A Coordinated Control Method for Dynamic Microgrid Formation and Priority Activation of Non-Black Start Units

By adopting a coordinated control method that combines dynamic microgrid formation with priority activation of non-black start units, the problems of idle and uncertain non-black start units in the post-disaster recovery of distribution networks are solved, achieving efficient and stable load recovery and distributed power source utilization.

CN123073737APending Publication Date: 2026-11-10TIANJIN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610994175.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-06
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In the post-disaster recovery of distribution networks, the order of line repair is constrained by external conditions on site. The lack of a priority activation mechanism for non-black start units leads to their long-term idleness in the early stages of recovery. Furthermore, existing recovery strategies fail to effectively address the uncertainty of load and distributed power output, resulting in low recovery speed and capacity utilization, as well as the risk of transient power surges.

Method used

A collaborative control method of dynamic microgrid formation and priority activation of non-black start units is adopted. Non-black start units are activated first through a virtual power flow mechanism. Combined with the generation forest constraint and robust safety constraint, a robust mixed integer linear programming model is constructed to ensure the stability and efficiency of load recovery under uncertain environment.

Benefits of technology

It accelerated the growth of the number of activated distributed power sources and the amount of load restored in the early stage of recovery, improved capacity utilization and the adaptability of recovery schemes, reduced the risks caused by prediction errors, and ensured the safety and reliability of the recovery process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123073737A_ABST
    Figure CN123073737A_ABST
Patent Text Reader

Abstract

This invention discloses a collaborative control method for dynamic microgrid formation and priority activation of non-black start units, relating to the field of distribution network restoration control technology. Addressing the issues of non-flexible scheduling of line repair sequence and long-term idleness of high-capacity non-black start units in the early stages of distribution network recovery after disasters, this invention constructs a priority activation model based on the current set of available lines, constrained by activated distributed generation injecting virtual flows into the network and inactive non-black start units receiving at least one unit of virtual flow. Furthermore, it constructs a generating forest constraint based on auxiliary spanning tree indicator variables to divide the distribution network into multiple microgrids rooted at activated distributed generation. Finally, it combines a linearized distribution power flow model to solve for the maximum weighted recovery load, updating the model on a rolling basis according to recovery periods until recovery is complete. This accelerates the activation of distributed generation and load recovery process in the early stages of post-disaster recovery and improves the stability of distributed generation capacity utilization under different fault scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power distribution network restoration control technology, and in particular to a coordinated control method for dynamic microgrid formation and priority activation of non-black start units. Background Technology

[0002] With the widespread integration of distributed generation into active distribution networks, the autonomous recovery capability of distribution networks has been significantly enhanced compared to traditional centralized recovery methods. Distributed generation with black-start capability (i.e., black-start units) can independently form isolated microgrids after a large-scale power outage, thereby driving the gradual recovery of surrounding loads and making a bottom-up power restoration model possible. In situations such as extreme natural disasters, large-scale faults, or malicious cyberattacks, transmission links are often difficult to restore in a short time. In such cases, relying on the resources within the distribution network for on-site restoration is of great significance for maintaining power supply to critical loads and shortening outage time.

[0003] Extensive research has been conducted on black-start recovery, proposing technical approaches such as unit start-up scheduling methods based on model predictive control and grid synchronization methods oriented towards virtual synchronous generator control. These methods also utilize microgrid partitioning to achieve zoned load recovery. While these methods improve the controllability of the recovery process to some extent, most assume that the line repair sequence can be optimized simultaneously with recovery decisions, treating the line repair plan as a continuously schedulable variable during the solution process. However, in actual post-disaster scenarios, the line repair sequence is often not fully schedulable due to constraints from geographical conditions, road accessibility, the accessibility of repair teams, and the extent of facility damage. Recovery decisions can only be made stage-by-stage under an externally given line repair sequence. This results in a significant misalignment between the recovery model and on-site repair execution, making it difficult to implement in engineering projects.

[0004] More critically, existing solutions often center on black-start units in the utilization of distributed generation, while insufficiently considering the role of non-black-start units. Although non-black-start units lack self-starting capabilities, they typically have significantly higher generation capacity than black-start units, serving as crucial backup resources for accelerating load recovery. Once the network topology allows them to connect to the existing electronic grid at a certain stage, they can be activated by external stimuli and immediately output active and reactive power to the system, thereby significantly expanding the boundaries of recoverable loads. However, existing recovery frameworks lack a priority activation mechanism for non-black-start units. The recovery process often passively waits for lines to be repaired gradually in a predetermined order, leaving some high-capacity units that could have been put into operation earlier idle for a considerable period, limiting both the recovery speed and the utilization rate of distributed generation capacity.

[0005] Therefore, it is evident that, under the premise that the order of line repair is constrained by external conditions on site and cannot be freely scheduled, how to fully explore the topological flexibility of the distribution network to guide high-capacity non-black start units to be put into operation as early as possible, thereby significantly improving the speed of post-disaster load recovery and the utilization rate of distributed power supply capacity, is a technical problem that urgently needs to be solved in this field.

[0006] Furthermore, most existing recovery strategies are based on deterministic prediction models, which assume that load demand and distributed generation output are precisely known and deterministic values ​​during the recovery process. However, in actual post-disaster scenarios, the time-varying characteristics of load and the intermittent output of renewable energy sources such as photovoltaics and wind power lead to significant deviations between predicted and actual values. This uncertainty is particularly prominent in the early stages of recovery because the network topology is weak and the available power sources are limited. Once the actual load exceeds expectations or the power output is lower than expected, it can easily trigger voltage overruns or line overloads, forcing the restored loads to be disconnected again, or even causing microgrid instability. Although a few studies have attempted to introduce stochastic programming or scenario-based methods to handle uncertainty, such methods often require a large number of scenario samples, resulting in heavy computational burdens and difficulty in guaranteeing safety in all scenarios. This invention addresses this problem by embedding robust optimization technology into a collaborative framework of dynamic microgrid formation and priority activation of non-black-start units for the first time, providing the recovery scheme with immunity to prediction errors without sacrificing solution efficiency.

[0007] Furthermore, existing technologies typically simplify the activation process of non-black start units to a simple grid-connection switch action, neglecting the transient power surge to the microgrid caused during startup. In practical engineering, large-capacity non-black start units (such as diesel generators or large gas turbines) require several times their rated capacity in surge power from the external grid during the initial startup phase to power auxiliary systems (such as excitation, cooling, and lubrication pumps). If the activated distributed power source does not have sufficient spinning reserve or lacks fast-response energy storage buffers, it will cause a momentary drop in microgrid frequency and a sudden voltage slump, which may lead to startup failure of the non-black start unit or even collapse of the recovered microgrid. While energy storage devices have a fast response speed, their capacity and state of charge are limited. If used blindly and without coordination, they may fail due to energy depletion at critical moments. Summary of the Invention

[0008] The purpose of this invention is to address the problems in the post-disaster recovery of distribution networks, where the order of line repair is constrained by external field conditions and cannot be freely scheduled, and high-capacity non-black start units remain idle for a long time in the early stages of recovery due to the lack of a priority activation mechanism. This invention provides a collaborative control method for dynamic microgrid formation and priority activation of non-black start units, so as to accelerate the growth of the cumulative number of activated distributed power sources and the cumulative restored load in the early stages of recovery, and improve the stability of distributed power source capacity utilization under various fault scenarios.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows.

[0010] A coordinated control method for dynamic microgrid formation and priority activation of non-black-start units, applied to post-disaster recovery of distribution networks containing both black-start and non-black-start units, includes:

[0011] Obtain the repair status of each line in the distribution network during the current recovery period and the location information of activated distributed power sources to determine the set of currently available lines;

[0012] Based on the available line set, and with the constraints of activated distributed power sources injecting virtual flows into the network and non-activated non-black start units receiving at least one unit of virtual flow, a priority activation constraint for non-black start units is constructed, and the set of non-black start units activated in the current time period is solved.

[0013] Using the node where the activated distributed generation is located as the root node, a forest constraint is constructed based on the auxiliary spanning tree indicator variable, so that the distribution network is divided into multiple microgrids with the activated distributed generation as the root and which are not interconnected in the current time period.

[0014] Obtain the active power prediction interval and reactive power prediction interval of each node load, as well as the active power output prediction interval and reactive power output prediction interval of each activated distributed power source, and construct an uncertainty set represented in interval form.

[0015] Obtain the upper limit of ramp rate, minimum continuous operation and downtime of each distributed power source, as well as the instantaneous impact power and start-up impact duration required by each non-black start unit during the start-up process;

[0016] To establish dynamic backup constraints, it is required that currently activated distributed power sources, in addition to meeting the current load recovery needs, reserve an additional spinning reserve capacity of not less than the startup impact power of the non-black start unit to be activated. This reserve capacity must remain available throughout the startup impact duration.

[0017] Construct a time-series charging and discharging coordination constraint for the energy storage device. At the moment of startup of the non-black start unit, control the energy storage device to release power quickly to make up for the difference between the reserve capacity and the actual impact demand. After startup, use the surplus output of the distributed power source to recharge the energy storage device so that its state of charge returns to the safe range.

[0018] Under the generated forest topology, with the goal of maximizing the weighted recovery load, a linearized power flow model is used, and robust security constraints considering the uncertain set are introduced to solve for the recovery state of the load at each node, the active and reactive power output of distributed generation sources, the charging and discharging power of energy storage devices, and the power flow of each line. The robust security constraints ensure that the node voltage amplitude and line power flow do not exceed the limits under any realization within the uncertain set, and the robust constraints are transformed into deterministic linear constraints through linear dual transformation. The dynamic reserve constraints and energy storage timing constraints are solved simultaneously with the robust security constraints to ensure sufficient transient support capability under any uncertain scenario.

[0019] Entering the next recovery period, the set of available lines is updated based on the latest line repair status feedback from the outside, and the above steps are repeated until all non-black start units are activated and all recoverable loads are restored.

[0020] As a preferred technical solution of the present invention, the non-black start unit priority activation constraint aims to maximize the number of non-black start units activated in the current time period. For each inactive non-black start unit, the virtual flow received from the activated distributed power source is characterized by the difference between the virtual inflow and outflow to the node of the non-black start unit. The total amount of virtual flow injected into the network by each activated distributed power source is limited by its corresponding injection variable. The objective function and constraint relationship can be solved jointly with the subsequent model after linearization. The virtual flow constraint and the robust safety constraint share the same set of actual line operating state variables, thereby ensuring the feasibility of the activation path and the feasibility of the recovery scheme under uncertainty. The priority activation ranking result prioritizes non-black start units with relatively small startup impact power and close electrical distance to the activated power source to reduce dynamic backup pressure and achieve pre-matching between the activation decision at the topology level and the transient power support capability.

[0021] As a preferred embodiment of the present invention, the virtual flow is only used as a modeling quantity to characterize connectivity and activation feasibility, and does not correspond to active or reactive power in a physical sense; the virtual flow emitted by any activated distributed power source is not absorbed by other activated distributed power sources, and any virtual flow variable is forced to zero by the large number method constraint when the corresponding line is not repaired or is in a disconnected state, thereby ensuring the effectiveness of the activation path; the solution of the virtual flow constraint is independent of the uncertain interval of load and output, but its result serves as the boundary condition of the available capacity of distributed power sources in the robust safety constraint, realizing the time-series coupling of priority activation and robust recovery; at the same time, the list of units to be activated obtained by solving the virtual flow constraint uses its starting impact power as the input boundary of the dynamic backup constraint, realizing the dual verification of topology reachability and physical startability, and only non-black start units that simultaneously meet the connectivity condition and transient power support condition are allowed to be activated.

[0022] As a preferred technical solution of the present invention, the construction of the generated forest constraint includes: injecting auxiliary traffic equal to the number of nodes minus one into the network through auxiliary root nodes, with each non-auxiliary root node receiving exactly one unit of auxiliary traffic; the auxiliary traffic carried on the line is limited by the auxiliary spanning tree indicator variable; the total number of lines included in the auxiliary spanning tree is equal to the number of nodes minus one; and limiting the actual operating state variable of the line to not exceed the corresponding auxiliary spanning tree indicator variable, thereby allowing some tree edges to break in actual operation to form multiple subtrees rooted in the activated distributed power source, so as to meet the requirements of radial operation; the generated forest constraint and the line power flow limit in the robust security constraint share the same line operating state variable, so that the radial structure remains unchanged under any uncertain scenario, avoiding non-convexity caused by state switching.

[0023] As a preferred technical solution of the present invention, for lines that have not yet been repaired, the actual operating state variable of the line is forcibly set to zero during the corresponding period. The virtual flow, auxiliary flow and actual active and reactive power flow on the line are shielded by the state variable and the large number method constraint, so as to avoid power being allocated to unavailable lines. The forced zeroing operation also applies to the line power flow variable in the robust safety constraint to ensure that the unrepaired line does not participate in any uncertainty transmission.

[0024] As a preferred embodiment of the present invention, the linearized power flow model establishes active power balance and reactive power balance relationships for each node. By introducing the method of large numbers constraint, the squared voltage amplitude of the node is coupled with the active power flow, reactive power flow, and actual operating state variables of the corresponding line. This ensures that the squared voltage amplitude relationship only applies when the line is in a closed state; for lines in an open state, the node voltages at both ends are independently determined by the source end of their respective microgrids. In the robust security constraint, the uncertainty of the node injected power is expressed as the algebraic sum of the load forecast deviation and the distributed generation output deviation, and a box-type uncertainty set is used to describe the uncertainty. The fluctuation range is described by introducing auxiliary variables and dual multipliers, transforming the original equilibrium equations and voltage constraints containing uncertain parameters into a system of linear inequalities containing only deterministic variables; the time-series charging and discharging coordination constraints of the energy storage device further include: in the preparatory stage before the non-black start unit starts, if the energy storage state of charge is lower than a preset threshold, the black start unit is used to charge it first to ensure the discharge capacity at the moment of start-up; in the recovery stage after start-up, the charging power of the energy storage is limited to not exceeding the total surplus capacity of the distributed power sources in the current microgrid, so as to avoid the charging process crowding out the load recovery power, and realize the full-cycle energy storage management of pre-charging, in-process support and post-replenishment.

[0025] As a preferred technical solution of the present invention, in the linearized power distribution flow model, the active and reactive power outputs of the inactive non-black start unit nodes are forced to zero, the active and reactive power outputs of the activated distributed power nodes are respectively limited to their capacity limits, the active and reactive power flows of each line are strictly limited to the product of its transmission capacity and the actual operating state variables of the line, and the square of the node voltage amplitude is limited to the pre-set upper and lower voltage limits to ensure that each device operates in a safe range during the recovery process; the robust safety constraint further requires that all the above limit inequalities hold true for all scenarios in the uncertainty set, that is, the worst-case criterion is used for verification, so as to obtain a recovery scheme that is immune to prediction errors.

[0026] As a preferred technical solution of the present invention, the non-black start unit priority activation constraint, the generated forest constraint, the linearized power flow model, and the robust security constraint are jointly constituted as a robust mixed-integer linear programming model under a single recovery period. The robust mixed-integer linear programming model is solved using a mixed-integer programming solver to obtain the microgrid partitioning, the set of activated non-black start units, the output of distributed power sources, and the load recovery decision for the current period. The robust equivalence transformation is achieved through the duality theory of linear programming, without introducing integer variables, thus maintaining the model solution efficiency.

[0027] As a preferred technical solution of the present invention, during the cyclic execution of the solution for each recovery period, if it is determined that all distributed power sources have been activated, the solution for the priority activation constraint of the non-black start unit is omitted in subsequent recovery periods. Only the generated forest constraint, the linearized power flow model, and the robust safety constraint are solved based on the updated set of available lines. If the recovery of all recoverable loads has been completed in the current period and there are no new non-black start units that can be activated, the loop ends. If the robust safety constraint is not feasible in the current period, the system reverts to the deterministic model and issues an early warning, prompting the operator to check the prediction interval or adjust the recovery strategy. If the dynamic standby constraint is not feasible, a tiered start strategy is triggered, and the non-black start units are activated in batches according to the impact power from small to large, or flexible load resources such as electric vehicles are temporarily called to supplement the standby capacity, so as to ensure that the recovery process can still be safely advanced when the transient support capacity is insufficient.

[0028] As a preferred technical solution of the present invention, the line repair status and the location information of the distributed power source are collected through an interface that connects with the on-site emergency repair management system and the distribution automation system. The obtained switch action commands, distributed power source start-up and shutdown commands, and load switching commands are sent to the corresponding remote terminal units to execute the recovery plan. At the same time, the load forecast interval and the distributed power source output forecast interval are updated online based on historical measurements and meteorological information using a data-driven method. The updated interval is used as the input of the uncertainty set of the next recovery period to achieve adaptive rolling optimization for robust recovery.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] Under the premise that the line repair sequence is constrained by external field conditions and cannot be freely scheduled, this invention fully explores the flexibility of the distribution network topology and introduces a virtual power flow mechanism. This allows each non-black start unit that has not yet been activated to be preferentially connected to the microgrid composed of activated distributed power sources and externally activated once its topology reachability conditions are met. This enables the high-capacity power generation resources to be put into operation earlier than the traditional passive recovery method based on the line repair progress, and accelerates the growth of the cumulative number of activated distributed power sources and the cumulative restored load in the initial stage of recovery. Combined with the microgrid dynamic partitioning constraints based on generated forest and the linearized distribution power flow constraints, this invention ensures that radial operation, voltage amplitude and line power flow meet safety limits, while significantly reducing the dispersion of distributed power source capacity utilization under different fault scenarios, and improving the adaptability of the recovery scheme to the diversity of post-disaster line conditions.

[0031] This invention, based on prioritizing the activation of non-black start units and dynamically partitioning the microgrid, introduces robust safety constraints based on interval uncertainty sets. This ensures that the proposed recovery scheme is effective not only in the most likely predicted scenarios but also guarantees that node voltage and line power flow do not exceed limits despite any fluctuations in load and distributed generation output. This characteristic improves the reliability and anti-interference capability of the recovery scheme in actual operation, avoiding repeated load shedding operations caused by prediction errors. Simultaneously, the introduction of robust constraints does not alter the mixed-integer linear programming nature of the original problem; high solution efficiency is maintained through linear dual transformation, making it suitable for online rolling applications.

[0032] This invention tightly couples robust safety constraints and priority activation mechanisms through line state variables and power supply available capacity boundaries to form a three-in-one collaborative optimization architecture of activation-topology-robust scheduling. This architecture not only leverages the large capacity advantage of non-black start units to accelerate recovery but also ensures safe operation in uncertain environments, resolving the contradiction between recovery speed and robustness in traditional methods.

[0033] By constructing dynamic backup constraints, the recovery scheme is upgraded from static capacity matching to dynamic power support. Traditional capacity constraints only focus on whether the total is sufficient, while this invention requires a sufficiently fast power response at the moment of startup. This greatly improves the success rate of on-site startup of non-black start units, avoids repeated startup failures caused by power surges, and thus accelerates the overall recovery process.

[0034] By introducing time-sequential charge and discharge constraints on energy storage devices, the contradiction between startup surge and steady-state recovery is resolved. The rapid response characteristics of energy storage provide a safety net during the initial surge, while the surplus output of distributed power sources replenishes the energy storage after successful startup, transforming energy storage from a simple backup component into a recyclable buffer. This mechanism of charging before startup and replenishing after startup, deeply complements the priority activation strategy, neither slowing down the activation speed nor sacrificing the sustainability of long-term recovery.

[0035] By simultaneously solving the ramp rate and minimum start / stop time constraints with robust safety constraints, the recovery scheme is ensured to meet not only steady-state safety limits but also dynamic adjustment capabilities. Under arbitrary load and output fluctuation scenarios, the system has sufficient adjustment margin to track changes, effectively solving the fatal flaw of static optimization and dynamic instability in existing recovery schemes. Attached Figure Description

[0036] Figure 1 This is a schematic diagram illustrating the evolution of the collaborative control method provided in the embodiments of the present invention during a typical recovery phase.

[0037] Figure 2 This is an overall flowchart of the collaborative control method provided in the embodiments of the present invention;

[0038] Figure 3 This is a schematic diagram of the improved IEEE 123-node test system topology used in an embodiment of the present invention;

[0039] Figure 4 This is a comparison chart of the average recovery load in embodiments of the present invention;

[0040] Figure 5 This is a comparison chart of distributed power supply capacity utilization in embodiments of the present invention;

[0041] Figure 6 This is a comparison chart of the standard deviation of capacity utilization rates in embodiments of the present invention. Detailed Implementation

[0042] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, so that those skilled in the art can clearly understand the technical solutions and implementation methods adopted by the present invention. The described embodiments are only some preferred examples of the present invention, and other modifications and equivalent substitutions made by those skilled in the art without creative effort are all within the protection scope of the present invention.

[0043] This embodiment addresses an active distribution network containing distributed generation. This network comprises several black-start units and several non-black-start units, both coupled to node loads via feeders. Black-start units possess self-starting capability, enabling them to start and establish a voltage reference based on their own energy storage or independent power units after a complete system outage. Non-black-start units lack the ability to establish a voltage reference on their own, but their rated capacity is significantly higher than that of black-start units, requiring external power sources to be present before they can be connected to the network. For clarity, this embodiment assumes the connection between the main feeder and the substation is unavailable, meaning the distribution network relies entirely on internal distributed generation for recovery during the post-disaster phase. However, this assumption does not limit the applicability of the invention; it also applies to situations where the main grid is partially available or where the main grid resumes power supply in the later stages of recovery.

[0044] like Figure 1 As shown, the recovery strategy proposed in this invention exhibits a typical three-step evolution process at a specific recovery stage. In the initial stage, only black-start units are energized in the system, while non-black-start units are still out of service, with the loads to be restored scattered throughout the network. Since the black-start units have limited capacity and cannot independently undertake large-scale load restoration tasks, the recovery process does not directly proceed towards maximizing load capacity. Instead, it first aims to activate non-black-start units, establishing a microgrid topology extending from the black-start units to the nearest accessible non-black-start unit. Once the non-black-start unit is integrated into the energized grid and starts up, its large-capacity power generation capacity is added to the dispatchable resource pool, forming a new power support together with the original black-start units. Based on this, the microgrid boundary is reorganized using the sum of the capacities of all currently operating distributed power sources as a limit, incorporating as many accessible loads as possible into the recovery scope.

[0045] To support the implementation of the above strategy, this embodiment introduces a virtual power flow mechanism to characterize the reachability of non-black-start units and the topological feasibility of the corresponding microgrid. It should be noted that the virtual power flow described here is only used as a modeling tool to characterize connectivity and activation feasibility, and does not correspond to physical active or reactive power flow. Let... This represents the set of nodes where the distributed power source is currently active. This represents the set of nodes containing non-black start units that have not yet been activated. Indicates a collection of power distribution lines. This represents the set of recovery periods. Decision variables are introduced. Used to characterize nodes The non-black start unit at the time period Has the current topology received data from an activated distributed power source? The emitted unit virtual stream introduces Indicates distributed power source Issued, during the time period via route The transmitted virtual traffic. Based on this, the topological constraints preferentially activated by non-black start units can be characterized by the following relationship:

[0046]

[0047]

[0048]

[0049]

[0050]

[0051] in, For distributed power Total amount of virtual flows injected into the network, binary variable Used to characterize the circuit During the period The running status, The above relationship expresses the physical meaning that: each active distributed power source injects a virtual flow into the network; each inactive non-black-start unit must receive a virtual flow from any active power source through the current topology; and the virtual flow emitted by any distributed power source must not be absorbed by other active power sources, thus ensuring the validity of the activation path. Lines that are not repaired or are disconnected are then processed through a large positive number. The constraints are directly shielded, making it impossible to handle the transmission of virtual streams.

[0052] Based on the virtual power flow constraints, this embodiment also needs to ensure that the microgrid formed in each recovery phase meets the requirements of radial operation of the distribution network, i.e., it is a spanning forest composed of several spanning trees with activated distributed generation sources as root nodes. To this end, spanning tree auxiliary variables are introduced. Characterizing the circuit Whether to include it in the auxiliary spanning tree structure and introduce an auxiliary root node. and its corresponding auxiliary flow Construct the following constraints for generating a forest:

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] The above relationships together construct an auxiliary spanning tree covering all nodes, where the virtual root node injects into the network. A unit of auxiliary flow is received by each non-root node, thus establishing tree connectivity; this is achieved by varying the actual operating variables of the line. Limit to no more than the spanning tree indicator variable By allowing some tree edges to be disconnected in the actual running topology, a spanning tree can be divided into multiple unconnected subtrees, forming several microgrids rooted at activated distributed power sources, thus constituting the required spanning forest structure. For time period The set of lines that have been repaired and are available; lines outside of this set remain disconnected during actual operation.

[0060] After determining the microgrid topology and the activation paths of non-black-start units, it is necessary to further determine the load recovery decisions of each node and the output of each distributed power source under this topology. To balance solution efficiency and accuracy, this embodiment uses a linearized power flow model to describe the active and reactive power balance of each branch and the node voltage relationships. Let... To represent nodes During the period A binary variable indicating whether the load has been recovered. For nodes Load priority weights, , They are nodes During the period Active and reactive loads , They are nodes Distributed power sources during time periods Those who contribute meritoriously and those who contribute without merit, , For the corresponding capacity limit, , The lines are respectively The tide of merit and inmerit on the surface , The lines are respectively Resistance and reactance, For nodes The square of the voltage amplitude, , These are the lower and upper limits of the square of the voltage amplitude. , The lines are respectively The active and reactive power transmission capacity. The load recovery model is expressed as follows:

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] By introducing large The method couples node voltage relationships with line state variables, ensuring that the voltage equations only apply between the two node ends when the line is closed. When the line is open, the node voltages at both ends are independently determined by the source end of their respective microgrids. Active and reactive power flows for each line are strictly limited to the product of its transmission capacity and line state variables, thus avoiding the application of power to disconnected or unrepaired lines. The entire load recovery model, together with the aforementioned virtual power flow constraints and spanning forest constraints, constitutes a mixed-integer linear programming model under a single recovery phase, which can obtain the global optimum within a reasonable timeframe using a commercial solver.

[0072] To address the uncertainties in load and distributed generation output during actual recovery, this embodiment further constructs robust safety constraints based on the aforementioned deterministic model. Specifically, for each node, its actual active load is represented as a predicted value plus a deviation term, which varies within a given interval. Similarly, the actual active output of each activated distributed generation is also represented as the sum of the predicted value and the deviation term, with the deviation also within an interval. The load deviations and power output deviations of all nodes together constitute a box-shaped uncertainty set. Reactive power is also described using the same interval-based approach.

[0073] When constructing robust constraints, it is required that for any set of deviation values ​​in the uncertainty set, the power balance equations of each node, the power flow limits of each line, and the voltage amplitude limits of each node must all be satisfied simultaneously. This means that the recovery scheme must be able to withstand the worst-case uncertainty. To transform this requirement of "holding true for any deviation" into a solvable form, this embodiment uses linear duality theory, introducing auxiliary dual variables for each constraint containing deviations, and using the strong duality theorem to equivalently transform the infinite number of scenario conditions into a finite number of linear inequalities containing only these auxiliary variables and the original deterministic variables. After the transformation, the entire model still maintains a mixed-integer linear programming structure, without the need to generate a large number of random scenarios, resulting in high computational efficiency.

[0074] During the rolling optimization process, before the start of each time period, the system updates the predicted intervals for the load of each node and the output of distributed power sources online using data-driven methods (such as interval estimation based on historical measurement errors or combined with weather forecast confidence). The updated intervals are then used as input to the uncertainty set of the current time period, enabling robust constraints to adaptively reflect the latest prediction uncertainties. If the robust model has no feasible solution for a certain time period, it indicates that the current prediction interval is too optimistic or that network resources are insufficient to handle all possible situations. In this case, the system will automatically revert to a deterministic model and issue an early warning signal, reminding dispatchers to check the data or manually adjust the recovery strategy to ensure that the algorithm has outputs under any circumstances.

[0075] It is important to emphasize that robust safety constraints do not operate independently, but rather work closely with virtual flow priority activation constraints and spanning forest constraints. Specifically, the set of currently activated distributed power sources determined by virtual flow constraints directly limits which power sources in the robust constraints can have non-zero output ranges; the output ranges of inactive power sources are forced to be zero. Simultaneously, the actual operating state variables of the lines determined by spanning forest constraints are directly used to limit robust power flow variables, ensuring that the topology remains fixed under uncertain disturbances and preventing non-convexity problems caused by state transitions. Through this design, priority activation, topology construction, and robust scheduling form a unified optimization framework, jointly ensuring the safety and efficiency of the recovery scheme in uncertain environments.

[0076] To verify the effectiveness of the robust constraints mentioned above, this embodiment conducted Monte Carlo simulations on an improved IEEE 123-bus system, adding ±20% random error to the load and photovoltaic output predictions, generating 500 uncertain scenarios. The results show that the scheme using the robust constraints proposed in this invention did not experience voltage over-limits or line overloads in any scenario; while the original deterministic scheme (without robust constraints) experienced over-limits in approximately 18% of the scenarios. This fully demonstrates that the newly added robust safety constraints can effectively improve the recovery scheme's resistance to prediction errors, and work well with the priority activation mechanism without affecting the power supply acceleration activation effect in the initial recovery phase.

[0077] To incorporate the startup transient process of non-black start units into the collaborative control, this embodiment further adds dynamic standby constraints to the model. Specifically, after determining the non-black start unit to be activated during each recovery period, the system obtains the startup impact power curve of the unit by looking up a table or based on historical measured data. This curve typically shows a high-power pulse lasting several seconds to tens of seconds at the moment of startup, with an amplitude that can reach 2 to 5 times the rated capacity of the unit. In this embodiment, this impact power is equivalent to a constant power demand during the startup duration.

[0078] Based on this, the system calculates the effective reserve capacity of all activated distributed generation sources within the current microgrid. This effective reserve capacity is defined as the difference between the maximum current generating power and the current actual output of each distributed generation source, and is also limited by its ramp rate, meaning the reserve capacity must be fully released within the duration of the impact. If the calculated total effective reserve capacity is less than the impact power requirement of the unit to be activated, the optimization model will refuse to activate that unit in the current time period, or instead select other non-black-start units with lower impact power, thus achieving a dual screening of topology-priority activation and physical start feasibility.

[0079] In this embodiment, energy storage devices are treated as a special type of fast-response power source. The model not only considers upper and lower limits for charging and discharging power and capacity, but also introduces a temporal coupling relationship for the state of charge (SBC), meaning the SBC at the end of the previous time period equals the SBC at the beginning of the current time period. Before the non-black-start unit starts, if the SBC of the energy storage is low, the model will force the activated distributed power sources to charge the energy storage device while simultaneously fulfilling load recovery requirements, ensuring it has sufficient discharge capacity. At startup, the energy storage device participates in support with maximum discharge power, compensating for insufficient spinning reserve. After startup, the system automatically switches the energy storage device that has completed its support task to charging mode, utilizing the current surplus power in the microgrid to restore its SBC. To balance load recovery, this embodiment specifically sets an upper limit constraint on charging power, meaning the charging power of the energy storage must not exceed a certain proportion of the current total surplus capacity of the microgrid, preventing excessive charging from crowding out load recovery space.

[0080] To address extreme situations, this embodiment employs a tiered startup strategy. If the dynamic reserve constraint has no feasible solution in the current time period, the system will not forcibly start the large-capacity unit. Instead, it will prioritize activating several small-capacity non-black start units with lower impact power. Once the small-capacity units are connected to the grid and the overall inertia and support capacity of the microgrid have increased, the large-capacity unit will be started in a subsequent time period. Simultaneously, the model incorporates minimum continuous operation and downtime constraints to prevent frequent start-ups and shutdowns of distributed power sources or energy storage devices within a short period, which could damage equipment lifespan.

[0081] The aforementioned dynamic reserve constraints and energy storage timing constraints are all linearized at the modeling level (for example, the ramp rate constraint is transformed into upper and lower limit constraints on the output change between adjacent time periods). Therefore, together with the original virtual flow constraints, generated forest constraints, and robust safety constraints, they maintain the structure of mixed-integer linear programming, and the solution efficiency is not significantly affected. In actual deployment, the above parameters (impact power, ramp rate, and real-time state of charge of energy storage) are obtained through real-time data interaction with the field equipment control system and energy management system, ensuring the consistency between the model input and the actual physical situation on site.

[0082] like Figure 2 As shown, this embodiment embeds the aforementioned single-stage optimization into a rolling optimization framework to adapt to the gradual advancement of the line repair process over time. Specifically, in the initial stage, the line damage and repair status obtained after the disaster are used as initial inputs to solve the aforementioned model to determine the currently active set of distributed generation sources. Based on the total available capacity under this set, load recovery decisions are further solved to obtain the current microgrid partitioning, distributed generation output, and recovered load nodes. Subsequently, in each new time period, the model is updated based on the latest line repair results from external feedback. The two-step solution process of priority activation and load restoration is then re-executed. If it is determined that all distributed power sources are already activated, the activation model will not be solved in subsequent stages; only load restoration under the latest topology will be optimized. If the current stage has completed the restoration of all recoverable loads and there are no new power sources to be activated, the rolling iteration ends. This process is repeated until all distributed power sources are activated and all recoverable loads are included in the restoration scope.

[0083] In engineering implementation, the above method can be undertaken by the recovery decision module deployed at the distribution master station or regional coordination control center. This module interfaces with the on-site emergency repair management system and distribution automation system to obtain line repair status, distributed power source operating status, and load measurement data. It then sends the solved switch action commands, distributed power source start / stop commands, and load switching commands to the corresponding remote terminal units, completing the actual execution of the recovery plan. At the solver level, a commercial mixed integer programming solver can be selected, and a reasonable upper limit for the solution time can be set according to the network size to meet real-time requirements.

[0084] like Figure 3 As shown, as a specific implementation example, this embodiment validates the proposed method on an improved IEEE 123-node test system. This system, based on the original standard example, configures three black-start units with a single unit capacity of 300 kW and eight non-black-start units with a single unit capacity of over 700 kW. To highlight the role of local distributed power sources, the main substation interconnection feeder is considered to be permanently unavailable, making the recovery process entirely dependent on internal distributed power sources. All models are constructed in a general mathematical programming environment and solved by a commercial mixed-integer programming solver. Compared with traditional radial passive recovery strategies and recovery strategies that rely solely on dynamic microgrid reconfiguration without introducing a priority activation mechanism, the method proposed in this embodiment can mobilize more distributed power sources to join the operation in the early stages of recovery, resulting in significantly higher cumulative activated units and cumulative restored load during critical periods such as Phase 1 and Phase 6 compared to the control method. In a Monte Carlo simulation of 500 random fault scenarios, the method in this embodiment achieves higher average restored load (e.g., Figure 4 As shown), distributed power supply capacity utilization (such as...) Figure 5 (as shown) and the standard deviation of capacity utilization (as shown) Figure 6 It shows a relatively stable advantage in indicators such as (as shown in the figure).

[0085] It should be noted that the network scale, unit capacity, and parameter values ​​given in the above embodiments are merely illustrative examples and do not constitute a limitation on the scope of protection of this invention. In practical applications, this invention is equally applicable to different network structures, different distributed power supply configurations, and even different load priority weight settings; for distribution networks equipped with energy storage devices, controllable loads, or flexible interconnection equipment, further expansion can be achieved by supplementing the corresponding operational constraints into the model of this embodiment. Those skilled in the art can also combine the virtual power flow mechanism with other types of connectivity constraints based on the technical concept of this invention, or replace the linearized distribution power flow model with a higher-precision power flow approximation to adapt to different engineering accuracy requirements. Such replacements or expansions are all equivalent implementations of this invention.

Claims

1. A coordinated control method for dynamic microgrid formation and priority activation of non-black start units, applied to post-disaster recovery of distribution networks containing black start units, non-black start units, and energy storage devices, characterized in that, include: Obtain the repair status of each line in the distribution network during the current recovery period and the location information of activated distributed power sources to determine the set of currently available lines; Based on the available line set, and with the constraints of activated distributed power sources injecting virtual flows into the network and non-activated non-black start units receiving at least one unit of virtual flow, a priority activation constraint for non-black start units is constructed, and the set of non-black start units activated in the current time period is solved. Using the node where the activated distributed generation is located as the root node, a forest constraint is constructed based on the auxiliary spanning tree indicator variable, so that the distribution network is divided into multiple microgrids with the activated distributed generation as the root and which are not interconnected in the current time period. Obtain the active power prediction interval and reactive power prediction interval of each node load, as well as the active power output prediction interval and reactive power output prediction interval of each activated distributed power source, and construct an uncertainty set represented in interval form. Obtain the upper limit of ramp rate, minimum continuous operation and downtime of each distributed power source, as well as the instantaneous impact power and start-up impact duration required by each non-black start unit during the start-up process; To establish dynamic backup constraints, it is required that currently activated distributed power sources, in addition to meeting the current load recovery needs, reserve an additional spinning reserve capacity of not less than the startup impact power of the non-black start unit to be activated. This reserve capacity must remain available throughout the startup impact duration. Construct a time-series charging and discharging coordination constraint for the energy storage device. At the moment of startup of the non-black start unit, control the energy storage device to release power quickly to make up for the difference between the reserve capacity and the actual impact demand. After startup, use the surplus output of the distributed power source to recharge the energy storage device so that its state of charge returns to the safe range. Under the generated forest topology, with the goal of maximizing the weighted recovery load, a linearized power flow model is used, and robust security constraints considering the uncertain set are introduced to solve for the recovery state of the load at each node, the active and reactive power output of distributed generation sources, the charging and discharging power of energy storage devices, and the power flow of each line. The robust security constraints ensure that the node voltage amplitude and line power flow do not exceed the limits under any realization within the uncertain set, and the robust constraints are transformed into deterministic linear constraints through linear dual transformation. The dynamic reserve constraints and energy storage timing constraints are solved simultaneously with the robust security constraints to ensure sufficient transient support capability under any uncertain scenario. Entering the next recovery period, the set of available lines is updated based on the latest line repair status feedback from the outside, and the above steps are repeated until all non-black start units are activated and all recoverable loads are restored.

2. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: The non-black start unit priority activation constraint aims to maximize the number of non-black start units activated in the current time period. For each inactive non-black start unit, the virtual flow received from the activated distributed power source is characterized by the difference between the virtual inflow and outflow to the non-black start unit node. The total amount of virtual flow injected into the network by each activated distributed power source is limited by its corresponding injection variable. The virtual flow constraint and the robust security constraint share the same set of actual line operating state variables, thereby ensuring the feasibility of the activation path and the feasibility of the recovery scheme under uncertainty. The priority activation ranking results prioritize non-black start units with relatively small start-up impact power and close electrical distance to the activated power supply, in order to reduce dynamic backup pressure and achieve pre-matching between topology-level activation decisions and transient power support capabilities.

3. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 2, characterized in that: The virtual flow is only used as a modeling quantity to characterize connectivity and activation feasibility, and does not correspond to active or reactive power in a physical sense. The virtual flow emitted by any activated distributed power source is not absorbed by other activated distributed power sources, and any virtual flow variable is forced to zero by the large number method constraint when the corresponding line is not repaired or is in a disconnected state. The solution of the virtual flow constraint is independent of the uncertain interval of load and output, but its result serves as the boundary condition of the available capacity of distributed power sources in the robust safety constraint, realizing the time-series coupling of priority activation and robust recovery. At the same time, the list of units to be activated obtained by solving the virtual flow constraint uses its starting impact power as the input boundary of the dynamic standby constraint, realizing the dual verification of topology reachability and physical startability. Only non-black start units that simultaneously meet the connectivity condition and transient power support condition are allowed to be activated.

4. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: The construction of the generated forest constraints includes: injecting auxiliary traffic equal to the number of nodes minus one into the network through auxiliary root nodes, with each non-auxiliary root node receiving exactly one unit of auxiliary traffic; limiting the auxiliary traffic carried on the lines by auxiliary spanning tree indicator variables; ensuring that the total number of lines included in the auxiliary spanning tree is equal to the number of nodes minus one; and restricting the actual operating state variables of the lines to not exceed the corresponding auxiliary spanning tree indicator variables, thereby allowing some tree edges to break during actual operation and form multiple subtrees rooted in activated distributed power sources; the generated forest constraints and the line power flow limits in the robust security constraints share the same line operating state variables, ensuring that the radial structure remains unchanged under any uncertain scenario and avoiding non-convexity caused by state switching.

5. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 4, characterized in that: For lines that have not yet been repaired, their actual operating state variables are forcibly set to zero during the corresponding time period, and the virtual flow, auxiliary flow, and actual active and reactive power flow on the line are jointly shielded by the state variable and the large number method constraint. This forced zeroing operation also applies to the line flow variables in the robust safety constraints, ensuring that unrepaired lines do not participate in any uncertainty propagation.

6. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: The linearized power flow model establishes active power balance and reactive power balance relationships for each node, and couples the square of the node voltage amplitude with the active power flow, reactive power flow and actual operating state variables of the corresponding line by introducing large number constraints, so that the square of the voltage amplitude relationship only takes effect at both ends of the line when the line is in a closed state. In the robust safety constraints, the uncertainty of node injected power is represented as the algebraic sum of load forecast deviation and distributed power output deviation, and its fluctuation range is described by a box-type uncertainty set. By introducing auxiliary variables and dual multipliers, the original balance equations and voltage constraints containing uncertain parameters are transformed into a set of linear inequalities containing only deterministic variables. The time-series charging and discharging coordination constraints of the energy storage device further include: in the preparatory stage before the start-up of the non-black start unit, if the energy storage state of charge is lower than a preset threshold, the black start unit is used to charge it first to ensure the discharge capacity at the moment of start-up; in the recovery stage after start-up, the charging power of the energy storage is limited to not exceed the total surplus capacity of the distributed power in the current microgrid, so as to avoid the charging process crowding out the load recovery power and realize the full-cycle energy storage management of pre-charging, in-process support and post-replenishment.

7. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 6, characterized in that: In the linearized power flow model, the active and reactive power outputs of inactive non-black-start unit nodes are forced to zero, the active and reactive power outputs of activated distributed power nodes are limited to their respective capacity limits, the active and reactive power flows of each line are strictly limited to the product of its transmission capacity and the actual operating state variables of the line, and the square of the node voltage amplitude is limited to the pre-set upper and lower voltage limits. The robust safety constraint further requires that all the above limit inequalities hold true for all scenarios in the uncertainty set, that is, the worst-case criterion is used for verification, so as to obtain a recovery scheme that is immune to prediction errors.

8. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: The non-black start unit priority activation constraint, the generated forest constraint, the linearized power flow model, and the robust security constraint together constitute a robust mixed-integer linear programming model under a single recovery period. The robust mixed-integer linear programming model is solved using a mixed-integer programming solver to obtain the microgrid partitioning, the set of activated non-black start units, the output of distributed generation, and the load recovery decision for the current period. The robust equivalence transformation is achieved through the dual theory of linear programming, without introducing integer variables, thus maintaining the model solution efficiency.

9. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: During the iterative execution of the solution for each recovery period, if it is determined that all distributed power sources have been activated, the solution for the non-black start unit priority activation constraint is omitted in subsequent recovery periods. Instead, the generated forest constraint, the linearized power flow model, and the robust safety constraint are solved based solely on the updated set of available lines. If all recoverable loads have been restored in the current period and no new non-black start units can be activated, the cycle ends. If the robust safety constraints are not feasible in the current period, the system reverts to the deterministic model and issues an early warning, prompting operators to check the prediction range or adjust the recovery strategy. If the dynamic standby constraints are not feasible, a tiered start strategy is triggered, activating non-black start units in batches according to their impact power from small to large, or temporarily calling on flexible load resources such as electric vehicles to supplement standby capacity, ensuring that the recovery process can still proceed safely when transient support capacity is insufficient.

10. The coordinated control method for dynamic microgrid formation and priority activation of non-black start units according to claim 1, characterized in that: The line repair status and the location information of the distributed power source are collected through the interface with the on-site emergency repair management system and the distribution automation system. The obtained switch action commands, distributed power source start-up and shutdown commands, and load switching commands are sent to the corresponding remote terminal units to execute the recovery plan. At the same time, the load forecast interval and the distributed power source output forecast interval are updated online based on historical measurements and meteorological information using a data-driven method. The updated interval is used as the input of the uncertainty set of the next recovery period to achieve adaptive rolling optimization for robust recovery.