A network-constructing type mobile energy storage transient voltage coordination support method

Abstract

This invention discloses a method for coordinated transient voltage support of grid-connected mobile energy storage. It employs a two-layer response mechanism combining fast response and consistency optimization. When each grid-connected mobile energy storage node detects a transient fault, it triggers a local fast response, calculates reactive power, and corrects the voltage outer loop reference setpoint of the grid-connected control. In parallel, consistency optimization is initiated. Under quasi-steady-state conditions, electrical parameters are measured, multi-factor comprehensive local weights are calculated, and a distributed consensus algorithm negotiates the weights and total reactive power demand to calculate optimized power and adaptively adjust voltage control parameters before smoothly switching to optimized output. In steady-state conditions, distributed power redistribution is triggered based on the SOC distribution. After the grid voltage stabilizes, each node negotiates and synchronously reduces power to smoothly exit and restore the baseline parameters. This invention achieves distributed coordinated support, balancing transient fast response and steady-state optimization, supporting plug-and-play and dynamic entry / exit of nodes, effectively improving the transient voltage stability of the power system and extending the energy storage support time.

Description

Technical Field

[0001] This invention relates to the field of transient voltage control technology in power systems, and in particular to a method for coordinated support of transient voltage in grid-connected mobile energy storage. Background Technology

[0002] During the power system's transition to a higher proportion of renewable energy, the grid inertia continues to decline, and transient voltage stability issues are becoming increasingly prominent. Distribution networks, microgrids, and other scenarios are prone to significant voltage drops under fault disturbances, affecting power supply reliability. Mobile energy storage, with its flexible deployment and rapid response, has become an important means of transient voltage support; however, existing mobile energy storage voltage support technologies still have many shortcomings: Relying on a centralized control architecture poses a risk of single point of failure, cannot adapt to the dynamic changes in the number and deployment location of mobile energy storage devices, and has poor scalability; The control strategy is too simple and lacks a hierarchical design that combines rapid transient response with steady-state coordination optimization. Either the response speed is too slow to suppress transient voltage drops, or the lack of global optimization leads to low energy storage capacity utilization efficiency and SOC imbalance. The access process is complex, requiring manual configuration of electrical parameters and communication information, lacking plug-and-play capability, and unable to be quickly put into operation in emergency fault scenarios. The control parameters are fixed values, and the electrical coupling differences between the mobile energy storage access point and the fault point are not considered. The transient response characteristics do not match the grid operating conditions. SOC balancing strategies rely on centralized computing, which is difficult to implement in distributed scenarios. During long-term support, some energy storage systems may prematurely exit due to excessively low SOC, reducing the overall support duration.

[0003] Network-based control technology enables mobile energy storage to proactively establish voltage and frequency references, eliminating reliance on grid phase-locked loops and providing superior voltage support capabilities in weak grid and transient disturbance scenarios. Therefore, it is crucial to integrate network-based control with distributed coordination technology to design a transient voltage coordination support scheme adapted to the characteristics of mobile energy storage, resolving the technical contradictions between rapid response and global optimization, centralized control and distributed deployment, and fixed parameters and operating condition adaptation. Summary of the Invention

[0004] The purpose of this invention is to provide a method for coordinated transient voltage support of grid-based mobile energy storage. It adopts a two-layer response mechanism of fast response and consistency optimization, combined with distributed consensus negotiation, plug-and-play self-organizing network, SOC-driven power redistribution and adaptive adjustment of control parameters, to achieve decentralized coordinated voltage support for mobile energy storage. This achieves the technical goals of rapidly suppressing voltage drops in the early stage of transients, optimizing capacity utilization in the steady state, rapid deployment of plug-and-play equipment, and long-term support for SOC balance, thereby improving the transient voltage stability of the power system.

[0005] To achieve the above objectives, the present invention provides the following solution: A method for coordinating and supporting transient voltage in a grid-based mobile energy storage system employs a two-layer response mechanism combining fast response and consistency optimization, comprising the following steps: S1, Initialization and Standby: After the grid-type mobile energy storage node is connected to the grid, it completes self-test grid connection, neighbor discovery, coordination group joining and parameter initialization in sequence to achieve plug-and-play and enter standby state; S2, Fault Detection and Rapid Response: Each grid-type mobile energy storage node independently monitors the voltage. When a transient fault is detected, a rapid response is triggered. The reactive power output is directly calculated based on the local voltage deviation, and the reference voltage setpoint of the outer voltage loop is corrected based on the voltage-reactive power droop characteristics of the grid-type control. S3, Consistency Negotiation and Optimization: After the fast response is started, consistency optimization is started in parallel. After the transient process enters the quasi-steady state, the steady-state voltage deviation and voltage recovery speed are measured, the multi-factor comprehensive local weight is calculated, the total weight and total reactive power demand are negotiated through distributed consistency iteration, the optimization target power is calculated, and the voltage control parameters are adaptively adjusted. Finally, the reactive power output is gradually transitioned from the fast response output to the consistency optimization output through smooth switching. S4, Steady-state coordination and SOC balance: The system enters steady-state coordination operation, monitors the SOC distribution in real time, and when the SOC deviation of a certain node meets the triggering condition, it performs distributed power redistribution to achieve global SOC balance. S5, Support Exit and Equipment Recovery: After the grid voltage is detected to have stabilized, each grid-type mobile energy storage node synchronously reduces reactive power output through consensus negotiation, smoothly exits support and restores the baseline control parameters, returns to standby state, and continues to monitor the grid; if a node needs to exit, a transparent exit process is executed to complete the task transfer.

[0006] Further, step S1, initialization and standby, specifically includes: After the network-type mobile energy storage node is deployed and connected, it sequentially performs device self-test and soft grid connection, broadcasts HELLO messages to discover neighbors, sends JOIN_REQUEST messages to join the coordination group, initializes parameters, completes plug-and-play access, and enters standby mode; in standby mode, it periodically sends HELLO and STATE_SYNC messages to maintain communication connection with neighboring nodes; The device self-test and soft grid connection include: performing core safety checks and reading real-time data from the battery management system, skipping non-critical self-tests; and performing fast soft grid connection after detecting that the voltage and frequency at the access point meet the grid connection conditions.

[0007] Furthermore, in step S2, the reactive power output is directly calculated based on the local voltage deviation. Specifically, it includes:

[0008] in, For fast response gain coefficient, This is due to local voltage deviation. Rated power, This is the SOC correction factor; Based on the voltage-reactive power droop characteristics of grid-type control, the reference voltage setpoint of the outer voltage loop is corrected, specifically by changing the no-load voltage setpoint. This allows the actual output reactive power to track the calculated reactive power output. .

[0009] Further, step S3, consensus negotiation and optimization, specifically includes: S301, Consistency Optimization Startup and Steady-State Measurement: After the fast response startup, consistency optimization is started in parallel. The system enters a quasi-steady state after the transient process ends by using a sliding window. S302, Multi-factor comprehensive weight and total reactive power demand negotiation: Each network-type mobile energy storage node calculates a comprehensive local weight, and iteratively negotiates with its neighbors through a distributed average consensus algorithm to achieve global consistency between the normalized weight and the total reactive power demand of the system. S303, adaptive adjustment of control parameters: Each grid-type mobile energy storage node calculates its relative position index based on the global maximum voltage deviation, and adaptively adjusts the voltage loop bandwidth and voltage droop coefficient according to the index. S304, Smooth Switching Between Fast Response and Optimized Output: Employs an exponential decay weighted strategy to gradually switch reactive power output from fast response output to consistent optimized output.

[0010] Further, in step S301, the transient process is determined to be over by using a sliding window. Specifically, a sliding window with a width of 50ms is used. When the absolute value of the voltage change rate in two consecutive sliding windows is less than 0.002pu / s, the transient process is determined to be over. The difference between the voltage before the fault and the current latched steady-state voltage is calculated as the steady-state voltage deviation.

[0011] Furthermore, in step S302, the comprehensive local weight is a power product of multiple factors, including voltage deviation factor, voltage recovery speed factor, capacity factor and SOC factor, wherein the voltage deviation factor plays a dominant role in the comprehensive weight. The basic iterative formula of the distributed average consensus algorithm is as follows: Node i In the k+ In the first iteration, update its estimated sum of weights:

[0012] in, For nodes The neighborhood group, For step size parameters, Represents a node In the The sum of weights estimated in the next iteration. Represents a node neighboring nodes In the The sum of weights estimated in the next iteration. Represents a node In the The sum of the estimated weights updated in the next iteration; The distributed average consensus algorithm adopts an adaptive step size strategy: a large step size is used for fast approximation in the early stage of iteration, a medium step size is used for stable convergence in the middle stage of iteration, and a small step size is used for precise convergence in the later stage of iteration; the initial estimated weight sum is the local weight multiplied by the number of neighbors plus 1. The convergence criterion is that the change in three consecutive iterations is less than 1% of the current value or the maximum number of iterations is reached. After convergence, each node obtains a consistent total weight. The normalized weight is obtained by dividing the local weight by the total weight, ensuring that the sum of all weights is 1.

[0013] Further, in step S303, after the relative position index obtains the global maximum voltage deviation through consensus negotiation, the ratio of the local deviation to the global maximum deviation is calculated. When adjusting the voltage loop bandwidth and voltage droop coefficient according to this index, the proportional coefficient and integral coefficient of the voltage outer loop PI controller are both obtained by multiplying the reference value by the adjustment factor. The adjustment factor is obtained by subtracting 0.3 to 0.4 times the relative position index from 1. The droop coefficient is adjusted according to the relative position index, and the reference droop coefficient is divided by 1 and then added to 0.8 times the relative position index.

[0014] Further, in step S304, the formula for the smooth switching between fast response and optimized output is as follows:

[0015] in, For nodes The final reactive power output, For nodes Fast-response reactive power output calculated directly based on local voltage deviation For nodes The optimized target reactive power is obtained through consensus negotiation. To switch weights; Switching weights As time decreases from 1 to 0, an exponential decay function is used:

[0016] in, For the steady-state determination time, This is the switching time constant.

[0017] Further, step S4, steady-state coordination and SOC equilibrium, specifically includes: Each node shares its current SOC through periodic messages; when the deviation between a node's SOC and the average SOC of its neighbors exceeds a preset threshold and the current voltage stability margin meets the conditions, a power adjustment request is initiated; neighboring nodes assess the amount of load they can share based on their own SOC margin and respond; the requesting node and the responding node jointly complete the power adjustment; during the SOC equalization adjustment, if a voltage disturbance is detected again, the SOC equalization task is immediately terminated to prioritize voltage stability. The transparent exit process described in step S5 includes: the exiting node broadcasts an exit notification in advance; after receiving the notification, other nodes quickly renegotiate task allocation and distribute the exiting node's tasks to the remaining nodes; the exiting node linearly reduces its power to zero and then disconnects the switch; other nodes synchronously increase their power to take over, ensuring a smooth transition in total output.

[0018] Furthermore, it also includes communication and anomaly handling steps: when communication quality degrades or communication is interrupted, the node automatically switches from coordination mode to independent support mode based on fast response rules; when a node is detected joining, leaving, or failing, the remaining nodes automatically update the neighbor topology and re-negotiate consensus, dynamically adjusting task allocation.

[0019] According to specific embodiments provided by the present invention, the method for coordinating and supporting transient voltage in grid-type mobile energy storage disclosed by the present invention has the following technical effects: The two-layer response mechanism balances speed and optimization: the fast response layer is completely local and autonomous, with no communication dependency, and starts responding in 50-100ms and reaches the reactive power output target in 100-200ms, effectively suppressing voltage drops in the initial transient phase; the consistency optimization layer achieves multi-objective optimization through distributed negotiation, improving energy storage capacity utilization efficiency by about 65% and solving the voltage oscillation problem caused by a single fast response. The decentralized distributed architecture is adapted to the characteristics of mobile energy storage: there is no central controller, each node is a peer, and global negotiation is completed through a consensus algorithm to avoid the risk of single point of failure. It supports dynamic addition / exit of devices and is adapted to scenarios where the deployment location and number of mobile energy storage devices change dynamically. Plug-and-play self-organizing networks enable rapid deployment: simplified self-testing, relaxed grid connection tolerance, and standardized message interaction enable rapid access in 5-10 seconds, which is 3-6 times faster than traditional manual configuration methods, meeting the rapid deployment needs in emergency fault scenarios. Adaptive adjustment of control parameters enhances transient adaptability: Based on the relative position index with the fault point, the voltage loop bandwidth and droop coefficient are dynamically adjusted to match the transient response characteristics of each node with the electrical coupling strength, thereby enhancing the stability of nodes near the fault point and improving the response speed of nodes far from the fault point. SOC-driven distributed power redistribution extends support time: Under the premise of stable voltage, global SOC balance is achieved without centralized calculation. Power transfer is completed through negotiation between neighbors, which gradually reduces the standard deviation of SOC distribution and extends the overall support time by 20%-30%. A robust anomaly handling mechanism enhances system robustness: Strategies are designed to handle abnormal scenarios such as communication degradation, node failure, and algorithm divergence. When communication is interrupted, the system automatically switches to independent support mode. When a node exits, task takeover is completed within 3-5 seconds. SOC balancing is terminated immediately during voltage disturbances, ensuring stable operation of the system under complex conditions. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a diagram of the two-layer response mechanism architecture of the present invention, illustrating the hierarchical relationship and collaborative workflow of the fast response layer, communication network layer, and consistency optimization layer; Figure 2 This is a time-coordination diagram of the fast response and consistency optimization of the present invention, showing the time variation of fast response output, optimized output, weighted final output and switching weights; Figure 3 This is a flowchart of the steady-state measurement and multi-factor weight calculation process of the present invention, showing the steps of steady-state determination, measurement value acquisition, calculation of each factor, and generation of comprehensive weight. Figure 4 This is a timing diagram for the plug-and-play fast access of the present invention, showing the complete process of fast access in 5 to 10 seconds; Figure 5 This is a flowchart of the SOC drive power redistribution process of the present invention, illustrating the dynamic process of SOC balancing; Figure 6 This is a comparison chart of voltage recovery effects in embodiments of the present invention, showing the voltage recovery curves and key performance indicators for three scenarios: no support, fast response only, and the two-layer collaboration of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Terminology Explanation: Grid-based mobile energy storage: Vehicle-mounted or containerized energy storage systems that adopt grid-based control strategies can actively establish voltage and frequency references and have the ability to be rapidly deployed and relocated.

[0024] Two-layer response mechanism: A hybrid control strategy combining fast local response (millisecond level) and distributed coordination optimization (second level) is adopted. The fast layer ensures immediate support in the early stage of transient response, while the optimization layer realizes coordination and cooperation in the steady-state stage.

[0025] Distributed consensus negotiation: Each mobile energy storage unit acts as a peer node, autonomously negotiating task allocation through neighbor communication and consensus algorithms, without the need for a central controller for coordination.

[0026] Plug and play self-organizing network: After mobile energy storage is connected, it automatically discovers neighbors, establishes communication, negotiates parameters and puts it into operation through standardized protocols, supporting dynamic addition and removal of devices.

[0027] This invention aims to provide a method for coordinating and supporting transient voltage in a grid-based mobile energy storage system. It employs a two-layer mechanism combining fast response and distributed optimization to achieve collaborative support for multiple mobile energy storage units. The main objectives are: to establish a decentralized distributed coordination mechanism; to design a two-layer control strategy combining fast local response and coordinated optimization; to propose a task-autonomous negotiation algorithm based on steady-state measurement; to achieve plug-and-play functionality with 5-10 second fast access; and to provide a SOC-driven distributed power redistribution mechanism.

[0028] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] like Figures 1-6 As shown, this invention provides a method for coordinating and supporting transient voltage in grid-type mobile energy storage, employing a two-layer response mechanism combining fast response and consistency optimization, including the following steps: S1, Initialization and Standby: After the grid-type mobile energy storage node is connected to the grid, it completes self-test grid connection, neighbor discovery, coordination group joining and parameter initialization in sequence to achieve plug-and-play and enter standby state; S2, Fault Detection and Rapid Response: Each grid-type mobile energy storage node independently monitors the voltage. When a transient fault is detected, a rapid response is triggered. The reactive power output is directly calculated based on the local voltage deviation, and the reference voltage setpoint of the outer voltage loop is corrected based on the voltage-reactive power droop characteristics of the grid-type control. S3, Consistency Negotiation and Optimization: After the fast response is initiated, consistency optimization is started in parallel. After the transient process enters a quasi-steady state, the steady-state voltage deviation and voltage recovery rate are measured. Multi-factor comprehensive local weights are calculated. The sum of weights and the total reactive power demand are negotiated through distributed consistency iteration. The optimization target power is calculated, and the voltage control parameters are adaptively adjusted. Finally, the reactive power output is gradually transitioned from the fast response output to the consistency optimization output through smooth switching. Step S3, Consistency Negotiation and Optimization, specifically includes: S301, Consistency Optimization Startup and Steady-State Measurement: After the fast response startup, consistency optimization is started in parallel. The system enters a quasi-steady state after the transient process ends by using a sliding window. S302, Multi-factor comprehensive weight and total reactive power demand negotiation: Each network-type mobile energy storage node calculates a comprehensive local weight, and iteratively negotiates with its neighbors through a distributed average consensus algorithm to achieve global consistency between the normalized weight and the total reactive power demand of the system. S303, adaptive adjustment of control parameters: Each grid-type mobile energy storage node calculates its relative position index based on the global maximum voltage deviation, and adaptively adjusts the voltage loop bandwidth and voltage droop coefficient according to the index. S304, Smooth Switching Between Fast Response and Optimized Output: Employs an exponential decay weighted strategy to gradually switch reactive power output from fast response output to consistent optimized output.

[0030] S4, Steady-state coordination and SOC balance: The system enters steady-state coordination operation, monitors the SOC distribution in real time, and when the SOC deviation of a certain node meets the triggering condition, it performs distributed power redistribution to achieve global SOC balance. S5, Support Exit and Equipment Recovery: After the grid voltage is detected to have stabilized, each grid-type mobile energy storage node synchronously reduces reactive power output through consensus negotiation, smoothly exits support and restores the baseline control parameters, returns to standby state, and continues to monitor the grid; if a node needs to exit, a transparent exit process is executed to complete the task transfer.

[0031] Specifically, this invention proposes a two-layer response mechanism for the transient voltage coordination support of grid-type mobile energy storage. The fast response layer ensures support in the initial transient phase, while the consistency optimization layer achieves steady-state coordination.

[0032] 1. Overall Architecture of Two-Layer Response Mechanism This invention employs a two-layer mechanism combining a fast response layer and a consistency optimization layer to resolve the conflict between transient response speed and distributed coordination optimization. For example... Figure 1 As shown, the bottom layer is the fast response layer (multiple mobile energy storage nodes, each independently detecting and making decisions, without communication connections), the middle layer is the communication network (mesh connections between nodes), and the top layer is the consistency optimization layer (visualization of the iterative negotiation process, such as convergence curves). The startup sequence of the two layers is represented by a time axis: the fast response layer starts from 0 to 200ms, and the consistency optimization layer starts after 200ms and continues to iterate. The outputs of the two layers are distinguished by color: fast response output (red), optimization output (blue), and smooth switching process (red gradually changing to blue). Key parameters are marked: fast response time less than 100ms, consistency convergence time 2 to 5 seconds, and switching time constant 4 seconds. Specifically, each grid-type mobile energy storage node integrates a grid-type converter, battery management system, local detection module, and local control module, enabling independent fault detection and local fast reactive power control. The communication network layer realizes state synchronization and message interaction between each grid-type mobile energy storage unit. The local control module of each grid-type mobile energy storage node integrates the control logic of the fast response layer and the consistency optimization layer, achieving layered control of transient fast response and steady-state coordinated optimization.

[0033] 1.1 Fast Response Layer Design The fast response layer starts immediately after fault detection and quickly outputs reactive power based on local measurements and preset rules, with a response time of less than 100ms.

[0034] Local fault detection: Each mobile energy storage unit independently monitors its local voltage, with a sampling frequency of no less than 1kHz. A fast response is triggered when the detected voltage deviation exceeds 5% of the rated value or the voltage change rate exceeds 0.05 pu / s (per unit). The detection delay is controlled within 20 to 50 ms.

[0035] Fast response rule: Reactive power output is calculated directly based on local voltage deviation, without the need for communication and negotiation with other nodes. Fast response reactive power calculation formula:

[0036] in, The fast response gain factor (recommended value 10, which can be adjusted in the range of 8 to 12 depending on the grid strength) is used. This represents the local voltage deviation (per unit). Rated power, This is the SOC correction factor.

[0037] The SOC correction factor is set according to the current SOC segment: when SOC is above 70%, the factor is 1.1, which is used preferentially; when SOC is between 40% and 70%, the factor is 1.0, which is used normally; when SOC is between 20% and 40%, the factor is 0.7, which is moderately reduced; when SOC is below 20%, the factor is 0.4, which is significantly reduced to protect the battery.

[0038] The fast response output takes capacity limitations into account, and the actual output is the smaller of the calculated value and 90% of the maximum capacity (safety margin).

[0039] Fast response execution: After calculation, based on the voltage-reactive power droop characteristic (QV Droop) of the grid-type control, the reference voltage setpoint of the outer voltage loop is corrected. This is achieved by changing the no-load voltage setpoint. This allows the actual output reactive power to be tracked and calculated. The maximum setpoint adjustment range is limited to ±5% of the rated voltage. Voltage adjustment is completed within 50 to 100 ms, and reactive power output rise time is 100 to 200 ms.

[0040] The fast response layer is characterized by: complete local autonomy with no communication dependencies; fast response speed to meet the initial transient requirements; simple rules and low computational load; and strong robustness, unaffected by communication failures. The fast response layer continues to run for 5 to 10 seconds after fault detection, providing time for the consistency optimization layer.

[0041] 1.2 Consistency Optimization Layer Design The consistency optimization layer starts after fault detection and optimizes power allocation through distributed algorithm negotiation, with a convergence time of 2 to 5 seconds. The optimization layer mainly plays its role after the transient process stabilizes (500ms to 1 second after the fault), improving the steady-state support effect.

[0042] Optimization goals: While meeting voltage support requirements, optimize multiple objectives such as electrical coupling strength, capacity utilization balance, and SOC balance to improve overall support efficiency and duration.

[0043] Optimization method: A distributed average consensus algorithm is adopted, in which each node achieves consensus on weight allocation through iterative communication. The consensus algorithm is executed in parallel after the fast response layer starts, and after convergence, the fast response output is gradually switched to the optimized output.

[0044] Inter-layer coordination: The outputs of the fast response layer and the optimization layer are switched smoothly through weighted averaging.

[0045] in, For nodes The final reactive power output, For nodes Fast-response reactive power output calculated directly based on local voltage deviation For nodes The optimized target reactive power is obtained through consensus negotiation. To switch weights; Switching weights As time decreases from 1 to 0, an exponential decay function is used:

[0046] in, The time constant is switched at the steady-state determination time. Take 4 seconds. Smooth switching avoids sudden changes in output power, ensuring a smooth transition.

[0047] 1.3 Collaborative Workflow of the Two-Tier Mechanism At time T0 (0 to 50ms after the fault occurs): Each node detects an abnormal voltage and triggers the fast response layer.

[0048] At time T1 (50 to 150 ms): the fast response layer calculates the reactive power output and begins execution, while the consistency optimization layer initiates negotiation.

[0049] At time T2 (150 to 500 ms): The fast response output rises to the target value, initially suppressing voltage drops. The consensus algorithm begins iterating.

[0050] At time T3 (500ms to 5 seconds): the transient process stabilizes and enters a quasi-steady state. The consensus algorithm continues to iterate and converges to obtain the optimized weights. Each node gradually switches from fast response to optimized output.

[0051] At time T4 (5 seconds later): Fully switch to optimized output, fast response layer exits. The system enters steady-state coordinated operation.

[0052] At time T5 (after fault clearing): the voltage is detected to have stabilized, the transient support is exited and the system returns to standby mode.

[0053] like Figure 2 As shown, the horizontal axis represents time (0 to 10 seconds), and the vertical axis represents reactive power output (0 to maximum value). Two sets of curves are plotted: the fast response output curve (rapidly rising to a high value, then smoothly declining), and the consistency optimization output curve (slowly rising from 0 to a steady-state value). The final output is represented by a superimposed curve (a weighted combination of the two). Key time points are marked: T1 (50ms, fast response starts), T2 (200ms, fast response reaches its target, consistency starts), T3 (2.5 seconds, consistency converges), and T4 (6 seconds, switching complete). The curve showing the change in switching weights is marked with a dashed line. The transient critical period (0 to 500ms) is marked with a shaded area to emphasize the important role of fast response during this period.

[0054] 2. Consensus negotiation based on steady-state measurement The core of the consistency optimization layer is task negotiation, which uses steady-state measurement to avoid time synchronization problems.

[0055] 2.1 Steady-state voltage deviation measurement To avoid time synchronization issues in transient measurements, steady-state voltage deviation is used as the basis for weighting calculations. Steady-state measurements are initiated 500ms to 1 second after a fault, waiting for the voltage fluctuations to initially subside.

[0056] Steady-state judgment criterion: A sliding window with a width of 50ms (i.e., 2.5 power frequency cycles) is used for judgment. When the voltage change rate within two consecutive sliding windows ( When the absolute values ​​of all values ​​are less than 0.002 pu / s, the transient process is considered to have ended, and the system enters a "quasi-steady state." At this time, the current voltage measurement value is latched as the "steady-state voltage deviation," which is the difference between the voltage before the fault (the sliding average value taken 100 ms before the fault) and the currently latched steady-state voltage. Steady-state measurement avoids the problem of asynchronous measurement times at different nodes during transient processes; all nodes measure the voltage of the "same state."

[0057] 2.2 Multi-factor comprehensive weight estimation Local weighting considers not only voltage deviation, but also multiple factors such as voltage recovery speed, capacity, and SOC.

[0058] Voltage deviation factor: Calculated based on steady-state voltage deviation. The larger the deviation, the larger the factor. A small constant (0.01) is added to the denominator during calculation to prevent division by zero.

[0059] Voltage recovery rate factor: This measures the rate at which the voltage recovers from its lowest point to a steady state. A slow recovery indicates weak support capability. The recovery rate is defined as the voltage change divided by the time interval. The reciprocal of the recovery rate is used as a weighting factor (slower recovery has a higher weight), and after normalization, the recovery rate factor is obtained.

[0060] Capacity factor: The ratio of a node's rated power to its average rated power. Nodes with larger capacity can handle more tasks.

[0061] SOC factor: Segmented settings: when SOC is above 70%, the factor is 1.15, which is given priority; when SOC is between 40% and 70%, the factor is 1.0, which is used normally; when SOC is between 20% and 40%, the factor is 0.7, which reduces the workload; when SOC is below 20%, the factor is 0.5, which significantly reduces the load to protect the battery.

[0062] Comprehensive local weighting: The factors are weighted using a power product approach to enhance node differentiation. The calculation formula is as follows:

[0063] in, The normalized voltage deviation factor, The normalized recovery rate factor, As capacity factor, This represents the SOC factor. The exponential parameters (0.4, 0.2, 0.2, 0.2) reflect the dominance of voltage deviation.

[0064] like Figure 3 As shown, the left side is the time axis, marked with fault occurrence (0ms), voltage minimum point (50 to 100ms), steady-state judgment (500ms), and steady-state measurement (500 to 600ms). The middle section shows the measurement values ​​acquired: steady-state voltage deviation, voltage recovery rate, capacity, and SOC. The right side shows the weight calculation: calculation of each factor (voltage deviation factor, recovery rate factor, capacity factor, SOC factor) → weighted combination → local weight. Different colors are used to distinguish the contribution ratio of each factor (pie chart).

[0065] 2.3 Iterative Negotiation of Consensus Algorithm Each node negotiates the normalized weights and total reactive power demand through a consensus algorithm.

[0066] Basic iterative formula: node i In the k+ In the first iteration, update its estimated sum of weights:

[0067] in, For nodes The neighborhood group, For step size parameters, Represents a node In the The sum of weights estimated in the next iteration. Represents a node neighboring nodes In the The sum of weights estimated in the next iteration. Represents a node In the The sum of the estimated weights updated in the next iteration; An adaptive step size strategy is adopted: a large step size of 0.3 is used for fast convergence in the initial stage (first 5 iterations), a medium step size of 0.15 is used for stable convergence in the middle stage (6 to 15 iterations), and a small step size of 0.08 is used for accurate convergence in the later stage (after 15 iterations). The adaptive step size speeds up the convergence speed in the early stage, while the small step size ensures accuracy in the later stage.

[0068] Initial value setting: The initial estimated total weight is the local weight multiplied by the number of neighbors plus 1 (including itself). This initial value utilizes the neighbor count information, which is closer to the true value than a fixed value, reducing the number of iterations.

[0069] Convergence criterion: Convergence is determined when the change in value is less than 1% of the current value after three consecutive iterations; or when the maximum number of iterations (25) is reached, the process is forcibly terminated. Simulations show that 25 iterations are sufficient to reduce the error to within 5%.

[0070] Normalized weight calculation: After convergence, each node obtains a consistent total weight. The normalized weight is obtained by dividing the local weight by the total weight, ensuring that the sum of all weights is 1.

[0071] 2.4 Negotiation of Total Reactive Power Demand A similar method is used to negotiate the total reactive power demand. Each node estimates its local demand based on the local steady-state voltage deviation and the estimated voltage-reactive power sensitivity. The sensitivity estimate is selected based on the grid strength: for typical weakly connected distribution network end scenarios with a short-circuit ratio (SCR) between 2.0 and 3.0, a sensitivity coefficient of 0.04 to 0.06 pu / Mvar is recommended; for extremely weak grids or medium- and low-voltage microgrids with an SCR less than 2.0, a sensitivity coefficient of 0.05 to 0.08 pu / Mvar is recommended. If access point information is provided at the time of access, a more accurate value can be obtained by querying the database.

[0072] The average of the total demand is negotiated through a consensus algorithm. This averaging method incorporates local measurements from each node, making it more robust than single-point measurements. The iterative process is the same as the weight negotiation and can be executed in parallel.

[0073] 2.5 Target Power Calculation After negotiation, each node calculates the optimized target reactive power, which is the normalized weight multiplied by the negotiated total demand. Considering capacity constraints, the actual target is taken as the smaller of the calculated value and 95% of the maximum capacity, with a limit factor of 0.95 to ensure no overload. The optimized target power gradually replaces the fast response output through an inter-layer switching mechanism.

[0074] 3. Adaptive adjustment of voltage control parameters Based on the relative position of the nodes and the supporting tasks, the voltage control parameters are adaptively adjusted to optimize transient response characteristics.

[0075] 3.1 Calculation of Relative Position Indicators The relative position index is calculated based on the steady-state voltage deviation. Through consensus negotiation, each node obtains the global maximum voltage deviation, and then the ratio of the local deviation to the maximum deviation is calculated as the relative position index. An index close to 1 indicates that the node has a large voltage deviation, is close to the fault point, or has strong electrical coupling; an index close to 0 indicates that the node is far away or has weak coupling.

[0076] 3.2 Voltage loop bandwidth adjustment The bandwidth of the voltage outer loop controller affects voltage response speed and stability. Nodes closer to the fault require stronger stability and should have their bandwidth appropriately reduced; nodes farther away can have their bandwidth increased to speed up the response.

[0077] The parameters of the voltage outer-loop PI controller are adjusted based on the relative position index. Both the proportional and integral coefficients are calculated by multiplying the reference value by an adjustment factor, which is obtained by subtracting 0.3 to 0.4 times the relative position index from 1. When the relative position index is large (closer distance), the PI parameters decrease, the bandwidth decreases, and stability is enhanced; when the index is small (greater distance), the PI parameters approach the reference value, maintaining normal response. The adjustment factor ensures that the parameter variation range is within a reasonable range (±30% to 40%).

[0078] 3.3 Voltage droop coefficient adjustment Network-based control utilizes voltage-reactive power droop characteristics. The droop coefficient affects voltage regulation capability; a smaller droop coefficient means a greater voltage regulation effect for the same reactive power output.

[0079] The droop factor is adjusted based on the relative position index. The baseline droop factor (typically 3% to 5%) is divided by 1 and then increased by 0.8 times the relative position index. A larger relative position index (closer distance) results in a smaller droop factor and enhanced voltage regulation capability; a smaller index (greater distance) results in a droop factor closer to the baseline value. An adjustment factor of 0.8 ensures the droop factor remains within 55% to 100% of the baseline value, preventing excessive differences that could lead to power distribution imbalance.

[0080] 3.4 Timing of Parameter Adjustment Parameter tuning is performed after the consensus algorithm converges, synchronously with power switching. The tuning process is smooth with a time constant of 3 to 5 seconds to avoid system oscillations caused by sudden parameter changes. After tuning, the parameters remain unchanged during the support period. After the fault is cleared, the parameters are gradually restored to the baseline value with a recovery time constant of 10 to 15 seconds.

[0081] 4. Plug and play self-organizing network protocol The plug-and-play protocol supports mobile energy storage with fast access in 5 to 10 seconds and transparent exit.

[0082] 4.1 Quick Access Process (5 to 10 seconds) T0 to T3 Stage (0 to 3 seconds) - Simplified Self-Test and Grid Connection: Performs core safety checks, including battery voltage (SOC greater than 15%), insulation resistance status confirmation (reading the real-time insulation monitoring register value of the Battery Management System (BMS), or based on historical data from the most recent shutdown self-test; data reading and judgment time is less than 200ms, avoiding time-consuming high-voltage injection insulation tests), and protection device function (response test). Skips non-critical items such as battery equalization and detailed internal resistance tests. Detects the connection point voltage and frequency (85% to 115% of rated value, 49.5 to 50.5Hz for 50Hz systems; 59.3 to 60.7Hz for 60Hz systems) to confirm grid connection capability. Performs fast soft grid connection, with tolerance relaxed to 5% amplitude and 10 degrees phase, closing the switch; grid connection inrush current is limited to within 15% of rated current. Total self-test and grid connection time is 2 to 3 seconds.

[0083] Phase T3 to T5 (3 to 5 seconds) - Neighbor Discovery: Immediately after network connection, a HELLO message is broadcast, including node ID, capacity, SOC, location, and communication parameters. A 1-second response timeout is set to collect neighbor replies and build a neighbor table. If there is no response within 1 second, the node is determined to be an isolated node and enters independent operation mode. Neighbor discovery takes 1 to 2 seconds.

[0084] Phase T5 to T7 (5 to 7 seconds) - Joining Coordination: Send a JOIN_REQUEST message; upon receiving a JOIN_RESPONSE message, obtain the current coordination status. Initialize the fast response layer parameters (using default values) and consistency variables (using current neighbor values ​​or capacity-based estimates). If the current power grid is in a fault state, immediately activate the fast response layer; otherwise, enter standby. Joining coordination takes 1 to 2 seconds.

[0085] Phase T7 to T10 (7 to 10 seconds) - Participation in Coordination: Begin sending STATE_SYNC messages and participate in consistency iterations. If a fault is being supported, gradually switch to optimized output after several iterations. If the power grid is normal, remain on standby, preparing to deal with future disturbances. The initial participation in coordination takes 2 to 3 seconds.

[0086] The total time for quick access is 5 to 10 seconds, which is 3 to 6 times faster than the traditional method that requires manual configuration (30 to 60 seconds).

[0087] like Figure 4As shown, the horizontal axis represents time (0 to 10 seconds), and the vertical axis represents participating entities (new node, neighbor A, neighbor B, power grid). Arrows represent message passing and actions, and message types and times are labeled. Key steps: T=0 Power on → T=2 Self-test complete → T=3 Grid connection successful, broadcast HELLO → T=3.5 Receive response → T=4 Send JOIN_REQUEST → T=4.5 Receive JOIN_RESPONSE → T=5 Initialize parameters → T=7 Participate in consensus iteration → T=10 Enter coordination operation. Different colors are used to distinguish stages: self-test grid connection (red), neighbor discovery (yellow), joining coordination (green), and participating in operation (blue). The time taken for each stage and the total cumulative time are labeled.

[0088] 4.2 Standard Access Procedure (15 to 30 seconds) Standard access performs a complete self-test and stable grid connection, suitable for non-emergency scenarios. The complete self-test includes battery balancing testing, internal resistance measurement, converter performance testing, and sensor calibration, taking 8 to 10 seconds. Soft grid connection uses a high-precision algorithm with tolerances controlled within 2% amplitude and 5 degrees phase, taking 2 to 3 seconds. Neighbor discovery waiting time is extended to 3 to 5 seconds to ensure all reachable neighbors are discovered. Parameter initialization is more refined, and local testing can be performed to verify access point characteristics. The total time for standard access is 15 to 30 seconds, ensuring access quality and system stability.

[0089] 4.3 Transparent Exit Process (3 to 5 seconds) Exit notification (T-3 seconds): The node broadcasts a LEAVE_NOTIFY message, notifying its neighbors 3 seconds in advance, explaining the reason for exiting and the current workload.

[0090] Task transfer negotiation (T-3 to T-1 seconds): After receiving the notification, other nodes quickly renegotiate task allocation. A large step size (0.4) is used to accelerate convergence, completing in 5 to 10 iterations (taking 1 to 2 seconds). The tasks of the exiting nodes are distributed to the remaining nodes.

[0091] Power descent and synchronous takeover (T-1 to T0 seconds): The exiting node linearly reduces power to zero over a 1-second period. Other nodes synchronously increase power to take over, ensuring a smooth transition in total output and keeping voltage fluctuations within 2%.

[0092] Decoupling and Topology Update (T0 seconds): After the power drops to zero, the switch is disconnected, and an exit confirmation is sent. Other nodes remove the node from their neighbor tables, update the topology, and continue coordination.

[0093] Transparent exit ensures that the exit of a single node does not affect the overall support, and the remaining nodes are automatically adjusted, demonstrating the robustness of the distributed architecture.

[0094] 4.4 Handling of Isolated Nodes Orphaned nodes operate in an independent support mode, based on fast response layer rules. A conservative fast response gain coefficient (8, lower than the 10 in coordinated mode) is used to avoid over-response. Output power is limited to 70% of rated capacity, leaving a margin. Orphaned nodes attempt to discover neighbors every 30 seconds, and immediately initiate the join process upon discovery. Orphaned mode performance is slightly lower than coordinated mode, but basic support functions are guaranteed.

[0095] 5. SOC-driven distributed power redistribution Over a long period of time, differences in SOC gradually become apparent, requiring dynamic adjustment of power allocation to achieve balance.

[0096] 5.1 SOC Status Sharing and Monitoring Each node includes its current State of Charge (SOC) in the STATE_SYNC message, which is shared with its neighbors. Nodes maintain a list of neighbor SOCs and calculate the average neighbor SOC, which is the sum of the node's SOC and all neighbor SOCs divided by the total number of nodes. The SOC deviation is calculated as the node's SOC minus the average neighbor SOC. When the SOC deviation is less than -15% (SOC significantly lower than the average), a power adjustment request is triggered.

[0097] 5.2 Constrained Power Adjustment Request Before initiating power adjustment, check the voltage stability margin constraints: Voltage margin check: If the current voltage deviation is greater than 5%, SOC equalization adjustment is prohibited; voltage stability should be prioritized. If the voltage deviation is less than 3%, SOC equalization is permitted.

[0098] Critical Node Protection: If a node's weight is greater than 0.35 (critical node), the SOC balance adjustment is limited to within 20% to avoid excessively weakening the supporting role of critical nodes. If the weight is less than 0.2 (edge ​​node), a large adjustment (±50%) is allowed.

[0099] Compensation Capability Verification: Calculate the power adjustment amount by multiplying a negative adjustment factor (typically 0.3) by the SOC deviation and then by the current target power. Send a POWER_ADJUST message to neighbors to inquire whether they can share the adjustment amount. Neighbors assess their own SOC (greater than 40%) and load factor (less than 80%) and reply with the amount they can share. Summarize neighbor responses; if the total sharing amount is less than 80% of the adjustment amount, reject the adjustment request and maintain the current output.

[0100] 5.3 Calculation and Execution of Shared Responsibility The share of the workload for each response node is proportional to its SOC margin. The SOC margin is defined as the current SOC minus the minimum allowed SOC (15%). The share is equal to the node's SOC margin divided by the sum of the SOC margins of all response nodes, and then multiplied by the total adjustment amount.

[0101] The requesting node reduces its output by the adjustment amount, and each responding node increases its output by its share. The adjustment process is completed smoothly in 5 to 10 seconds. A voltage-priority circuit breaker mechanism is established: During SOC balancing adjustment, if a sudden change in the absolute value of the local voltage deviation exceeds 1.5% or the voltage change rate exceeds 0.01 pu / s, it is determined that the power grid has been disturbed again, and the SOC balancing task is immediately and forcibly terminated. All nodes return to the voltage support output state of the previous moment within 100 ms, prioritizing voltage stability.

[0102] 5.4 Emergent Effect of SOC Equalization Although only negotiating with neighbors, through multiple rounds of adjustments (intervals of 5 to 10 minutes, depending on support strength), a global SOC equilibrium naturally emerges in the system. Nodes with low SOC request to reduce power, while nodes with high SOC respond and increase power; this mechanism gradually makes the SOC converge. Under ideal conditions (connected communication topology), the standard deviation of the SOC distribution gradually decreases, eventually converging to a weighted uniform distribution. Simulations show that distributed SOC equilibrium can extend support time by 20% to 30%.

[0103] like Figure 5 As shown, the left side is the initial state histogram (SOC and power output of each energy storage unit), the middle side is the decision-making process (SOC deviation calculation → voltage margin check → power adjustment request → neighbor response → load sharing calculation), and the right side is the adjusted state histogram (more balanced SOC distribution, power redistribution). Arrows indicate the direction and amount of power transfer (from nodes with low SOC to nodes with high SOC). The bottom shows the SOC standard deviation over time (gradually decreasing) to illustrate the balancing effect. Constraints are indicated: adjustment is allowed only if the voltage margin is less than 3%, and the adjustment range for critical nodes is ≤20%.

[0104] 6. Communication and Exception Handling 6.1 Communication Methods and Requirements Primary communication method: Deploy a WiFi router on-site to form a local area network. WiFi latency is 10 to 50 ms, packet loss rate is 1% to 3%, meeting the consistency algorithm requirements (latency less than 100 ms, packet loss rate less than 5%). Communication cycle is 100 to 200 ms (STATE_SYNC message), bandwidth requirement is approximately 5 to 10 kbps per node (message size approximately 100 to 200 bytes).

[0105] Alternative communication methods: 4G / 5G cellular networks, suitable for scenarios where WiFi coverage is unavailable. 4G latency is 50 to 200ms, with a packet loss rate of 5% to 15%. When using 4G, the communication cycle is extended to 200 to 500ms, the consensus algorithm step size is reduced (to 0.1), and the convergence time is extended to 5 to 10 seconds.

[0106] Communication quality testing: During the plug-and-play phase, new nodes test communication latency and packet loss rate with their neighbors. If the latency exceeds 200ms or the packet loss rate exceeds 10%, the node refuses to join the coordination group and switches to independent operation mode. During operation, communication quality is checked every 30 seconds, and an alarm is issued when the quality deteriorates.

[0107] 6.2 Degradation Operation Strategy Complete Communication Interruption: If a node does not receive any messages from its neighbors for five consecutive communication cycles (0.5 to 1 second), a communication failure is determined. It immediately switches to standalone operation mode, employing fast response layer rules, limiting reactive power output to 70% of rated capacity. It periodically attempts to rebuild communication (broadcasting "HELLO" every 30 seconds). After communication is restored, it smoothly switches back to coordination mode, with a recovery time of 5 to 10 seconds.

[0108] Partial communication interruption: If only some neighbors are lost, remove the lost nodes from the neighbor table and continue the consistency iteration using the remaining neighbors. The algorithm may temporarily fail to converge after a topology change, but will converge again after additional iterations.

[0109] Communication congestion: If a sudden increase in packet loss rate (greater than 20%) is detected, the communication frequency is automatically reduced, and the STATE_SYNC period is extended from 100ms to 500ms to alleviate network load. Critical messages (JOIN, LEAVE) employ a retransmission mechanism to ensure reliable delivery.

[0110] 6.3 Node Fault Handling Local fault rapid exit: When a node detects a local fault (battery overheating above 60°C, inverter fault, SOC less than 15%), it immediately initiates protection, reducing power to zero within 1 to 2 seconds. Simultaneously, it sends a LEAVE_NOTIFY message (if communication is normal), and other nodes quickly negotiate to take over the task (within 1 to 2 seconds).

[0111] Neighbor Failure Detection: If a neighbor suddenly loses contact and there is no LEAVE_NOTIFY message, it is determined to be a failure rather than a normal exit. Other nodes immediately initiate rapid negotiation (large step size 0.5), and complete the task redistribution in 5 to 10 iterations (1 to 2 seconds) to ensure support continuity.

[0112] Multi-node fault alarm: If multiple nodes fail within 10 seconds (more than 30% of the total), the remaining nodes will determine whether it is a systemic problem. If the remaining capacity is insufficient to maintain voltage stability (total capacity less than 70% of demand), an alarm will be sent to the superior dispatch system to request reinforcements or prepare to cut off the load.

[0113] 6.4 Handling Algorithm Convergence Anomalies Divergence detection: If the change in a variable is still greater than 10% after 15 consecutive iterations, it is judged as divergence or oscillation. Possible causes: excessive communication delay, disconnected topology, or inappropriate step size parameter.

[0114] Divergence handling: Automatically reduce the step size to 50% of the current value to enhance stability. If divergence still occurs, switch to conservative mode and use fixed weights (based on average capacity distribution, i.e., node rated power divided by the sum of all node rated power), abandoning optimization but ensuring stable system operation.

[0115] Improve topology connectivity: If convergence continues, increase the frequency of HELLO messages (once every 1 second) to attempt to discover more neighbors and improve topology connectivity. If the topology is indeed disconnected (isolated subnets exist), each subnet coordinates independently.

[0116] 7. Complete control process 7.1 Initialization and Standby (0 to 10 seconds) After mobile energy storage deployment and access: Device self-test and soft grid connection (2 to 3 seconds) → Broadcast HELLO to discover neighbors (1 to 2 seconds) → Send JOIN_REQUEST to join coordination (1 to 2 seconds) → Initialize parameters and enter standby (continuous). In standby mode, periodically send HELLO (every 5 seconds) and STATE_SYNC (every 200ms, containing standby status) to maintain connection with neighbors.

[0117] 7.2 Fault Detection and Rapid Response (0 to 200ms) Each node independently detects its local voltage and triggers a rapid response: record the voltage before the fault → calculate the voltage deviation → calculate the reactive power output according to the rapid response formula → adjust the reference voltage amplitude → rapidly increase the reactive power output (rise time 100 to 200 ms). The rapid response layer begins execution within 50 to 150 ms after detection and reaches the initial output within 200 ms.

[0118] 7.3 Consistency Negotiation and Optimization (200ms to 5 seconds) After the fast response is initiated, parallel consistency optimization is initiated: wait for steady state (500ms to 1 second) → measure steady state voltage deviation and recovery speed → calculate local weights of multiple factors → negotiate the sum of weights and total demand in consistency iteration (20 to 25 iterations, 2 to 5 seconds) → calculate optimized power → adjust voltage control parameters → smoothly switch to optimized output (3 to 5 seconds).

[0119] 7.4 Steady-state coordination and SOC equilibrium (5 seconds to several minutes) Enter steady-state coordinated operation: continuously send STATE_SYNC synchronization status (every 200ms) → re-execute consensus negotiation every 1 to 2 seconds, fine-tune weights (considering SOC changes, only 3 to 5 iterations are needed) → monitor SOC distribution and trigger power redistribution (interval 5 to 10 minutes) → monitor voltage recovery and determine whether to exit.

[0120] 7.5 Support Exit and Resume (5 to 15 seconds) When voltage stability recovery is detected (deviation less than 3% and lasting for more than 5 seconds): Exit time through consensus negotiation (fast negotiation, 1 to 2 seconds) → Synchronously reduce reactive power output (power reduction time 5 to 10 seconds) → Voltage control parameters are restored to the reference value (recovery time 10 to 15 seconds) → Return to standby state and continue monitoring the power grid.

[0121] Example 1: Three mobile energy storage units with two-layer response coordination support Scenario: A 35kV distribution network is connected to a 20MW photovoltaic system, with a short-circuit ratio of 2.5. Three mobile energy storage systems are temporarily deployed: Energy Storage A (3MW / 6MWh, SOC 80%), Energy Storage B (2MW / 4MWh, SOC 65%), and Energy Storage C (2.5MW / 5MWh, SOC 50%). The three energy storage systems are networked via WiFi (latency 20-40ms, packet loss rate 2%), forming a fully connected topology.

[0122] Fault event: After 10 minutes of operation, a three-phase short circuit fault occurred in the distribution network, lasting for 200ms. After the fault was cleared, the voltage measured by energy storage A dropped from 1.0 pu to 0.85 pu (a drop of 0.15 pu), energy storage B dropped to 0.72 pu (a drop of 0.28 pu), and energy storage C dropped to 0.78 pu (a drop of 0.22 pu).

[0123] Fast response layer execution (0 to 200ms): Three energy storage units detected voltage anomalies almost simultaneously (detection delay 30 to 50 ms). Each unit calculated its fast-response reactive power output, using a fast-response gain coefficient of 10 and SOC correction coefficients of 1.1 (SOC 80% > 70%), 1.0 (SOC 65% within the 40% to 70% range), and 1.0 (SOC 50% within the 40% to 70% range). Calculation results: Energy storage A outputs 10 × 0.15 × 3 × 1.1 = 4.95 Mvar, but considering capacity limitations (90%), it is limited to 2.7 Mvar; Energy storage B outputs 10 × 0.28 × 2 × 1.0 = 5.6 Mvar, limited to 1.8 Mvar; Energy storage C outputs 10 × 0.22 × 2.5 × 1.0 = 5.5 Mvar, limited to 2.25 Mvar.

[0124] The total fast response output was approximately 6.75 Mvar. The three energy storage systems began increasing reactive power output within 50 to 100 ms after detection, reaching the target value within 150 to 200 ms. At this point, the voltage initially recovered: Energy Storage A recovered to 0.90 pu, Energy Storage B to 0.82 pu, and Energy Storage C to 0.86 pu. The fast response layer effectively suppressed the lowest point of voltage drop.

[0125] Consistency optimization layer execution (200ms to 5 seconds): After rapid response startup, consistency optimization is initiated in parallel. After 500ms, the voltage enters a quasi-steady state. The steady-state voltage deviations for each energy storage unit are as follows: Energy storage A, 0.90 pu (deviation 0.10 pu); Energy storage B, 0.82 pu (deviation 0.18 pu); Energy storage C, 0.86 pu (deviation 0.14 pu). Voltage recovery speeds are measured: Energy storage A, 0.5 pu / s; Energy storage B, 1.0 pu / s; Energy storage C, 0.8 pu / s.

[0126] The local weights of multiple factors are calculated, taking into account four factors: voltage deviation, recovery speed, capacity, and SOC. The comprehensive weight of energy storage A is approximately 1.42 (considering the SOC factor of 1.15), energy storage B is approximately 0.78, and energy storage C is approximately 0.92.

[0127] Consistent iterative negotiation was conducted using an adaptive step size, initially 0.3 and later 0.08. After 18 iterations (approximately 2.5 seconds), the total weights converged to 3.12. Normalized weights: Energy storage A = 0.46, Energy storage B = 0.25, Energy storage C = 0.29.

[0128] Total demand was negotiated, and each energy storage facility estimated its local demand based on local voltage deviation and estimated sensitivity (0.06 pu / Mvar), which were 1.67, 3.0, and 2.33 Mvar, respectively. The average value was 2.33 Mvar.

[0129] Optimized power calculation: Energy storage A target 0.46 × 2.33 = 1.07 Mvar, Energy storage B target 0.25 × 2.33 = 0.58 Mvar, Energy storage C target 0.29 × 2.33 = 0.68 Mvar, total 2.33 Mvar.

[0130] Smooth transition (2.5 to 6 seconds): After the optimization layer converges, each energy storage unit gradually switches from a fast-response output to an optimized output. The switching process takes 3.5 seconds and uses an exponential decay function. Energy storage A smoothly decreases from 2.7 Mvar to 1.07 Mvar, energy storage B from 1.8 Mvar to 0.58 Mvar, and energy storage C from 2.25 Mvar to 0.68 Mvar. During the switching process, the voltage fluctuates slightly (approximately ±0.02 pu), but remains within a safe range.

[0131] Steady-state coordinated operation (6 seconds to 15 minutes): After the switchover is complete, the system enters steady-state coordination. The voltage stabilizes at: 0.96 pu for energy storage A, 0.94 pu for energy storage B, and 0.95 pu for energy storage C. The weights are renegotiated every 2 seconds (requiring only 5 iterations for fine-tuning) and dynamically adjusted according to changes in SOC.

[0132] SOC equilibrium (supported for 10 minutes): Ten minutes later, the SOC of energy storage A dropped to 68%, energy storage B to 55%, and energy storage C to 38%. Energy storage C's SOC was significantly lower than the average (54%), with a deviation of -16%, exceeding the trigger threshold (-15%). Voltage stability margin check: The current voltage deviation is approximately 2%, less than the 3% threshold, allowing for SOC equalization. Energy storage C requested a power reduction of 0.2 Mvar, which was shared by energy storage A and B (based on SOC margin allocation, energy storage A shared 0.13 Mvar, and energy storage B shared 0.07 Mvar). After adjustment, the rate of SOC decrease became more consistent.

[0133] Support withdrawal (15 minutes after failure): Once the grid voltage stabilizes and returns to the range of 0.98 to 1.02 pu for more than 5 seconds, the three energy storage units negotiate to shut down. After a smooth 5-second power reduction, reactive power output drops to zero, voltage control parameters return to the baseline value, and the system returns to standby mode.

[0134] Summary of Results: Figure 6 The horizontal axis represents time (0 to 18 seconds), and the vertical axis represents voltage (0.6 to 1.12 pu). The figure shows three sets of voltage recovery curves: unsupported (dashed line), fast response only (dotted line), and the two-layer synergy of the present invention (solid line). Figure 6 The document lists a comparison of the key performance indicators (maximum drop, recovery time) of the three schemes. Figure 6 The green-filled area visually illustrates the voltage support gain of this invention compared to the unsupported case, while the red markings highlight the oscillation issues present in existing technologies (fast response only). Three key time points—fault clearing (0.2s), consistency convergence (2.5s), and handover completion (6s)—are delineated by vertical dashed lines.

[0135] like Figure 6 As shown, the voltage recovery curves for the three scenarios are compared: 1. Benchmark comparison (no support): voltage drop depth (0.65 pu) and recovery time of more than 10 seconds.

[0136] 2. Existing technical pain points (fast response only): Although the lowest voltage point can be raised to 0.72 pu, during the recovery phase (5 to 8 seconds), due to the lack of target coordination among multiple machines, obvious voltage oscillations and overshoot phenomena occur (as shown in red in the figure), which is not conducive to system stability.

[0137] 3. Effects of the invention (two-layer synergy): It has the same support capability as fast response in the initial transient phase (minimum point 0.72pu); in the recovery phase, through the continuous support gain shown in the green filled area in the figure, and the smooth inter-layer switching mechanism, oscillations are eliminated, and the voltage recovery time is shortened to less than 3 seconds. Figure 6 The quantitative data in the illustration further confirms the advantages of this invention in balancing speed and stability. Furthermore, the optimization layer reduces the total output requirement from 6.75 Mvar to 2.33 Mvar, improving efficiency by approximately 65%.

[0138] Example 2: Degraded Operation in Communication-Constrained Scenarios Scenario: A remote 35kV distribution network with no WiFi coverage, only 4G network with average signal quality (latency 80-150ms, packet loss rate 8%-12%). Deploy two mobile energy storage units: Energy Storage D (4MW / 8MWh, SOC 85%) and Energy Storage E (3MW / 6MWh, SOC 60%).

[0139] Access process: After energy storage devices D and E are connected, a 4G network is attempted. Communication quality testing: average latency 120ms, packet loss rate 10%, critically low (requirements of less than 200ms and less than 10%, which are just met). Degradation parameters are adopted: communication cycle extended to 300ms, consistency step size reduced to 0.1, expected convergence time 5 to 8 seconds.

[0140] Fault Response: A single-phase ground fault occurred in the distribution network, causing a voltage drop. The D and E energy storage layers activated independently (within 50-100ms), outputting 3.2 Mvar and 2.4 Mvar respectively, initially supporting the voltage. The consistency optimization layer activated, but due to large communication delays and packet loss, the iterative process was unstable. After 30 iterations (approximately 9 seconds), it barely converged, but with a significant error (approximately 8%). Optimized power allocation: Energy storage D 2.8 Mvar, Energy storage E 2.0 Mvar.

[0141] Communication interruption: After 5 minutes of operation, the 4G network experienced a sudden congestion, with the packet loss rate surging to 25%. Energy storage devices D and E failed to receive messages from each other for 5 consecutive cycles (1.5 seconds), indicating a communication interruption. The system immediately switched to independent operation mode, operating based on rapid response rules, with reactive power output limited to 70% of rated capacity (energy storage D decreased to 2.8 Mvar, and energy storage E decreased to 2.1 Mvar). While total output decreased slightly in independent mode, basic support was maintained.

[0142] Communication restored: Two minutes later, network congestion eased and communication resumed. Energy storage devices D and E detected each other's HELLO messages and re-established their connection. A smooth switchback to coordination mode was performed, resuming consensus negotiation (switching completed in 5 seconds). After recovery, coordination continued until the fault was resolved.

[0143] Summary of Results: In communication-constrained scenarios, the fast response layer still functions normally (without communication dependencies), ensuring support during the initial transient phase. The consistency optimization layer, limited by communication quality, experiences prolonged convergence time and decreased accuracy, but still provides some optimization (total output reduced by approximately 14%). Automatic degradation to independent mode is implemented upon communication interruption, ensuring no loss of basic functionality. A smooth switch back to coordination mode after communication is restored demonstrates the system's adaptability and robustness. This scenario validates the effectiveness of the degradation strategy and proves the feasibility of the method under adverse communication conditions.

[0144] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for coordinating and supporting transient voltage in a grid-type mobile energy storage system, characterized in that, A two-layer response mechanism combining fast response and consistency optimization is adopted, including the following steps: S1, Initialization and Standby: After the grid-type mobile energy storage node is connected to the grid, it completes self-test grid connection, neighbor discovery, coordination group joining and parameter initialization in sequence to achieve plug-and-play and enter standby state; S2, Fault Detection and Rapid Response: Each grid-type mobile energy storage node independently monitors the voltage. When a transient fault is detected, a rapid response is triggered. The reactive power output is directly calculated based on the local voltage deviation, and the reference voltage setpoint of the outer voltage loop is corrected based on the voltage-reactive power droop characteristics of the grid-type control. S3, Consistency Negotiation and Optimization: After the fast response is started, consistency optimization is started in parallel. After the transient process enters the quasi-steady state, the steady-state voltage deviation and voltage recovery speed are measured, the multi-factor comprehensive local weight is calculated, the total weight and total reactive power demand are negotiated through distributed consistency iteration, the optimization target power is calculated, and the voltage control parameters are adaptively adjusted. Finally, the reactive power output is gradually transitioned from the fast response output to the consistency optimization output through smooth switching. S4, Steady-state coordination and SOC balance: The system enters steady-state coordination operation, monitors the SOC distribution in real time, and when the SOC deviation of a certain node meets the triggering condition, it performs distributed power redistribution to achieve global SOC balance. S5, Support Exit and Equipment Recovery: After the grid voltage is detected to have stabilized, each grid-type mobile energy storage node synchronously reduces reactive power output through consensus negotiation, smoothly exits support and restores the baseline control parameters, returns to standby state, and continues to monitor the grid; if a node needs to exit, a transparent exit process is executed to complete the task transfer.

2. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 1, characterized in that, Step S1, initialization and standby, specifically includes: After the network-type mobile energy storage node is deployed and connected, it sequentially performs device self-test and soft grid connection, broadcasts HELLO messages to discover neighbors, sends JOIN_REQUEST messages to join the coordination group, initializes parameters, completes plug-and-play access, and enters standby mode; in standby mode, it periodically sends HELLO and STATE_SYNC messages to maintain communication connection with neighboring nodes; The device self-test and soft grid connection include: performing core safety checks and reading real-time data from the battery management system, skipping non-critical self-tests; and performing fast soft grid connection after detecting that the voltage and frequency at the access point meet the grid connection conditions.

3. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 1, characterized in that, In step S2, reactive power output is directly calculated based on local voltage deviation. Specifically, it includes: in, For fast response gain coefficient, This is due to local voltage deviation. Rated power, This is the SOC correction factor; Based on the voltage-reactive power droop characteristics of grid-type control, the reference voltage setpoint of the outer voltage loop is corrected, specifically by changing the no-load voltage setpoint. This allows the actual output reactive power to track the calculated reactive power output. .

4. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 1, characterized in that, Step S3, consensus negotiation and optimization, specifically includes: S301, Consistency Optimization Startup and Steady-State Measurement: After the fast response startup, consistency optimization is started in parallel. The system enters a quasi-steady state after the transient process ends by using a sliding window. S302, Multi-factor comprehensive weight and total reactive power demand negotiation: Each network-type mobile energy storage node calculates a comprehensive local weight, and iteratively negotiates with its neighbors through a distributed average consensus algorithm to achieve global consistency between the normalized weight and the total reactive power demand of the system. S303, adaptive adjustment of control parameters: Each grid-type mobile energy storage node calculates its relative position index based on the global maximum voltage deviation, and adaptively adjusts the voltage loop bandwidth and voltage droop coefficient according to the index. S304, Smooth Switching Between Fast Response and Optimized Output: Employs an exponential decay weighted strategy to gradually switch reactive power output from fast response output to consistent optimized output.

5. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 4, characterized in that, Step S301: Determine the end of the transient process by using a sliding window. Specifically, a sliding window with a width of 50ms is used. When the absolute value of the voltage change rate in two consecutive sliding windows is less than 0.002pu / s, the transient process is determined to be over. The difference between the voltage before the fault and the current latched steady-state voltage is calculated as the steady-state voltage deviation.

6. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 4, characterized in that, In step S302, the comprehensive local weight is a power product of multiple factors, including voltage deviation factor, voltage recovery speed factor, capacity factor and SOC factor, wherein the voltage deviation factor plays a dominant role in the comprehensive weight. The basic iterative formula of the distributed average consensus algorithm is as follows: Node i In the k+ In the first iteration, update its estimated sum of weights: in, For nodes The neighborhood group, For step size parameters, Represents a node In the The sum of weights estimated in the next iteration. Represents a node neighboring nodes In the The sum of weights estimated in the next iteration. Represents a node In the The sum of the estimated weights updated in the next iteration; The distributed average consensus algorithm adopts an adaptive step size strategy: a large step size is used for fast approximation in the early stage of iteration, a medium step size is used for stable convergence in the middle stage of iteration, and a small step size is used for precise convergence in the later stage of iteration; the initial estimated weight sum is the local weight multiplied by the number of neighbors plus 1. The convergence criterion is that the change in three consecutive iterations is less than 1% of the current value or the maximum number of iterations is reached. After convergence, each node obtains a consistent total weight. The normalized weight is obtained by dividing the local weight by the total weight, ensuring that the sum of all weights is 1.

7. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 4, characterized in that, In step S303, after the relative position index obtains the global maximum voltage deviation through consensus negotiation, the ratio of the local deviation to the global maximum deviation is calculated. When adjusting the voltage loop bandwidth and voltage droop coefficient according to this index, the proportional coefficient and integral coefficient of the voltage outer loop PI controller are both obtained by multiplying the reference value by the adjustment factor. The adjustment factor is obtained by subtracting 0.3 to 0.4 times the relative position index from 1. The droop coefficient is adjusted according to the relative position index, and the reference droop coefficient is divided by 1 and then added to 0.8 times the relative position index.

8. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 4, characterized in that, Step S304, the formula for smooth switching between fast response and optimized output is as follows: in, For nodes The final reactive power output, For nodes Fast-response reactive power output calculated directly based on local voltage deviation For nodes The optimized target reactive power is obtained through consensus negotiation. To switch weights; Switching weights As time decreases from 1 to 0, an exponential decay function is used: in, For the steady-state determination time, This is the switching time constant.

9. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 1, characterized in that, Step S4, steady-state coordination and SOC equilibrium, specifically includes: Each node shares its current SOC through periodic messages; when the deviation between a node's SOC and the average SOC of its neighbors exceeds a preset threshold and the current voltage stability margin meets the conditions, a power adjustment request is initiated; neighboring nodes assess the amount of load they can share based on their own SOC margin and respond; the requesting node and the responding node jointly complete the power adjustment; during the SOC equalization adjustment, if a voltage disturbance is detected again, the SOC equalization task is immediately terminated to prioritize voltage stability. The transparent exit process described in step S5 includes: the exiting node broadcasts an exit notification in advance; after receiving the notification, other nodes quickly renegotiate task allocation and distribute the exiting node's tasks to the remaining nodes; the exiting node linearly reduces its power to zero and then disconnects the switch; other nodes synchronously increase their power to take over, ensuring a smooth transition in total output.

10. The method for coordinating and supporting transient voltage in grid-type mobile energy storage according to claim 1, characterized in that, It also includes communication and anomaly handling steps: when communication quality degrades or communication is interrupted, nodes automatically switch from coordination mode to independent support mode based on fast response rules; when a node is detected joining, leaving, or failing, the remaining nodes automatically update the neighbor topology and re-negotiate consensus, dynamically adjusting task allocation.

Images