method defined therein. The method flow of this example is based on the system architecture and data foundation constructed in Example 1.

[0034] Acquire the synchronization dataset, device constraint set, vehicle and charging facility available capacity set, and baseline power trajectory; based on the synchronization dataset and device constraint set, generate a power hard constraint mapping and a branch-pile execution reference by calculating multi-band energy indicators and identifying the power-to-DC ripple transfer characteristics; based on the synchronization dataset, device constraint set, vehicle and charging facility available capacity set, and power hard constraint mapping, generate a local fast control configuration set by solving the virtual impedance feasible region and allocating the compensation budget, and perform online compensation according to the local fast control configuration set to obtain the aligned fast channel execution trajectory; fuse the branch-pile execution reference, the local fast control configuration set, and the aligned fast channel execution trajectory, perform priority arbitration under the constraints of the power hard constraint mapping, and output the final execution power trajectory.

[0035] The process of generating a power hard constraint mapping and a charging station execution reference based on aggregated scenario and resource data includes: parsing the synchronization dataset, equipment constraint set, and available vehicle and charging facility capability set from the aggregated scenario and resource data; calculating the power change energy of the low-frequency band, twice the grid frequency band, and the switching sideband based on the synchronization dataset to obtain the three-band energy index; injecting a short-time power sequence into the synchronization dataset to identify the amplitude-frequency characteristics of power to DC voltage ripple and obtain the power to DC ripple transfer characteristics; based on the DC ripple limit in the equipment constraint set, combined with the power to DC ripple transfer characteristics and the three-band energy index, deriving the allowable peak power and rate of change upper limit for each frequency band to form a power hard constraint mapping; and fusing the available vehicle and charging facility capability sets and using the power hard constraint mapping for pruning to allocate targets for charging facilities and generate a charging station execution reference.

[0036] In other words, such as Figure 1 As shown, a method for virtual energy storage aggregation and control of green electricity in ports includes:

[0037] Step one: Acquire aggregated scenario and resource data. This step can be specifically referred to in Implementation Example 1 regarding the data acquisition, cleaning, constraint loading, and resource aggregation process. Its purpose is to provide comprehensive, accurate, and synchronized basic data input for subsequent regulatory decisions, specifically including the synchronized dataset, device constraint set, and available vehicle and charging facility capability set generated in Implementation Example 1.

[0038] Step two involves generating a power hard constraint mapping and a branch-specific execution reference based on aggregated scenario and resource data, by calculating multi-band energy indicators and identifying system transmission characteristics. This step is a crucial mapping link from power quality to power control in this invention. In this embodiment, to avoid the impact of severe power fluctuations on the DC bus and the triggering of overvoltage or overcurrent protection, this step innovatively maps power quality indicators that are not easily controlled directly (such as DC ripple) into time-varying power domain constraints that the controller can directly execute (i.e., power hard constraint mapping), and plans a safe and smooth power command (i.e., branch-specific execution reference) for each charging pile based on this constraint. The detailed implementation of this step will be further described in subsequent embodiments (such as embodiments three and four).

[0039] Step 3: Based on the aggregated scenario and resource data and the power hard constraint mapping, a local fast control configuration set is generated by solving the feasible region of virtual impedance and allocating the compensation budget. The aligned fast channel execution trajectory is then obtained through online compensation. This step aims to establish a local compensation channel, i.e., a fast channel, capable of rapidly responding to grid disturbances (such as frequency and voltage fluctuations). To ensure the stability of a large number of charging piles connected to the grid, this step introduces virtual impedance design based on passive theory. The range of virtual impedance parameters that ensures stable operation is solved by actively identifying grid characteristics. Simultaneously, the total allowable compensation capacity of the entire station (limited by the power hard constraint mapping generated in Step 2) is rationally allocated to each charging pile. These parameters together constitute the local fast control configuration set, which is then distributed to each terminal for execution. The detailed implementation of this step will be further elaborated in subsequent embodiments (such as Embodiments 5 and 6).

[0040] Step four involves fusing the branch-pile execution reference, the local fast control configuration set, and the aligned fast-channel execution trajectory, and performing priority arbitration under the constraints of the power hard constraint mapping to output the final execution power trajectory. This step is the final instruction synthesis and execution stage. It can be seen that the branch-pile execution reference generated in step two belongs to the medium-to-low speed planned power trajectory (which can be called the slow channel), while the fast-channel execution trajectory generated in step three is a high-speed compensation power for disturbances. This step fuses the two and uses the power hard constraint mapping generated in step two as the highest priority constraint for verification. When the fused instruction exceeds the constraints, arbitration and pruning will be performed according to preset priority rules (e.g., frequency support for grid security takes precedence over planned tracking), outputting the final power instruction issued to the charging pile for execution. The detailed implementation of this step will be further described in subsequent embodiments (such as Embodiment 7).

[0041] In summary, the process in this embodiment constructs a two-layer control architecture of slow channel planning + fast channel compensation, and ensures the coordinated operation of fast and slow channels and the safety and stability of the entire system through the core link of power hard constraint mapping.

[0042] Example 3 describes the power hard constraint mapping generation process based on three-band energy, that is, the process of inversely mapping power quality constraints (DC ripple) to power domain hard constraints.

[0043] In this embodiment, the aggregation control method includes the following steps: based on the aggregation scenario and resource data, by calculating multi-band energy indicators and identifying system transmission characteristics, a power hard constraint mapping and a sub-charging pile execution reference are generated. In this embodiment, in order to fundamentally avoid the voltage surge on the DC bus of the port area microgrid caused by drastic power changes of a large number of distributed power sources (charging piles), which could trigger equipment protection actions or even shutdowns, this invention proposes a method to transform DC ripple, which is not easily observed and controlled directly, into upper and lower limits of power commands (i.e., power hard constraints) that the controller can directly understand and execute.

[0044] Specifically, such as Figure 2 As shown, this step, based on the device DC ripple limits contained in the aggregated scenario and resource data, reverse-engineers the peak power and rate of change upper limits of the power domain, forming a hard power constraint mapping. This reverse-engineering process does not employ a fixed margin or limit, but rather a dynamic, adaptive calculation method, such as... Figure 3As shown, it includes: calculating the power change energy in three windows—low frequency band, twice the grid frequency band, and switch sideband—based on aggregated scenario and resource data, to obtain the three-frequency band energy index; similarly based on aggregated scenario and resource data, identifying the amplitude-frequency transfer characteristics of power change to DC voltage ripple, to obtain the power to DC ripple transfer characteristics; combining the three-frequency band energy index and the power to DC ripple transfer characteristics, and using the DC ripple limit as the boundary condition, to determine the peak power and the upper limit of the rate of change.

[0045] To achieve the above process, the aggregation and control main station first performs the following preparatory work:

[0046] Online identification of power-to-DC ripple transfer characteristics: During gaps in normal port operations (e.g., brief intervals between operations of two quay cranes), the system sends a short-duration, low-amplitude power disturbance sequence to some charging piles via the aggregation control master station. Optionally, this sequence can be a multi-sine wave signal or a pseudo-random binary sequence (PRBS), whose frequency components cover the range of interest but deliberately avoid known sensitive frequency bands. For example, the amplitude of the disturbance sequence can be set to within 2% of the rated power of the charging pile, with a duration of approximately 1-3 seconds. During the disturbance injection, the master station synchronously acquires the total injected power disturbance signal sequence and the voltage ripple response signal sequence on the key DC bus at high frequency. Based on these two sets of input-output data, a frequency domain model is calculated using the cross-spectrum method or the frequency domain estimation method in the system identification toolbox. This model represents the power-to-DC ripple transfer characteristics, whose gain and phase at different frequency points quantitatively describe the magnitude of the same-frequency voltage ripple generated on the DC bus by the injected unit power disturbance under the current topology and load of the port system. For example, if the gain at 100Hz is identified as 0.05pu, it means that under the current operating conditions, injecting a 1kW amplitude 100Hz sinusoidal power fluctuation will generate a 1kW amplitude on the DC bus. 100Hz voltage ripple of 0.05 / P_base (where P_base is the power reference value).

[0047] Calculation of Three-Band Energy Indicators: The main station continuously analyzes the time-series data of the total active power of the PCC point in the port area from the synchronous dataset obtained in Example 1. By designing three sets of parallel digital bandpass filters, the power signal is decomposed into three key frequency bands. In this example, these three frequency bands are preferably set as follows:

[0048] Low frequency band: such as 1Hz to 20Hz. Power fluctuations in this range are usually related to slow dynamic processes such as the start-up and shutdown of large equipment and changes in mechanical load. Excessive energy may cause system oscillation.

[0049] Twice the grid frequency band: for example, 95Hz to 105Hz (for a 50Hz grid). Power fluctuations in this range are mainly generated by the rectifier / inverter stages and are one of the main sources of DC ripple.

[0050] Switching sideband: For example, 18kHz to 22kHz (for a 20kHz switching frequency). Power fluctuations in this range are directly related to the switching action of the power electronic converter itself. Excessive energy may lead to high-frequency noise and electromagnetic interference problems.

[0051] For the filtered power signals of each frequency band, the square integral is performed within a sliding time window (e.g., a 20-second window for the low-frequency band and a 500-ms window for the high-frequency band) to obtain the power change energy of each frequency band. These three energy values ​​together constitute the three-band energy index vector, which dynamically reflects the distribution of recent power fluctuation energy in different frequency domains.

[0052] Based on this, the main station executes the core joint solution process to determine the power hard constraints. Specifically, for any frequency band among the low-frequency band, twice the grid frequency band, and the switching sideband, the basic power margin is calculated based on the power-to-DC ripple transfer characteristics and DC ripple limits, and the energy suppression factor is calculated based on the energy indices of the three frequency bands. The energy suppression factor is inversely proportional to the square root of the energy index of the corresponding frequency band. The basic power margin and the energy suppression factor are combined to determine the upper limit of the power peak value and rate of change of the frequency band.

[0053] This joint process can be achieved through the following formula: P allow,b =α b Vlim(||G) P→V || b (1 / sqrt(E)) b +∈)); where: P allow,b The maximum peak power (in kW) allowed to contribute to frequency band b is a component of the upper bound of the power to be solved. b is the frequency band index, taking values ​​for the low-frequency band, twice the grid frequency band, and the switching sideband, respectively. α b The safety factor for frequency band b is a dimensionless parameter used to maintain a certain safety margin. Preferably, its value ranges from [0.6, 0.9], and it can be set differently according to the importance of different frequency bands. lim The DC bus ripple voltage limit (unit: V or pu) defined for the equipment constraint set is the boundary condition for this calculation. ||G P→V || b E represents the power-voltage coupling amplification factor of the system in that frequency band, which is the norm of the power-to-DC ripple transfer characteristic identified above (e.g., the maximum gain or L2 norm in that band). bThe calculated energy index for frequency band b is given by the three-band frequency band (unit: (kW)^2·s). ∈ is a small positive number, such as 10, used to prevent the denominator from being zero. -6 .

[0054] It can be seen that the calculation logic of this formula is divided into two parts: the first part V lim / ||G P→V || b The first part, 1 / sqrt(E_b+ε), can be considered the basic power margin. Based on the physical characteristics of the system, it directly transforms the voltage constraint into a power constraint. The second part, 1 / sqrt(E_b+ε), can be considered the energy suppression factor, which is dynamically adjusted according to the current operating state of the system. When the energy E_b of a certain frequency band is already high, this suppression factor will decrease, thereby significantly tightening the power margin of that frequency band and playing a role in actively suppressing potential instability.

[0055] Calculate P for each of the three frequency bands. allow,b Then, the minimum value among the three is taken, and after processing by a smoothing filter (such as a first-order low-pass filter), the final power peak hard constraint at the current moment can be obtained. The hard constraint of the power change rate can be derived in a similar way using the derivative or difference relationship of the transfer characteristic. This set of time-varying power peak and change rate curves together constitute the power hard constraint mapping.

[0056] like Figure 2 Finally, the power hard constraint mapping is used to constrain and allocate resources in the aggregated scenario and resource data, generating a sub-charging pile execution reference. This is an ideal, smooth total power curve predicted based on the port area's production plan. The main station compares this curve with the power hard constraint mapping, and performs peak shaving, valley filling, or slope limiting on the portions exceeding the constraints to obtain a corrected total power target. Then, based on the available capacity set of vehicles and charging facilities, the total power target is decomposed to each available charging pile in a proportional or optimized manner, forming the final sub-charging pile execution reference issued to each terminal.

[0057] As an optional implementation, the identification of transport characteristics does not need to be performed in real time. Instead, it can be performed offline in advance under several typical port operation scenarios (such as peak daytime operations and concentrated charging at night) and stored in the database. During runtime, the system automatically identifies the scenario based on the current time or operating conditions and calls the corresponding transport characteristic model, thereby reducing the complexity of online calculations.

[0058] Example 4 describes a two-stage smooth switching process for polarity-maintaining window adaptation and boundary, providing a refined improvement to the generation and execution process of the sub-stub execution reference. It is specifically designed to address the stability and safety issues when aggregated power crosses near zero (i.e., from charging state to discharging state, or vice versa).

[0059] In this embodiment, the method further includes: monitoring the number of zero-point crossings to obtain a zero-point crossing count; aggregating low-frequency energy and insulation monitoring alarms reflected by aggregated scene and resource data to generate a zero-crossing risk index; and dynamically adjusting the polarity holding window duration and buffer bandwidth used to constrain the sub-pile execution reference based on the zero-point crossing count and the zero-crossing risk index.

[0060] Specifically, when the power needs to switch from one polarity (e.g., charging, where the power is positive) to another polarity (e.g., discharging, where the power is negative), if the switching speed is too fast, it can easily cause system oscillations and DC voltage instability near the zero point. To address this, this embodiment introduces the concepts of a polarity maintenance window and a buffer band.

[0061] The polarity retention window defines a minimum time during which the polymerization power should maintain its polarity, avoiding frequent back-and-forth crossings around zero.

[0062] The buffer zone defines a small power range centered at zero power (e.g., ±5% of the rated power). When power enters this range, a special slow control strategy must be followed.

[0063] To enable these two parameters to adapt to changes in system state, this embodiment employs an adaptive adjustment mechanism. The master station continuously performs the following calculations:

[0064] Zero-point crossing count: Within a sliding time window (e.g., the past 5 minutes), count the number of times the zero point is crossed in Example 3.

[0065] Zero-Risk Index: This is a comprehensive index that aggregates multiple risk factors through a weighted sum. For example, Zero-Risk Index = w1 × (Low-Frequency Band Energy) + w2 × (Number of Insulation Monitoring Alarms). Here, w1 and w2 are weighting coefficients. High low-frequency band energy indicates a risk of low-frequency oscillation in the system; insulation monitoring alarms may be related to DC system grounding, with an even higher risk during power reversal.

[0066] Dynamic Adjustment: Based on the above indicators, the window duration and buffer bandwidth are adjusted according to preset rules. For example: Window duration = T_base + k1 × (zero-point crossing count) + k2 × (zero-crossing risk index); Buffer bandwidth = B_base + h1 × (zero-crossing risk index). Where T_base and B_base are baseline values, and k1, k2, and h1 are positive coefficients. This means that when the system detects frequent crossings or a high zero-crossing risk, it will automatically extend the polarity maintenance window and widen the buffer band, thereby forcing the system to perform power reversal in a more conservative and slower manner.

[0067] Furthermore, in order to ensure an extremely smooth process when crossing the buffer zone, the method of this embodiment also includes: performing a two-stage smooth switching at the boundary of the polarity maintenance window, wherein, firstly, the power change rate is reduced to near zero by using trend feedforward; then, second-order ramp control is enabled to complete the zero-crossing switching of power polarity, and during the switching, if an increase in the energy index of either frequency band is detected, the execution duration of the second-order ramp is extended.

[0068] The two-phase switching process is implemented as follows:

[0069] Phase 1: Trend-Fedforward Deceleration. When the aggregated power trajectory is about to enter the buffer zone, the controller does not simply decelerate linearly. Instead, it activates a trend-feedforward module, which predicts the inertia of the power trajectory as it enters the buffer zone based on the short-term rate of change of power before reaching the boundary. Based on this prediction, the controller applies a reverse, smooth control force, actively reducing the rate of change of power to a very small value (near zero), allowing it to gently reach the boundary of the buffer zone.

[0070] Phase Two: Second-Order Slope Zero Crossing. Once power enters the buffer zone, the control strategy switches to a second-order slope. This means that not only is the rate of change of power limited, but the rate of change of the rate of change itself (i.e., acceleration) is also limited, thus achieving the smoothest zero crossing. During this slope execution, the master station continuously monitors the energy indicators of the three frequency bands in Example 3 at high frequency. If an abnormal rise in energy is detected in any frequency band (especially the low frequency band and twice the grid frequency band) and exceeds the warning threshold, it indicates that this zero crossing may be triggering an unstable mode in the system. At this time, the controller will immediately extend the execution time of the second-order slope in Phase Two online, for example, extending the originally planned 200ms zero crossing time to 350ms. By further slowing down the zero crossing speed, the development of oscillations is actively suppressed, ensuring the absolute safety of the switching process.

[0071] As an optional implementation, the transition points and parameters of the extended logic for the two-stage switching can be optimized online by an expert system or reinforcement learning agent to adapt to more complex operating conditions. Furthermore, before entering the first stage, the master station can send an early warning message about the impending power reversal to the local controller, causing it to reduce the compensation gain in advance, thus forming closer coordination.

[0072] Example 5 describes the configuration of a local fast control channel based on passive constraints, and explains how to configure the parameters of the local fast control channel for each distributed power source (charging pile) to achieve fast and stable grid support.

[0073] In large port areas, numerous charging piles operate as power electronic converters (Inverters) connected to the grid. Their high-frequency switching characteristics and control loop behavior may interact with the equivalent impedance of the power grid, inducing resonance or instability in specific frequency bands. To fundamentally address this issue, this invention proposes a proactive method to ensure system stability. Before issuing specific compensation commands, a set of configuration parameters is generated for each local controller to ensure its passivity. This process involves: based on the aggregation of scenario and resource data and the mapping of power hard constraints, solving the feasible region of virtual impedance and allocating the compensation budget to generate a local fast control configuration set.

[0074] Specifically, such as Figure 4 As shown, this step includes the following core components:

[0075] Online identification of the equivalent impedance of the power grid: By injecting a power disturbance sequence and acquiring the response, the set of frequency points of the equivalent impedance of the power grid is identified. The purpose of this process is to obtain a dynamic characteristic profile of the local power grid at different frequencies.

[0076] The equivalent impedance frequency set of a power grid refers to a dataset that contains the complex impedance values ​​(including amplitude and phase angle) presented on the grid side as seen from the common grid connection point (PCC) of the charging pile cluster at a series of discrete frequency points.

[0077] The main control station plans and executes a low-amplitude, short-window, multi-frequency separated disturbance injection. For example, a 3-second time window with relatively stable operating load is selected, and a small, additional power disturbance reference is synchronously sent to all or some charging piles in the system. This disturbance signal is preferably a superimposed multi-sine wave signal with its frequency points distributed within the target frequency band of 5Hz to 2kHz, but avoiding known sensitive areas such as power frequency (50Hz) and second harmonic (100Hz); its total amplitude is strictly limited to within 2% of the rated power of the entire station to ensure no disturbance to normal production. During the injection, high-speed monitoring units (such as PMUs or high-frequency waveform recorders) deployed at the PCC point synchronously record the instantaneous values ​​of voltage and current at the injection point at a sampling rate of not less than 10kHz. After data acquisition, a Fourier Transform (FFT) is performed on the voltage and current response sequences. At each injected frequency point ω, the complex estimate of the grid admittance, Y_grid(jω) = ΔI(jω) / ΔV(jω), is calculated; where ΔI(jω) is the phasor of the current response and ΔV(jω) is the phasor of the voltage response. Finally, by taking the reciprocal Z_grid(jω) = 1 / Y_grid(jω), the equivalent grid impedance at that frequency point is obtained. This process is repeated for all frequency points to form the required impedance frequency set.

[0078] By actively and safely identifying impedance characteristics that reflect the actual operating conditions of the current power grid, the traditional control design approach that relies on inaccurate system models or typical parameters can be obtained. This provides an accurate data foundation for subsequent stability design and is a prerequisite for achieving highly reliable adaptive control.

[0079] Solving the feasible region of virtual impedance: Based on the preset frequency domain passivity condition and combined with the equivalent impedance frequency point set of the power grid, the feasible region of virtual impedance parameters that satisfy the condition is solved.

[0080] Virtual impedance is a control algorithm that improves the dynamic performance of a system by modifying the voltage or current control loop of the converter to exhibit additional, programmable resistance, inductance, or capacitance characteristics. The feasible region of virtual impedance parameters is the set of all parameter points in a multidimensional parameter space (such as virtual resistance R_v, virtual inductance L_v) that allow the system to satisfy passive constraints.

[0081] In this embodiment, the preset frequency domain passivity condition is: at any frequency point within the target frequency band, the real part of the series impedance formed by the set of frequency points of the equivalent impedance of the equipment and the equivalent impedance of the power grid is not less than the preset passivity margin δ. This condition can be expressed by the formula: Re{Z device (jω)+Z grid (jω)}≥δ;

[0082] Where: Re{...} represents the operation of taking the real part of a complex number; Z_grid(jω) is the grid impedance identified in the previous step; Z_device(jω) is the equivalent output impedance of the charging pile converter, which can be expressed as Z_device(jω)=Z_phy(jω)+Z_v(jω), where Z_phy(jω) is the inherent physical impedance of the converter, and Z_v(jω) is the virtual impedance to be designed, whose expression is related to parameters such as the virtual resistance R_v and virtual inductance L_v to be solved; δ is a preset passivity margin, a small positive real number (e.g., 0.02pu), used to ensure that the system remains stable under parameter perturbations or model uncertainties. Expanding the above inequalities on a series of discrete frequency points within the target frequency band (e.g., 5Hz-300Hz) forms a system of linear inequalities about R_v and L_v. Solving this system of inequalities using a grid search method or a convex optimization toolkit yields a two-dimensional or multi-dimensional feasible region.

[0083] Based on rigorous passivity theory, the system stability problem is transformed into a mathematical constraint that can be explicitly solved. By solving the feasible region, the fuzzy objective of ensuring stability is explicitly expressed as the specific range of values ​​for the controller parameters (virtual impedance), transforming stability design from post-event verification to pre-event constraints, greatly improving the design efficiency and reliability of the control system.

[0084] Allocation of compensation budget: After determining the stability domain, it is also necessary to determine the upper limit of output of each device when performing rapid compensation.

[0085] Determine the total compensation bandwidth budget: The total compensation bandwidth budget for the entire station is determined based on the power hard constraint mapping. Specifically, the aggregation control master station obtains the time-varying power change rate hard constraint curve generated in Example 3, and takes its minimum value over a future control cycle as Rate_limit_hard. Simultaneously, the sum of the physical maximum power change rates of all available charging piles is aggregated as Rate_limit_phy. The total compensation bandwidth budget Budget_total for the entire station is prudently set to min(Rate_limit_hard, Rate_limit_phy) × k_safety, where k_safety is a safety factor (e.g., 0.8). This tightly couples the dynamic capabilities of the fast channel with the global safety constraints of the slow channel.

[0086] Calculate the compensation participation factor: Based on the available capabilities of each device in the aggregated scenario and resource data, calculate the compensation participation factor for each device. For the i-th device, its participation factor Factor_i can be weighted and synthesized from its available power P_avail_i, available energy E_avail_i, etc., for example, Factor_i=w_p×(P_avail_i / ΣP_avail)+w_e×(E_avail_i / ΣE_avail).

[0087] Allocating device-level budgets: Following the compensation participation factor, the total compensation bandwidth budget is allocated to each device. The initial budget for the i-th device is Budget_i = Budget_total × Factor_i. Optionally, to address the situation where the budget allocated to some devices exceeds their own physical limits, an iterative algorithm of peak shaving and backfeedback can be used: if the Budget_i of a device i exceeds its Rate_limit_phy_i, the excess portion ΔB = Budget_i - Rate_limit_phy_i is proportionally backfeeded to other unsaturated devices until the budgets of all devices are within their physical constraints.

[0088] Finally, a set of parameters is selected from the feasible domain of virtual impedance parameters (for example, under the premise of satisfying constraints, R_v with the minimum additional loss or L_v with the best damping is selected), together with the above-assigned device-level gain upper limit budget, as well as the preset second-order ramp parameters and priority arbitration table, and encapsulated into a data packet, namely the local fast control configuration set, and sent to each local controller.

[0089] Example 6: Describe the online execution and status adaptation of the local fast control channel, and elaborate on the online application and dynamic execution process of the local fast control configuration set. It describes how the local controller of the charging pile autonomously and intelligently performs fast compensation in each local control cycle (e.g., every 10 ms).

[0090] The process is as follows: The compensated fast channel execution trajectory is obtained through online execution compensation, as specifically shown in Figure 5 the figure.

[0091] Communication health assessment and state machine decision: At the beginning of each control cycle, the local controller first evaluates the communication quality with the master station. It collects communication metrics such as communication delay, jitter, and packet loss rate, and synthesizes them into a communication credibility score; based on the communication credibility score and a preset threshold, it determines the state machine status label, and the state machine status label includes any one of normal, out-of-step, conforming, and derating.

[0092] The controller maintains a communication quality monitor, calculates the delay mean and jitter by analyzing the timestamps of received messages, and detects the packet loss rate through message sequence numbers. An exemplary scoring function is: Score = w_d × Delay_norm + w_j × Jitter_norm + w_l × Loss_rate. According to the value of Score, the controller switches between the four states:

[0093] Normal: Score ≤ θ_normal, good communication, all functions enabled.

[0094] Out-of-step: θ_normal < Score ≤ θ_lost, the communication deteriorates slightly. At this time, the controller will actively adaptively adjust the gain of power compensation according to the state machine status label. For example, it will halve the compensation gain and operate with a conservative strategy.

[0095] Conforming: Score > θ_lost, or multiple consecutive messages are lost, and the communication deteriorates severely. The controller enters the conforming state, freezes the last received sub-pile execution reference, and continues to perform local compensation based on this as a benchmark to maintain the power waveform and wait for the communication to recover.

[0096] Derating: The communication is interrupted for a long time or there is an internal fault in the controller, and it enters the safe derating or shutdown state.

[0097] Budgeted compensation and online passivity maintenance: In the normal or out-of-step state, the controller measures the deviation of the local bus frequency and voltage, and calculates an original power compensation instruction according to its compensation gain (which has been adaptively adjusted in the out-of-step state).

[0098] Virtual impedance superposition: The controller superimposes the virtual impedance configuration onto the power compensation control loop. This means that before generating the final PWM drive signal, the control algorithm simulates the effect of the virtual resistance R_v and virtual inductance L_v configured in Example 5.

[0099] Online Passivity Margin Assessment: Simultaneously, using the equivalent impedance frequency set of the power grid as a benchmark, the passivity margin for the current cycle is assessed online, and a passivity compliance flag is generated. The controller uses the most recently received Z_grid(jω) data from the master station, combined with its current Z_device(jω), to calculate Re{Z_device+Z_grid} in real time. If this value is lower than the margin δ at any frequency point, a non-compliance flag is generated.

[0100] Safety Arbitration: Safety arbitration is performed on the power compensation results based on the passive compliance flag, resulting in a final compensation instruction. If the flag is non-compliant, the arbitration logic will immediately take measures, such as temporarily increasing the value of the virtual resistor R_v or further reducing the compensation gain, until the passive margin is restored, thereby ensuring that the system operates within the stable region at all times.

[0101] Error-constrained ramp reset: When communication recovers from severe degradation (conformal state), the local controller receives a completely new stub execution reference, which may deviate significantly from the currently executing power value. To avoid instruction jumps, when the state machine state label enters the conformal state, the deviation between the slow power instruction and the fast channel execution trajectory is detected; and based on the deviation, the error-constrained ramp reset mechanism is initiated, realigning the fast channel execution trajectory to the latest slow power instruction at a controlled slope. In other words, the controller does not immediately follow the new instruction, but rather smoothly transitions from the current power value to the new reference value using a ramp constrained by the upper limit of the power change rate (determined by the budget in Example 5).

[0102] Coupled Execution and Online Perturbation Identification: To achieve continuous system characteristic monitoring, this method injects an orthogonal perturbation sequence into the final compensation command. This orthogonal perturbation sequence has the characteristics of zero mean, bounded amplitude, and cross-correlation with perturbation sequences from other devices below a preset threshold. For example, each controller is assigned a unique pseudo-random binary sequence as its perturbation signal, with its amplitude limited to within 5% of the compensation command. Furthermore, the injection of the orthogonal perturbation sequence, the error-constrained ramp reset mechanism, the superposition of virtual impedance configurations, and the application of device-level gain upper limit budgets are all coupled and executed under the unified scheduling of state machine state tags. This is a highly integrated control logic: for example, perturbation injection is only activated in the normal state; the virtual impedance is maintained in both normal and out-of-step states, but may be adjusted to a more conservative value in conformal states; the ramp reset mechanism is triggered when exiting a conformal state. This coupled design ensures that various advanced functions can operate safely and harmoniously in complex communication and power grid environments.

[0103] Finally, the power command generated after all the above steps is the aligned fast-channel execution trajectory. It is sent to the converter's underlying control loop for execution and uploaded to the main station for subsequent fusion arbitration.

[0104] Example 7 describes the fast and slow channel fusion arbitration and cross-cycle adaptive correction process, and explains how to safely and intelligently fuse the medium and low speed planned power trajectory (slow channel) and the high speed dynamic compensation trajectory (fast channel) in the final execution stage of the aggregation control method, and how the system achieves cross-cycle self-learning and optimization by evaluating the execution results.

[0105] In this embodiment, the core steps of the method are to fuse the stub execution reference, the local fast control configuration set and the aligned fast channel execution trajectory, and to perform priority arbitration under the constraints of the power hard constraint mapping to output the final execution power trajectory.

[0106] Fusion and Priority Arbitration: In each control cycle, the aggregated control master station or the edge controller with the function decentralization will synthesize the stub execution reference (slow channel, denoted as P_ref(t)) obtained from Example 3 and the aligned fast channel execution trajectory (fast channel, denoted as P_fast(t)) obtained from Example 6.

[0107] The initial synthesized power command P_syn(t) is the algebraic sum of the two commands: P_syn(t) = P_ref(t) + P_fast(t). Simultaneously, its rate of change dP_syn(t) / dt is calculated. Subsequently, the controller continuously monitors whether the synthesized power command exceeds the boundaries of the power hard constraint mapping. These boundaries refer to the upper limit of the power peak value and the upper limit of the power rate of change generated in Example 3, which vary with time.

[0108] The arbitration mechanism is activated when P_syn(t) or dP_syn(t) / dt is detected to exceed the boundary. If the boundary is exceeded, the stub execution reference is pruned according to the priority rule of frequency support first, voltage support second, and trajectory tracking third, to ensure that the fusion result conforms to the power hard constraint mapping.

[0109] The fast-channel execution trajectory P_fast(t) can be decomposed into P_fast(t) = P_freq(t) + P_volt(t), where P_freq(t) is the compensation component generated in response to the grid frequency deviation, and P_volt(t) is the compensation component generated in response to the voltage deviation.

[0110] Arbitration logic: When a conflict occurs, the controller reduces or trims instructions in the following order:

[0111] First, attempt to reduce the lowest priority trajectory tracking component, i.e., P_ref(t) of the slow channel. For example, if dP_syn(t) / dt exceeds the limit, prioritize limiting the rate of change of P_ref(t), or even temporarily reverse it, to make room for higher priority compensation.

[0112] If reducing P_ref(t) alone is insufficient to bring the synthesized instruction back within the boundary, the controller begins to reduce the next-priority voltage support component, P_volt(t).

[0113] The highest priority frequency support component P_freq(t) will be preserved to the maximum extent possible, and will only be slashed in extreme cases.

[0114] This priority rule embodies the highest principle of ensuring the overall safety and stability of the power grid. The stability of the power grid frequency is the primary concern for system safety, followed by the stability of local voltage, while the port area's own economic operation plan (reflected in P_ref(t)) has the lowest priority. Through this clear arbitration mechanism, it is ensured that in any emergency, the virtual energy storage system can provide the most favorable support to the power grid first and to the greatest extent possible, while never exceeding the hard power constraint boundary determined by its own physical characteristics. This achieves the goal of providing strong external support while ensuring absolute internal safety.

[0115] To enable the system to have self-learning and self-optimization capabilities, the present invention further includes: comparing the final execution power trajectory with the sub-stake execution reference, and combining the three-band energy index of the execution cycle to calculate an evaluation index set; based on the evaluation index set, decomposing and generating budget adjustment amount, virtual impedance fine-tuning amount, and window and buffer band correction amount for the next cycle; updating the local fast control configuration set using the budget adjustment amount and virtual impedance fine-tuning amount, and updating the polarity holding window and buffer band parameters using the window and buffer band correction amount.

[0116] The main station's evaluation module is activated once every 15 minutes over a longer timescale.

[0117] The evaluation index set is a set of numerical values ​​that quantify the system performance, such as: the integral value of frequency / voltage deviation, the number of times the arbitration mechanism is triggered, the peak and average values ​​of the three-band energy index, the root mean square value of DC ripple, and the tracking error between the final executed power trajectory and the sub-stake execution reference.

[0118] Generate Correction Quantities: A decision engine based on expert rules or fuzzy logic generates correction quantities for key control parameters based on a set of evaluation metrics. For example:

[0119] If the assessment finds that the arbitration mechanism is frequently triggered due to exceeding the power change rate constraint, the decision engine will generate a negative budget adjustment amount, appropriately reducing the total compensation bandwidth budget allocated in Implementation Example 5 in the next cycle, so that the behavior of the fast channel is more convergent.

[0120] If the assessment finds that the energy index of a certain frequency band remains high even within hard constraints, a positive virtual impedance fine-tuning amount may be generated to slightly increase the virtual impedance value of that frequency band in the next cycle to provide stronger damping.

[0121] If the assessment finds that the actual process of the power zero-crossing point is still accompanied by energy spikes, a positive window and buffer band correction will be generated to further widen the adaptive window parameters in Example 4 in the next cycle.

[0122] This closed-loop correction mechanism constructs a top-level learning loop spanning multiple core modules. It enables the system to learn from its historical performance, automatically discover and correct potential parameter mismatches or suboptimal configurations, thereby continuously converging towards a safer, more efficient, and more stable operating point without human intervention, demonstrating a high degree of intelligence.

[0123] Example 8 describes a group coordination control process based on timing misalignment, which can replace the online execution compensation method based on passive constraints in Example 6, and is especially suitable for scenarios with limited communication bandwidth or low requirements for controller computing power.

[0124] In this embodiment, the step of obtaining the aligned fast-channel execution trajectory through online compensation is replaced by: calculating the response time constant based on the rated power and maximum power change rate of each device in the aggregated scenario and resource data, and dividing the device into at least two response groups according to the response time constant; determining the power command issuance timing misalignment interval between at least two response groups, and issuing commands to each response group sequentially according to the timing misalignment interval.

[0125] Dynamic grouping of devices based on responsiveness:

[0126] First, calculate the response time constant τ_i = P_rated_i / (dP / dt)_max_i for each available charging pile i; where P_rated_i is its rated power and (dP / dt)_max_i is its maximum power change rate. The smaller the value of τ_i, the faster the device response. Then, based on the value of τ_i, use a clustering algorithm (such as K-Means) or a simple thresholding method to divide all devices into several response groups. For example, devices with τ < 50ms are divided into the fast response group, those with 50ms ≤ τ < 200ms are divided into the medium response group, and those with τ ≥ 200ms are divided into the slow response group.

[0127] This grouping method transforms a large number of heterogeneous devices into several homogeneous groups with similar response characteristics. This approach of consolidating disparate components greatly simplifies control complexity, eliminating the need for the controller to independently schedule hundreds or thousands of devices. Instead, it manages only a few groups of devices with consistent behavior.

[0128] Determining and distributing timing misalignment intervals in sequence:

[0129] When compensation of a total power of P_total is required, this power is allocated to various response groups. The controller plans the startup sequence for each group, typically prioritizing the fast response group. To avoid a massive impact on the power grid caused by simultaneous operation of all devices, the startup times of the groups are artificially staggered by a timing misalignment interval Δt_group. This interval can be calculated based on the allowable rate of change of the bus voltage, for example, Δt_group = P_step_group / (dV / dt)_allow; where P_step_group is the power step generated by the startup of a response group, and (dV / dt)_allow is the allowable rate of change of the bus voltage defined in the device constraint set. The controller issues commands sequentially according to the timing sequence of fast response group -> Δt_group -> medium response group -> Δt_group -> slow response group.

[0130] The core idea of ​​this method is to transform simultaneity into sequence. By staggering the timing, a large power surge is decomposed into multiple small, tolerable power steps. Thus, without relying on complex stability control (such as virtual impedance), a programmed, open-loop scheduling method is used to smooth the dynamic process of the system.

[0131] Furthermore, in order to introduce feedback regulation into this open-loop scheduling, this embodiment also includes: monitoring the bus voltage change caused by the initiation of any response group, and dynamically adjusting the timing misalignment interval of subsequent response groups based on the bus voltage change; and when the actual response power of any response group is found to be lower than the planned value, calculating the compensation time and advancing the instruction issuance time of the next response group accordingly.

[0132] Interval adaptive adjustment: For example, after the fast response group is started, the controller monitors the actual bus voltage sag ΔV_measured. If ΔV_measured is much smaller than the allowable value, the controller will dynamically shorten Δt_group before the next group starts to speed up the overall response. Conversely, if ΔV_measured is close to the limit, Δt_group will be extended.

[0133] Compensation shift mechanism: If the controller detects that the actual power output P_actual of the fast response group within the specified time is lower than the planned value P_plan, it will calculate the power deficit ΔP = P_plan - P_actual and estimate the time Δt_comp required for the next response group (medium response group) to make up this deficit. Then, the original start time of the medium response group will be advanced by Δt_comp to ensure the timeliness of the total power response.

[0134] Example 11 describes the core algorithm in Example 3, namely, how to generate a specific power hard constraint value based on the system state and constraints.

[0135] Setting input parameters and boundary conditions: Assume that at a certain time t, the aggregated control master station obtains the following input data:

[0136] The rated voltage of the DC bus is 750V. The DC ripple limit V_lim defined in the equipment constraint set is that the peak-to-peak value does not exceed 5% of the rated voltage, that is, V_lim=750V×5%=37.5V.

[0137] Through online identification (as in Example 3), the gain norms ||G_P→V||_b of the power-to-DC ripple transfer characteristics under the current operating conditions in three key frequency bands were obtained as follows:

[0138] Low frequency band (b=1): ||G_P→V||_1=0.1V / kW;

[0139] Double the grid bandwidth (b=2): ||G_P→V||_2=0.5V / kW;

[0140] Switching sideband (b=3): ||G_P→V||_3=0.2V / kW;

[0141] Through real-time calculation (as in Example 3), the current three-band energy index E_b was obtained as follows:

[0142] Low frequency band (b=1): E_1=4.0(kW)²·s.

[0143] Twice the grid bandwidth (b=2): E_2=1.0(kW)²·s.

[0144] Switching sideband (b=3): E_3=0.25(kW)²·s.

[0145] The safety factor α_b for all frequency bands is uniformly set to 0.8.

[0146] The value ε for preventing the division of zero into small positive numbers is set to 10^-6.

[0147] Step-by-step calculation process: Based on the formula P_allow,b=α_b in Example 3. (V_lim / ||G_P→V||_b) The calculation is performed using (1 / sqrt(E_b+ε)).

[0148] Step 1: Calculate the fundamental power margin V_lim / ||G_P→V||_b for each frequency band;

[0149] Low-frequency margin = 37.5V / 0.1V / kW = 375kW;

[0150] Twice the grid bandwidth margin = 37.5V / 0.5V / kW = 75kW;

[0151] Switching margin = 37.5V / 0.2V / kW = 187.5kW.

[0152] Step 2: Calculate the energy suppression factor 1 / sqrt(E_b+ε) for each frequency band.

[0153] Low-frequency factor = 1 / sqrt(4.0+10^-6)≈0.500.

[0154] Twice the power grid frequency band factor = 1 / sqrt(1.0+10^-6)≈1.000.

[0155] The switching sideband factor = 1 / sqrt(0.25+10^-6)≈2.000.

[0156] Step 3: Calculate the independent allowable peak power P_allow,b for each frequency band, taking into account the safety factor.

[0157] The low-frequency band allowable value P_allow,1 = 0.8 × 375kW × 0.500 = 150kW.

[0158] The allowable value for the power grid frequency band is P_allow,2 = 0.8 × 75kW × 1.000 = 60kW.

[0159] The allowable value for the switch sideband is P_allow,3 = 0.8 × 187.5 kW × 2.000 = 300 kW.

[0160] Step 4: Determine the final power hard constraint

[0161] The aggregation control master station takes the minimum value among all allowed values ​​of frequency bands as the final power peak hard constraint P_hard_limit at the current moment.

[0162] P_hard_limit=min(P_allow,1,P_allow,2,P_allow,3)=min(150,60,300)=60kW.

[0163] Based on the above calculations, the peak power hard constraint generated by the system at time t is 60kW. It can be seen that although the system can withstand power fluctuations exceeding 150kW in the low-frequency band and switching sideband, the final power hard constraint is determined by the calculation results for this frequency band because the current twice-grid frequency band has the highest transfer gain (i.e., the system is most sensitive to power disturbances in this band), and there is already a certain amount of energy accumulation in this band. This calculation result will serve as an inviolable red line in the fusion arbitration process of Example 7 to ensure that no power command will cause DC bus ripple to exceed the limit, thereby guaranteeing the system's power quality and safety stability.

[0164] Example 9 describes an adaptive grouping and customized control process based on charging behavior characteristics. Instead of treating all charging piles as homogeneous resources, it identifies unique behavioral profiles by deeply mining their operational data and then dynamically groups and differentiated, refined collaborative controls based on these profiles.

[0165] In this embodiment, the steps of acquiring aggregated scene and resource data and generating stub execution references further include: extracting behavioral feature vectors containing the standard deviation of power change rate and power ripple ratio for each device based on the power curves in the synchronous dataset; and dynamically grouping the devices according to the similarity of the behavioral feature vectors to form a hierarchical group structure, wherein dynamic grouping includes merging, splitting or disbanding the groups.

[0166] Online behavioral feature extraction and updating: The aggregation control master station maintains a dynamically updated behavioral feature vector F_i for each online charging pile i. For example, within a 5-minute rolling time window, the master station extracts the power curve of the charging pile from the synchronous dataset and calculates the following features:

[0167] The standard deviation of the power change rate σ_P: std(dP / dt), reflects the stability of the equipment's power output. Equipment with frequent start-stop cycles and large power fluctuations has a higher value.

[0168] Power ripple ratio RPP: (P_max-P_min) / P_avg, reflects the magnitude of power ripple during steady-state operation of the device.

[0169] N_trans: The number of times the device switches between charging, discharging, and standby states, reflecting the intermittency of its tasks.

[0170] Charging efficiency η: P_dc·dt / P_ac·dt reflects the energy conversion efficiency of the equipment.

[0171] Battery State of Charge (SOC) and Rated Power (P_rated): These are included as static or slowly varying characteristics.

[0172] These features collectively constitute a multidimensional behavioral feature vector F_i=[σ_P,RPP,N_trans,η,SOC,P_rated]. To ensure the real-time performance and stability of the features, an exponentially weighted average method can be used to smoothly update the vector: F_i(t)=β×F_i(t-1)+(1-β)×F_i_new; where β is the forgetting factor, for example, 0.8. This allows the feature vector to smoothly reflect the recent behavioral patterns of the device.

[0173] Hierarchical Dynamic Clustering: Based on the aforementioned feature vectors, the system employs a hierarchical clustering strategy. First, coarse-grained stratification is performed based on the rated power P_rated, for example, dividing devices into high-power fast charging groups (e.g., >50kW), medium-power slow charging groups (e.g., 7-22kW), and V2G bidirectional groups. Subsequently, within each large hierarchy, a density-based clustering algorithm (e.g., DBSCAN) is used to perform fine-grained dynamic clustering based on the similarity (e.g., Euclidean distance) of other dynamic behavioral features (e.g., σ_P, RPP).

[0174] This clustering is dynamic; the system periodically monitors the compactness and separation of each cluster. When the distance between the centroids of two clusters is less than the merging threshold θ_merge, they will be merged. When the internal variance of a cluster is greater than the splitting threshold θ_split, it will be split into two or more more compact subclusters. When a cluster has too few members (e.g., less than two), the cluster will be disbanded, and its members will become discrete points or be reallocated. This dynamic management mechanism ensures that the clustering results always accurately reflect the current state of the system.

[0175] Based on this, a set of customized control strategies is matched for each group in the hierarchical group structure according to its behavioral characteristics, forming a group-customized control strategy set; wherein, the customized control strategy includes at least configuring differentiated power ripple limits or power margins for different groups, and generating stub execution references based on the group-customized control strategy set.

[0176] The system tailors control strategies for groups with different behavioral characteristics. For example:

[0177] For dynamic task-oriented fast-charging clusters with high σ_P and high N_trans behavior characteristics (typically serving high-frequency operation vehicles such as AGVs), the system matches them with a fast response strategy. This strategy allows for greater power ripple (e.g., allowing ±5% ripple) when participating in frequency regulation and gives them a higher compensation participation factor to fully utilize their fast response capabilities.

[0178] For static energy storage V2G clusters with low σ_P and low N_trans behavior (typically serving employee vehicles parked for extended periods or dedicated energy storage piles), the system is matched with a smoothing support strategy. This strategy imposes strict ripple limits on their power output (e.g., no more than ±2%) and reserves a large power margin (e.g., always maintaining a 20% power reserve), making them the main force for providing stable and reliable voltage support or peak shaving and valley filling services.

[0179] For suspected anomalies exhibiting high RPP behavior, the system applies a conservative observation strategy, temporarily reducing their weight in all power grid services and pushing diagnostic alerts to the operation and maintenance system.

[0180] When generating the final sub-stake execution reference, the aggregation control master station no longer simply distributes the total power target among all devices, but adopts a hierarchical allocation mode: First, the total power target (such as total frequency regulation capacity, total peak shaving power) is allocated among groups according to the customized strategies of each group. Then, within each group, the workload allocated to that group is further decomposed to each device within the group according to the group's internal coordination rules (e.g., considering SOC balancing).

[0181] Through this adaptive clustering and customized control, the system can achieve refined and differentiated management of massive heterogeneous resources. It goes beyond the extensive aggregation based solely on equipment nameplate parameters, instead using data-driven behavioral insights to match the right tasks to the right resource groups, thereby achieving higher control efficiency, stronger system stability, and longer equipment lifespan on a macro level.

[0182] Example 10 describes a method for generating a safety boundary based on a dynamic power envelope, providing an alternative scheme for generating a hard power constraint map. Instead of relying on complex system identification and frequency domain analysis, it employs a data-driven method based on historical data statistics and future periodic predictions to construct a dynamic, adaptive safety operating envelope.

[0183] In this embodiment, the step of generating the power hard constraint mapping is replaced by: constructing a statistical power envelope containing the power standard deviation based on historical execution data within the rolling time window; constructing a predicted power envelope using the daily periodicity and weekly similarity of the historical execution data; and weightedly fusing the statistical power envelope and the predicted power envelope to form the power hard constraint mapping.

[0184] Construction of the statistical and predictive envelope:

[0185] Statistical Power Envelope: The aggregated control master station maintains a rolling historical power database, for example, covering the past 24 hours. Within a short sliding time window T (e.g., 15 minutes), the maximum value (max(P), minimum value (min(P), and standard deviation σ_P) of the historical power P(t) within this window are calculated. Based on this, the upper envelope of the statistical power P_env_up_stat(t) = max(P(tT:t)) + k_σ × σ_P(t) and the lower envelope P_env_low_stat(t) = min(P(tT:t)) - k_σ × σ_P(t) are constructed; where k_σ is the standard deviation coefficient (e.g., 1 or 2), used to control the width of the envelope. This envelope reflects the recent actual operating range and the degree of fluctuation of the system.

[0186] Considering the strong daily and weekly periodicity of port operations, the system utilizes this prior knowledge to construct a prediction model. A simple and effective model is: P_pred(t) = P_hist(t-24h) × k_day × k_week. Here, P_hist(t-24h) is the historical power value at the same time 24 hours ago; k_day and k_week are the daily and weekly correction factors (e.g., used to distinguish load differences between weekdays and weekends). Based on this prediction curve, a predicted power envelope can also be constructed.

[0187] The final power envelope is obtained by weighted fusion of the above two: P_env_final = w_stat × P_env_stat + w_pred × P_env_pred. For example, the weights can be set to w_stat = 0.6 and w_pred = 0.4. The purpose of fusion is to take into account both the system's short-term actual fluctuations (reflected by the statistical envelope) and long-term operating patterns (reflected by the predictive envelope). This fused P_env_final serves as the basic power hard constraint mapping.

[0188] Furthermore, to enhance the intelligence and adaptability of the envelope, this embodiment also introduces dynamic margin and scene management.

[0189] First, a dynamic safety margin is calculated and applied to adjust the power hard constraint map; where the magnitude of the dynamic safety margin is proportional to the distance between the current real-time power value and the envelope boundary defined by the power hard constraint map.

[0190] At any given time t, the safety margin Margin(t) is dynamically calculated, for example: Margin(t) = M_base × (1 + k_m × dist(P_actual(t), P_env_final(t)) / Bandwidth(t)); where M_base is the base margin; k_m is the gain coefficient; P_actual(t) is the current real-time power; dist(...) is the distance from the current power to the nearest envelope boundary; and Bandwidth(t) is the total width of the current envelope. This dynamic margin can be used to adjust the controller's gain or the allowable compensation range. When the real-time power P_actual(t) is near the center of the envelope, dist(...) is at its maximum, and Margin(t) is also at its maximum, allowing the system to respond faster with greater amplitude. As P_actual(t) approaches the envelope boundary, dist(...) decreases, and Margin(t) tightens accordingly, thereby automatically suppressing aggressive control behavior that may lead to out-of-bounds violations.

[0191] Secondly, based on the current time and real-time power value information, the current operating scenario is identified; and according to the operating scenario, the corresponding scenario-related parameters such as envelope width and safety factor are extracted from the scenario parameter library to configure the power hard constraint mapping; and the envelope utilization rate of the power hard constraint mapping is periodically calculated, and the parameters in the scenario parameter library are adjusted across cycles based on the envelope utilization rate to achieve self-learning.

[0192] The system has a built-in scene recognition module that matches the system status to a preset scene library based on the current time (e.g., 7:00-9:00), total power level, number of online devices, etc., such as morning peak operation, midday stable charging, evening peak V2G support, and nighttime off-peak maintenance.

[0193] Each scenario is associated with a set of optimal envelope generation parameters in the parameter library (such as statistical window length T, fusion weight w_stat, base margin M_base, etc.). Once the current scenario is identified, the system loads the corresponding parameter set to generate a customized power envelope for that scenario.

[0194] After each scenario ends (e.g., after the daily morning rush hour scenario ends), the system calculates the envelope utilization rate for that period, i.e., Utilization = |P_actual|dt / P_env_finaldt. The system uses an ideal utilization target (e.g., 85%) as a benchmark. If the actual utilization is too low (e.g., <70%), it indicates that the envelope is too loose, and the system will automatically tighten the parameters for the same scenario the next day (e.g., reduce k_σ). If the utilization is too high (e.g., >95%), it indicates that the envelope is too tight, limiting the normal operation of the system, and the parameters will be automatically relaxed. Through this daily fine-tuning, the system can autonomously learn and optimize the power envelope for each scenario, making it both safe and efficient.

[0195] Example 12 provides a refined and enhanced implementation method for the data aggregation and subsequent instruction issuance process in Example 1.

[0196] First, in the data preprocessing stage, to improve the accuracy of subsequent algorithms, this embodiment further introduces a noise spectrum-based weighted processing mechanism based on the cleaning steps of Embodiment 1. Specifically, the aggregation control master station performs offline or online statistical analysis on the synchronous datasets of each channel collected historically to estimate the noise power spectral density of each measurement channel (e.g., the bus voltage of a specific area, the power feedback of a certain type of charging pile). It can be seen that some channels may exhibit significant noise in specific frequency bands due to factors such as sensor quality and electromagnetic environment. When performing subsequent calculations dependent on the data, such as identifying the power-to-DC ripple transfer characteristics in Embodiment 3, or extracting trends in Embodiment 4, the algorithm uses this noise spectrum information as a weighting factor. For data channels with lower noise levels in key frequency bands, their weight in the calculation is higher, and vice versa. In this way, noise interference can be effectively suppressed, ensuring that the identified model and extracted trends are closer to the real physical process.

[0197] Secondly, at the command issuance and execution level, to address the issue of command execution desynchronization caused by inconsistent network latency in distributed systems, this embodiment employs a collaborative control mechanism based on effective timestamps. Specifically, the time synchronization strategy module in the aggregation control master station orchestrates a unified effective time for all slow commands and parameter packets requiring collaborative actions. For example, after the master station completes a calculation and generates a new local fast control configuration set (from Embodiment 5) and a sub-pile execution reference (from Embodiment 3), it does not immediately issue and require execution. Instead, the master station encapsulates these two data packets and attaches a unified, future effective timestamp to them, for example, September 2, 2025, 14:30:00.500. Upon receiving this timestamped command packet, the local controller of each charging pile stores it in a buffer and continues executing the old commands. When the local controller's own clock, synchronized via GPS / BeiDou, reaches the precise moment of the effective timestamp, it instantaneously and atomically activates the new configuration set and execution reference. Understandably, through this mechanism, even if the command packets arrive at different controllers at different times due to different network paths, all devices can eventually switch to the new working state at the same time, thereby ensuring the precise synchronization of virtual energy storage resources in the entire port area in terms of control logic and power targets, and avoiding system transient oscillations that may be caused by inconsistent states.

[0198] Example 13: This example aims to provide a more detailed description of several key details in the core control algorithm of this invention, and to explain the robustness strategy designed to ensure that the system can still operate stably under non-ideal or abnormal conditions.

[0199] In the online compensation execution process of Example 6, a ramp controller is commonly used to limit the rate of power change. In a preferred implementation of this example, a second-order power ramp is employed. Compared to a traditional first-order ramp (where power changes linearly at a constant rate of change), the power output curve of a second-order ramp is S-shaped. This means that the power change first smoothly increases from zero and then smoothly decreases back to zero, and the rate of change of its rate of change (i.e., power acceleration) is limited. In discrete implementations, a first-order ramp only limits the difference between the target and the current output, while a second-order ramp performs a second limiting filter on top of the first-order ramp. This significantly reduces the high-frequency components in the command changes, avoids unnecessary switching stress on power electronic devices, and reduces high-frequency noise injected into the grid, which is beneficial for improving power quality and extending equipment life.

[0200] In the compensation budget allocation stage of Example 5, to ensure that the budget allocated to each device does not exceed its physical limit, this example employs an iterative allocation algorithm for peak shaving and recharge. The specific process is as follows: First, the total compensation budget is initially allocated based on the compensation participation factors of each device. Then, it is checked whether any device's allocated budget exceeds its maximum power change rate limit. If so, all excess portions are summed to form a recharge pool. In the next iteration, the budget in the recharge pool is redistributed according to the relative proportions of the participation factors of all unsaturated (i.e., devices that have not reached their physical limits). This process is repeated until the recharge pool is empty or all devices are saturated. For example, with a total budget of 100 kW / s, the participation factors of devices A, B, and C are 0.5:0.3:0.2, and the physical limit of B is 25 kW / s. The initial allocation yields A=50, B=30, and C=20. B exceeds its limit by 5 kW / s. This 5 kW / s was placed into the reinjection tank and redistributed according to the participation factor ratio of A to C of 5:2. A increased by 5 × (0.5 / (0.5+0.2)) ≈ 3.57, and C increased by 5 × (0.2 / (0.5+0.2)) ≈ 1.43. The final allocation result was approximately A = 53.57, B = 25, and C = 21.43. This refined allocation method ensured that the total budget was fully utilized while strictly respecting the physical constraints of each device.

[0201] In the process of generating the branch station execution reference in Example 3, to improve efficiency and control targeting, this example introduces intermediate steps of charge-only / discharge-only grouping and power corridor shaping within each group. When the total power target is net discharge, the system does not require all vehicles to reduce charging. Instead, an optimization solver prioritizes selecting a group of vehicles with higher SOC and allowing discharge to form a discharge-only group, while selecting another group of vehicles to form a charging-only group, such that the algebraic sum of the power of the two groups is exactly equal to the total power target. This reduces unnecessary bidirectional energy flow and improves the overall system efficiency. After forming the groups, the system designs a dynamic power corridor for each group (i.e., upper and lower power limits and upper and lower limits of the rate of change). This corridor comprehensively considers factors such as the average SOC of all members in the group, the highest battery temperature, and the expected off-site time. The final power allocation is carried out under the constraints of this corridor.

[0202] Rollback strategy under abnormal operating conditions: To improve the robustness of the system, anomaly handling and rollback paths are designed.

[0203] Optionally, in the power grid impedance identification stage of Embodiment 5, if the signal-to-noise ratio of the response signal collected after the disturbance is too low, resulting in the confidence level of the identification result being lower than a preset threshold (e.g., coherence less than 0.7), the system will abandon the identification result and automatically load a set of default virtual impedance parameters that are pre-stored locally and applicable to most operating conditions, while sending an alarm for impedance identification failure to the upper-level system.

[0204] Optionally, in the power hard constraint mapping calculation stage of Embodiment 3, if the calculated upper limit of the power change rate is lower than the minimum controllable step size that all devices can physically achieve, it indicates that the current system state is very fragile. At this time, the system will trigger linkage protection. On the one hand, it will forcibly extend the polarity maintenance window and widen the buffer band through the mechanism of Embodiment 4. On the other hand, it will actively impose a greater slope limit on the slow channel to reduce the demand for power fluctuations from the source until the hard constraint calculation result recovers to a reasonable range.

[0205] In a preferred implementation of this embodiment, the system recognizes that the power-to-DC ripple transfer characteristics (Embodiment 3) or the grid equivalent impedance (Embodiment 5) are not static, but are closely related to the overall operation of the port area. For example, the dynamic characteristics of the system differ greatly between two scenarios: multiple quay cranes operating simultaneously during the day (providing shore power) and only AGVs charging at night.

[0206] The system identifies various typical, repeatable operating scenarios (e.g., peak daytime operation, concentrated nighttime charging, and low weekend load) through offline analysis or online learning, and establishes a version library of transfer characteristics or grid impedance. Each version is associated with a scenario tag. During actual operation, a scenario identification module determines in real time which predefined scenario the system is in based on information such as the current time, total load level, and operating status of key equipment. Once the scenario is determined, the controller loads the optimal model parameters matching that scenario from the version library for subsequent power hard constraint calculations or virtual impedance configuration.

[0207] This scenario-adaptive mechanism enables the core algorithm of this invention to always make decisions based on the most accurate system model. Compared with using a single, averaged model, the accuracy and stability of its control are greatly improved.

[0208] To facilitate system maintenance, fault diagnosis, and continuous optimization, this embodiment constructs a complete closed-loop audit and data archiving system.

[0209] After each major regulatory event (e.g., a complete power grid frequency support process), or at fixed time intervals (e.g., hourly), the system automatically generates and stores an audit and replication instance. This instance is a highly structured data snapshot that includes not only the final execution result but also the entire chain of decision-making basis leading to that result. Specifically, an instance includes at least:

[0210] Input data: A snapshot of the synchronized dataset at the moment the event occurred.

[0211] Intermediate calculation results: the set of equivalent impedance frequency points of the power grid identified at that time, the calculated power hard constraint mapping curve, the generated local fast control configuration set, etc.

[0212] Commands and outputs: the issued sub-slab execution reference, the fast channel execution trajectory uploaded by each device, and the final execution power trajectory of the final fusion.

[0213] Status and Alarms: Records of status machine label changes for all devices during the event and any alarm information generated.

[0214] By establishing such an auditing mechanism, when any unexpected behavior occurs in the system, operations and maintenance personnel can easily retrieve the corresponding simulated instances and replay the entire event process in an offline environment, thereby quickly locating the root cause of the problem. Furthermore, these accumulated massive amounts of instances also constitute a valuable database, which can be used for regression testing of algorithms, iterative optimization, and training of more advanced AI decision-making models, thus forming a complete closed loop of data-decision-execution-audit-optimization, providing a solid foundation for the long-term evolution of the system.

[0215] This invention constructs a power-to-DC ripple transfer characteristic model and calculates energy indicators for three key frequency bands—low-frequency band, twice the grid frequency band, and switching sideband—in real time. It then combines this with a DC ripple limit for solution, creatively transforming a frequency-domain, difficult-to-control power quality safety boundary (DC ripple) into a time-domain, controller-executable hard power constraint (i.e., variable power peak value and rate of change upper limit). This completely solves the fundamental problem of decoupling control objectives and safety boundaries in the background technology. In the practical scenario of port virtual energy storage, this means that when the aggregation control system responds to any grid dispatch command (such as fast frequency regulation), its issued power command naturally incorporates the prediction and protection of the internal DC bus stability. The system no longer relies on a static, conservative power change rate limit, but can dynamically calculate the maximum power change capacity it can safely withstand at any given moment based on the current system state (reflected in the frequency band energy indicators) and system characteristics (reflected in the transfer function). Therefore, when the system is in good condition, it can unleash its full potential, providing faster and greater response and achieving higher returns; while when the system is near the unstable boundary, it can instantly and accurately self-regulate to avoid triggering internal protection due to overreaction, ensuring the power supply safety of the port area's critical production loads and achieving a unity of high-performance external response and high-reliability internal operation.

[0216] This invention proactively injects minute disturbances into the power grid to identify its equivalent impedance online. Based on this, and using the passive theory Re{Z_device+Z_grid}≥δ, the feasible region of the virtual impedance parameters is solved, thus transforming the complex system stability problem into a pre-solvable parameter configuration problem with well-defined boundaries. This method directly solves the high-bandwidth control instability risk inherent in the background technology due to the interaction of numerous converters with an unknown power grid. In a port virtual energy storage system, this means that before activating the high-bandwidth power grid support function, the system has already configured itself with a set of stable parameters (i.e., virtual impedance parameters) through scientific calculations to ensure peaceful coexistence with the current power grid and to prevent resonance or oscillation. Compared with existing technologies that rely on simulation or on-site trial-and-error parameter adjustments, this invention provides an adaptive, online-executable stability guarantee mechanism. Therefore, even if the port power grid topology changes due to factors such as ship berthing (activation of shore power) or internal load switching, leading to changes in grid impedance, this system can maintain itself within the stable operating region through periodic re-identification and recalculation. This greatly enhances the reliability and credibility of virtual energy storage systems as commercial ancillary service providers, enabling them to safely participate in more demanding grid services without concern for causing instability to the port area or the upstream grid.

[0217] This invention addresses the challenges of balancing multiple control objectives and achieving long-term system performance evolution under complex constraints by constructing a fast-slow channel fusion and priority arbitration mechanism, and designing a closed-loop system for cross-cycle adaptive correction at the top level. This represents an integration and enhancement of the two core innovations mentioned above at the system level. The priority arbitration mechanism ensures that, under any circumstances, support for grid frequency has the highest priority, followed by voltage support, and finally, tracking of internal planned power. This design enables the Virtual Energy Storage System (VESS) to contribute its capacity to the maximum extent without hesitation when the grid urgently needs support, which is crucial for improving the grid's emergency recovery capabilities and preventing the situation from escalating. The cross-cycle adaptive correction mechanism, on the other hand, continuously evaluates the actual operating performance of the system (such as arbitration trigger frequency and ripple suppression effect) and fine-tunes key parameters such as the conservatism of power hard constraints, virtual impedance parameters, and the width of the zero-crossing window. This self-learning capability enables the system to continuously accumulate experience during use, autonomously adapt to seasonal changes in port load characteristics or performance drift caused by equipment aging, and maintain itself in a near-optimal operating range in the long term. This maximizes its technical and economic benefits while ensuring absolute safety and significantly reduces reliance on manual continuous tuning and maintenance.

[0218] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A port green electricity virtual energy storage aggregation regulation method, characterized in that, The method comprises the following steps: obtaining aggregated scenario and resource data, the aggregated scenario and resource data comprising: a synchronous data set of virtual energy storage, a baseline power trajectory, a device constraint set, and a vehicle and charging facility available capacity set; based on the aggregated scenario and resource data, generating a power hard constraint mapping and a sub-priority execution reference by calculating a multi-band energy index and identifying system power transmission characteristics, the energy index comprising: power quality indicators; based on the aggregated scenario and resource data and the power hard constraint mapping, generating a local fast control configuration set by solving a virtual impedance feasible region and apportioning a compensation budget, and obtaining an aligned fast channel execution trajectory by online compensation; fusing the sub-priority execution reference, the local fast control configuration set, and the aligned fast channel execution trajectory, and performing priority arbitration under the constraint of the power hard constraint mapping to output a final execution power trajectory of the virtual energy storage; generating the power hard constraint mapping and the sub-priority execution reference comprises: based on the device direct current ripple limit value contained in the device constraint set in the aggregated scenario and resource data, reversely deducing the upper limit of the power peak value and the change rate of the power domain to form the power hard constraint mapping; and using the power hard constraint mapping to constrain and allocate the baseline power trajectory in the aggregated scenario and resource data to generate the sub-priority execution reference; the reversely deduced upper limit of the power peak value and the change rate of the power domain comprises: based on the aggregated scenario and resource data, calculating the power change energy in the low frequency band, the twice grid frequency band, and the switching sideband to obtain a three-band energy index; based on the aggregated scenario and resource data, identifying the amplitude-frequency transfer characteristics of the power change to the direct current voltage ripple to obtain the power to direct current ripple transmission characteristics; combining the three-band energy index and the power to direct current ripple transmission characteristics, and solving with the direct current ripple limit value as the boundary condition to determine the upper limit of the power peak value and the change rate; the generated power hard constraint mapping and the sub-priority execution reference further comprise: monitoring the zero-crossing times of the aggregated power to obtain a zero-crossing count; aggregating the low-frequency band energy reflected by the aggregated scenario and resource data and the insulation monitoring alarm to generate a zero-crossing risk index; and dynamically adjusting the polarity maintaining window length and the buffer bandwidth used to constrain the sub-priority execution reference according to the zero-crossing count and the zero-crossing risk index.

2. The port green electricity virtual energy storage aggregation regulation method according to claim 1, characterized in that, Further comprising: at the boundary of the polarity maintaining window, performing two-stage smooth switching, wherein the power change rate is slowed down to near zero by using trend feedforward; enabling a second-order ramp control to complete the zero-crossing switching of the power polarity, and during the switching, if the energy index of any frequency band is monitored to rise, the execution time of the second-order ramp is extended.

3. The port green electricity virtual energy storage aggregation regulation method according to claim 1, characterized in that, generating the local fast control configuration set comprises: based on the aggregated scenario and resource data, identifying an equivalent impedance frequency point set of the power grid by injecting a power disturbance sequence and collecting the response; solving a virtual impedance parameter feasible region that satisfies the preset frequency domain passivity condition in combination with the equivalent impedance frequency point set of the power grid; selecting a group of parameters from the virtual impedance parameter feasible region to form a virtual impedance configuration, and taking it as a component of the local fast control configuration set.

4. The port green electricity virtual energy storage aggregation regulation method according to claim 3, characterized in that, online compensation to obtain the aligned fast channel execution trajectory comprises: In each local control cycle, the virtual impedance configuration is superimposed to the power compensation control loop; Based on the equivalent impedance frequency point set of the power grid, the passive margin of the current cycle is evaluated online, and a passive compliance flag is generated; According to the passive compliance flag, the result of power compensation is executed for safety arbitration, and the final compensation instruction is formed.

5. The port green electricity virtual energy storage aggregation regulation method according to claim 3, characterized in that, Generating the local fast control configuration set further includes: determining the total compensation bandwidth budget of the entire station according to the power hard constraint mapping; Based on the aggregated scenario and resource data, the compensation participation factor of each device is calculated; According to the compensation participation factor, the total compensation bandwidth budget is allocated to each device to form a device-level gain upper limit budget, which is included in the local fast control configuration set.

6. The method of claim 4, wherein, The online execution of the compensation obtained aligned fast channel execution trajectory further includes: Collecting communication indicators such as communication delay, jitter and packet loss rate, and integrating them into a communication reliability score; According to the communication reliability score and the preset threshold, the state machine state label is determined, which includes any one of normal, step out of step, shape preserving and derating; Before executing the final compensation instruction, the gain of power compensation is adaptively adjusted according to the state machine state label.

7. The port green virtual energy storage aggregation regulation method according to claim 1, characterized in that, Fusing the segmented execution reference, the local fast control configuration set and the aligned fast channel execution trajectory, and performing priority arbitration under the constraint of the power hard constraint mapping includes: When fusing the segmented execution reference and the aligned fast channel execution trajectory, it is detected in real time whether the synthesized power instruction exceeds the boundary of the power hard constraint mapping; If it exceeds the boundary, according to the priority rules of frequency support first, voltage support second, and trajectory tracking third, the segmented execution reference is cropped to ensure that the fusion result complies with the power hard constraint mapping.

Images