Three-line defense adaptive and collaborative prevention and control method for grid-type new energy power systems

By integrating multi-dimensional security margins and adaptive defense line switching, the problem of traditional power systems being unable to adapt to changes in grid-connected new energy sources has been solved, enabling the effective utilization of new energy sources and coordinated cooperation between defense lines, thereby improving the safety and stability of the power system.

CN123073759APending Publication Date: 2026-11-10이너 몽골리아 일렉트릭 파워 그룹 컴퍼니 리미티드 이너 몽골리아 일렉트릭 파워 리서치 인스티튜트 브랜치
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611585205.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-14
Publication Date
2026-11-10

Smart Images

  • Figure CN123073759A_ABST
    Figure CN123073759A_ABST
Patent Text Reader

Abstract

This invention relates to a three-tiered adaptive and coordinated control method for grid-connected renewable energy power systems, belonging to the field of power system safety and stability control technology. The method involves collecting system data, calculating five-dimensional safety margins (frequency, voltage, synchronization stability, current, and energy), fusing these margins into a dynamic grid safety margin using a smoothing minimum operator, and performing short-term prediction and conservative contraction. The three-tiered control mode is switched based on thresholds and hysteresis times. The first tier optimizes active power reserve, dynamic reactive power, and state of charge (SOC). The second tier weights the current based on frequency and voltage risk, calculates the projected current within current circle constraints, and performs power and energy verification to prevent saturation. The third tier actively disconnects grids based on grid capacity, dynamic reactive power, energy, and active power balance constraints, and calculates adaptive load shedding. This achieves unified quantitative triggering and coordinated control of the three tiers. This invention realizes unified quantitative triggering and coordinated continuous and discrete safety margin control for grid-connected renewable energy systems participating in the three-tiered control system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of power system safety and stability control technology, specifically relating to a three-line adaptive and coordinated prevention and control method for grid-type new energy power systems. Background Technology

[0002] As the proportion of grid-connected renewable energy generation units in the power system continues to increase, the dynamic characteristics of the power system have changed significantly. Unlike traditional synchronous power sources, grid-connected renewable energy is connected to the grid through power electronic converters, possessing external voltage source characteristics and being able to actively provide frequency and voltage support. However, its overcurrent capacity is weak, its energy release is limited, and its transient response is constrained by the controller. Its dynamic behavior during faults is fundamentally different from that of synchronous machines.

[0003] The existing three lines of defense in the power system mainly include: the first line of defense centered on relay protection; the second line of defense centered on safety and stability control devices; and the third line of defense centered on safety automatic devices such as out-of-step disconnection and low-frequency, low-voltage load shedding. Traditional defenses are typically based on offline simulation and fixed threshold settings, with each line of defense functioning relatively independently, lacking unified safety margin metrics and coordination mechanisms between lines. After grid-connected renewable energy is integrated, the system's inertia level, short-circuit capacity, frequency, and voltage support characteristics all change over time. Traditional fixed-setting defenses are ill-suited to complex and variable operating conditions, potentially leading to problems such as unreasonable timing of defense actions and contradictions between control measures and renewable energy responses. Furthermore, grid-connected renewable energy inherently possesses rapid and continuous active and reactive power regulation capabilities, but the existing defense architecture fails to incorporate this into a unified control system, thus failing to effectively utilize the control potential of grid-connected renewable energy.

[0004] Therefore, a three-line adaptive and coordinated control method suitable for grid-type new energy power systems is needed to achieve unified quantification of dynamic safety margin, automatic switching of defense mode, and coordinated cooperation of control measures of each defense line, thereby improving the safe and stable operation level of the power system. Summary of the Invention

[0005] The purpose of this invention is to provide a three-line adaptive and coordinated prevention and control method for grid-type new energy power systems, which realizes unified quantitative assessment of dynamic safety margin, automatic switching of the three-line mode, adaptive coordination of control measures of each line of defense, and full utilization of controllable resources of grid-type new energy, thereby improving the safe and stable operation level of the power system.

[0006] To achieve the above objectives, the technical solution of the present invention is: a three-line adaptive and coordinated prevention and control method for a grid-type new energy power system, comprising:

[0007] Acquire the frequency, rate of change of frequency, key bus voltage, power angle of grid-connected devices, internal angular frequency, dq axis current, upper limit of converter current, state of charge of energy storage and available active power data of the power system including synchronous generators, grid-connected new energy, grid-connected new energy / grid-connected energy storage and AC network during operation.

[0008] Calculate the frequency amplitude margin and frequency change rate margin based on the obtained frequency and frequency change rate, respectively; calculate the voltage margin based on the critical bus voltage; calculate the synchronization stability margin based on the power angle, kinetic energy and critical energy of the grid-type device; calculate the current margin based on the dq axis current and the upper limit of the converter current; and calculate the energy margin based on the available energy of energy storage and active reserve.

[0009] The frequency amplitude margin, frequency change rate margin, voltage margin, synchronization stability margin, current margin, and energy margin are fused using the smoothing minimum operator to obtain the current dynamic network safety margin.

[0010] Based on the system state, short-time prediction is used to calculate the prediction safety margin within the future prediction window. The prediction safety margin is then conservatively reduced by combining the state estimation error, available power prediction error, or model mismatch uncertainty to obtain the conservative prediction margin.

[0011] The conservative prediction margin is used to determine the defense mode of the line of defense based on the first threshold, the second threshold and the hysteresis recovery threshold: when the conservative prediction margin is not lower than the first threshold, it enters the first defense mode; when the conservative prediction margin is lower than the first threshold but not lower than the second threshold, it enters the second defense mode; when the conservative prediction margin is lower than the second threshold, it enters the third defense mode.

[0012] In the first line of defense mode, the active power reserve, reactive power reserve, energy storage charge status and operation mode of the grid-type device are optimized according to the conservative forecast margin constraint, and preventive control commands are generated.

[0013] In the second line of defense mode, the active and reactive current tracking weights are determined based on frequency risk and voltage risk. The initial dq axis current command is projected and solved within the converter current circle constraint to generate the current command of the grid-type converter after current limiting. The energy storage energy and available active power are verified and anti-saturation processing is performed.

[0014] In the third line of defense mode, active disconnection and partitioning are carried out based on the grid capacity, dynamic reactive power, releaseable energy and active power balance constraints. Adaptive load reduction is calculated according to system inertia, power deficit and active power that the grid can support, and disconnection and load reduction instructions are generated.

[0015] The generated preventive control commands, converter current commands, disconnection commands, and load reduction commands are sent to grid-type devices, energy storage devices, and safety automatic devices for execution.

[0016] Furthermore, the frequency amplitude margin is calculated using the following formula:

[0017]

[0018] The frequency change rate margin is calculated using the following formula:

[0019]

[0020] Using smooth minimum operator pairs and The fusion is used to obtain a comprehensive frequency margin, which is used to determine the current dynamic network security margin and subsequent defense mode.

[0021] For a single busbar, voltage margin The minimum voltage margin of each bus in the set of critical buses is taken as the system voltage margin.

[0022] in, For system frequency, For the rated frequency, , For frequency safety lower and upper limits, The rate of change of frequency, To allow the maximum rate of change of frequency, For the critical bus voltage, These are the lower and upper limits for voltage safety.

[0023] Furthermore, the synchronization stability margin is calculated based on the potential energy, kinetic energy, and critical energy of the unstable equilibrium point of the network-type device, taking:

[0024]

[0025] in, The current transient total energy is obtained by adding the kinetic and potential energy of the network-type device. The critical energy barrier corresponding to the unstable equilibrium point. To prevent zero decimals;

[0026] The current margin is calculated using the following formula:

[0027]

[0028] in, These are the active shaft and reactive shaft currents of the grid-type converter, respectively. This is the upper limit of the short-time current of the converter;

[0029] The energy margin is calculated based on the ratio of the energy required by the prediction window to the energy that can be released from the storage, and the ratio of active power reserve to the maximum additional active power required by the prediction. The results are then combined into a comprehensive energy margin using a smoothing minimum operator.

[0030] Furthermore, the current dynamic network security margin adopts a weighted smooth minimum operator:

[0031]

[0032] in, For comprehensive frequency margin, For system voltage margin, To ensure synchronous stability margin, For current margin, To consider the overall energy margin, the weighted smoothing minimum operator is used. for:

[0033]

[0034] Where n represents the number of margin components, in this case n=5. It is any margin component to be merged. The normalized weights for each margin component, and ; For smoothing parameters;

[0035] The predicted safety margin is calculated by substituting the future state obtained from the system state equation or second-order extrapolation into the same margin function. The conservative predicted margin... , To predict margin, Determined by state covariance through margin gradient propagation and / or model uncertainty. is the confidence contraction coefficient.

[0036] Furthermore, the defense mode determination is set with a first threshold. Second threshold Hysteresis recovery threshold and continuous confirmation time Hysteresis recovery threshold includes first backoff threshold. Second backoff threshold ,and , , ;

[0037] When the conservative forecast margin is no less than Maintain or enter the first line of defense mode at all times;

[0038] When the conservative forecast margin is lower than and not less than It then enters the second line of defense mode;

[0039] When the conservative forecast margin is lower than Enter the third line of defense mode;

[0040] When in the second line of defense mode and the conservative forecast margin recovers to The above and continue for no less than When the time comes, revert to the first line of defense mode; when in the third line of defense mode and the conservative forecast margin recovers to The above and continue for no less than At that time, retreat to the second line of defense mode.

[0041] Furthermore, the preventative optimization in the first line of defense mode constructs an objective function based on the active power reserve, reactive power reserve, energy storage state of charge deviation from the reference value, and relaxation variables of the grid-type device, and satisfies:

[0042] The conservative forecast margin plus slack variables should not be lower than the first threshold.

[0043] The active power reserve of each network-type device shall not exceed its maximum adjustable reserve, and the sum of reactive power reserve and current reactive power output shall not exceed the dynamic reactive power limit.

[0044] The energy storage state of charge is maintained above the minimum allowable value, the normal operating current command of the grid-type device does not exceed the reserved limit current, and the reserved limit current is less than the short-time current limit of the converter.

[0045] Adjust the active power reserve, reactive power reserve, energy storage charging / discharging status, or network operation mode according to the current minimum dimensional margin.

[0046] Furthermore, in the second line of defense mode, the initial active current command is generated based on the frequency difference and / or the rate of frequency change, and the initial reactive current command is generated based on the voltage deviation and / or the rate of voltage change; when the magnitude of the initial dq-axis current command exceeds the short-time current limit of the converter... At that time, risk based on frequency and voltage risk Determine the weights:

[0047]

[0048]

[0049] in, For comprehensive frequency margin, For voltage margin, , For risk sensitivity coefficient, Based on weights, This is a risk amplification factor. These are the active axis current tracking error weights and the reactive axis current tracking error weights, respectively.

[0050] by , Construct a weighted objective for dq-axis current tracking error, in The constraint-based current-limiting current calculation command is as follows: These are the initial active current command and the initial reactive current command, respectively.

[0051] Furthermore, the current-limiting command is used to construct the Lagrange function:

[0052]

[0053] From the partial derivative stationary point condition, we obtain:

[0054]

[0055] in, These are the active shaft and reactive shaft currents of the grid-type converter, respectively. For Lagrange multipliers;

[0056] When the initial command is located within the current circle When the initial command is outside the current circle, press to make the reconstructed current satisfy... non-negative Solve this problem;

[0057] The obtained optimal active axis current Optimal reactive axis current The active and reactive current commands are converted into those of the grid-type converter, and saturation is limited according to the available active power, the energy stored and released, and the predetermined support time. When saturation occurs, the difference between the actual support and the commanded support is fed back to the frequency outer loop integral state for inverse integral saturation.

[0058] Furthermore, the active disconnection under the third line of defense mode constructs an optimization problem using bus partitioning variables, line disconnection variables, and network-type device attribution variables, and satisfies:

[0059] Each bus belongs to only one isolated network partition;

[0060] The total rated capacity of network-type devices deployed in each isolated network partition shall not be less than the minimum network capacity requirement of that partition.

[0061] The total dynamic reactive power of the network-type devices in each isolated network partition shall not be less than the dynamic reactive power demand of that partition.

[0062] The total energy that can be released by grid-type energy storage in each isolated grid zone is no less than the energy demand that the zone can continuously support.

[0063] The active power generated, active power of the load, active power reserve of the grid, and the remaining load after load shedding in each isolated grid zone meet the active power balance constraints.

[0064] Based on the cost of line disconnection, the active power imbalance and load reduction of each zone, construct the target and output the disconnection scheme and load shedding scheme.

[0065] Furthermore, the adaptive load reduction is calculated based on the system's equivalent inertia, system capacity baseline, rated frequency, allowable frequency drop, predicted power deficit, and average active power support of the grid:

[0066]

[0067] in, The shortfall was due to meritorious service in the accident. To predict the average active power support of the internal structure of the mesh-type device, Let be the system's equivalent inertia constant. This is the baseline value for system capacity. To allow frequency drops, For the rated frequency, For the third line of defense frequency targets, For frequency braking time window;

[0068] Will As the minimum load reduction command, it is issued to the load reduction device; before disconnection or reconnection, parallel verification is performed to ensure that the frequency difference, voltage difference and power angle difference between islanded grids do not exceed the allowable values ​​for parallel connection during the same period.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] 1. Establish normalized safety margins in five dimensions: frequency, voltage, synchronous stability, current, and energy. Then, fuse these margins into a dynamic grid-based safety margin using a smoothing minimum operator. This enables a unified, continuous, and differentiable assessment of the system's safety status, overcoming the shortcomings of traditional single indicators that cannot comprehensively reflect the safety level of grid-based new energy systems.

[0071] 2. By using second-order state extrapolation combined with uncertainty contraction to obtain conservative prediction margin, it is possible to identify safety over-limit risks in the early stage of fault evolution, provide advanced judgment basis for defense actions, and improve the predictability and reliability of prevention and control measures.

[0072] 3. Based on conservative prediction margin, dual thresholds, and hysteresis time, the three-line defense mode adaptive switching is realized, which not only ensures rapid escalation of the defense line in the event of a severe fault, but also avoids frequent switching of the defense line when the margin fluctuates around the threshold, thereby improving the stability and engineering practicality of the control mode switching.

[0073] 4. The first line of defense is constrained by the prediction margin not being lower than the first threshold. It optimizes the adjustment of active power reserve, dynamic reactive power and energy storage SOC, and adaptively adjusts the control focus according to the weakest dimension at present, so as to match the continuous adjustment capability of grid-type new energy with the system security requirements.

[0074] 5. The second line of defense proposes a risk-weighted current circle projection method, which dynamically adjusts the dq-axis current tracking weights based on frequency and voltage risks. Within the current circle constraint, the analytical projection is obtained using the KKT condition (Karush-Kuhn-Tucker condition, first-order optimality condition of nonlinear programming), which can take into account both overcurrent safety and frequency / voltage support requirements. Furthermore, the support power is saturated and limited by checking available active power and releaseable energy, and an inverse integral saturation design is adopted to suppress the integral accumulation after energy depletion and avoid secondary impacts.

[0075] 6. The third line of defense, proactive disconnection, comprehensively considers grid capacity, dynamic reactive power, releaseable energy, and active power balance constraints to ensure that each isolated grid has independent and stable operation capabilities after disconnection; the adaptive load reduction is calculated based on equivalent inertia, power deficit, and the grid's supportable power to determine the minimum load reduction requirement, minimizing load loss while ensuring that the frequency does not fall below the safety lower limit.

[0076] 7. This invention incorporates the continuous regulation capability, current limiting protection constraints, active disconnection and adaptive load reduction of grid-connected renewable energy into a unified three-line defense collaborative prevention and control framework, realizing a smooth connection and coordinated cooperation from normal regulation to emergency control and then to guaranteed control, significantly improving the safety and stability control level of power systems with a high proportion of grid-connected renewable energy. Attached Figure Description

[0077] Figure 1 This is the overall architecture for adaptive and collaborative prevention and control with three lines of defense proposed in this invention.

[0078] Figure 2 This is a typical research system structure for grid-type new energy sources and AC power grids.

[0079] Figure 3 The frequency response of different strategies under complex disturbance scenarios.

[0080] Figure 4 This represents the voltage response of the critical bus under a complex disturbance scenario.

[0081] Figure 5 This diagram illustrates the current safety margin, the predicted safety margin, and the thresholds for the three lines of defense.

[0082] Figure 6 Frequency response under extreme multiple fault scenarios.

[0083] Figure 7 This is a comparison of the final load reduction under extreme multiple fault scenarios.

[0084] Figure 8 This is an example of the sensitivity of the converter current upper limit to the voltage minimum point.

[0085] Figure 9 This is an example of the sensitivity of available energy to frequency / load reduction performance. Detailed Implementation

[0086] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0087] This invention provides a three-line adaptive and coordinated prevention and control method for grid-type new energy power systems, comprising:

[0088] Acquire the frequency, rate of change of frequency, key bus voltage, power angle of grid-connected devices, internal angular frequency, dq axis current, upper limit of converter current, state of charge of energy storage and available active power data of the power system including synchronous generators, grid-connected new energy, grid-connected new energy / grid-connected energy storage and AC network during operation.

[0089] Calculate the frequency amplitude margin and frequency change rate margin based on the obtained frequency and frequency change rate, respectively; calculate the voltage margin based on the critical bus voltage; calculate the synchronization stability margin based on the power angle, kinetic energy and critical energy of the grid-type device; calculate the current margin based on the dq axis current and the upper limit of the converter current; and calculate the energy margin based on the available energy of energy storage and active reserve.

[0090] The frequency amplitude margin, frequency change rate margin, voltage margin, synchronization stability margin, current margin, and energy margin are fused using the smoothing minimum operator to obtain the current dynamic network safety margin.

[0091] Based on the system state, short-time prediction is used to calculate the prediction safety margin within the future prediction window. The prediction safety margin is then conservatively reduced by combining the state estimation error, available power prediction error, or model mismatch uncertainty to obtain the conservative prediction margin.

[0092] The conservative prediction margin is used to determine the defense mode of the line of defense based on the first threshold, the second threshold and the hysteresis recovery threshold: when the conservative prediction margin is not lower than the first threshold, it enters the first defense mode; when the conservative prediction margin is lower than the first threshold but not lower than the second threshold, it enters the second defense mode; when the conservative prediction margin is lower than the second threshold, it enters the third defense mode.

[0093] In the first line of defense mode, the active power reserve, reactive power reserve, energy storage charge status and operation mode of the grid-type device are optimized according to the conservative forecast margin constraint, and preventive control commands are generated.

[0094] In the second line of defense mode, the active and reactive current tracking weights are determined based on frequency risk and voltage risk. The initial dq axis current command is projected and solved within the converter current circle constraint to generate the current command of the grid-type converter after current limiting. The energy storage energy and available active power are verified and anti-saturation processing is performed.

[0095] In the third line of defense mode, active disconnection and partitioning are carried out based on the grid capacity, dynamic reactive power, releaseable energy and active power balance constraints. Adaptive load reduction is calculated according to system inertia, power deficit and active power that the grid can support, and disconnection and load reduction instructions are generated.

[0096] The generated preventive control commands, converter current commands, disconnection commands, and load reduction commands are sent to grid-type devices, energy storage devices, and safety automatic devices for execution.

[0097] The following is a detailed implementation process of the present invention.

[0098] 1. System Structure, Research Object and Basic Assumptions

[0099] 1.1 Research Subjects

[0100] Considering regional power systems that include synchronous generator equivalents, grid-connected renewable energy (GFL), grid-connected renewable energy / grid-connected energy storage (GFM), integrated loads, and AC networks (see [link to relevant documentation]). Figure 2 The network-type device employs a virtual synchronous machine / drooping equivalent outer loop, and its fast internal voltage-current loop is equivalent to a first-order element in the electromechanical-control coupling model of this invention. For millisecond-level electromagnetic transient processes and power electronic device switching processes, the control logic of the method of this invention remains unchanged, but the relevant control parameters of the network-type device need to be further verified in the Electromagnetic Transients Model (EMT model).

[0101] The EMT model is a detailed simulation model used to describe the electromagnetic transient process of a power system. By establishing a fast dynamic model of the power electronic converter, grid components and control system, it reflects the voltage, current and controller response characteristics during fault disturbances and fast control processes.

[0102] The electromechanical-control coupling model used in this invention is mainly used to describe the dynamic processes such as system frequency, voltage, power angle, converter current constraint, and energy storage energy change, and is used for dynamic safety margin calculation, short-time prediction, and three-line defense coordinated control; the EMT model is used to verify the applicability of the above control parameters in the fast electromagnetic transient process before engineering implementation.

[0103] Among them, the parameters that need to be verified by the EMT model include the virtual inertia parameters, virtual damping parameters, voltage-reactive power control parameters, current limiting parameters, and energy storage support-related parameters of the grid-type device, in order to verify the effectiveness of the parameters in fault transients, current constraints, and rapid power regulation processes.

[0104] Therefore, the EMT model is an engineering verification and parameter calibration model for the control strategy of this invention, rather than a real-time calculation model used for online dynamic safety margin calculation and defense mode discrimination.

[0105] 1.2 Per-unit system and symbol conventions

[0106] The system uniformly adopts the three-phase capacity base value Line voltage effective value base value and rated angular frequency Unless otherwise specified, voltage, current, active power, and reactive power are expressed in per-unit values; frequency is expressed in Hz; time is expressed in s; stored energy can be expressed in MWh or converted to pu·s in the dynamic equation. It is stipulated that the active and reactive power injected into the AC system by a grid-connected power source is positive.

[0107] 2. Unified Dynamic Model of Grid-type New Energy

[0108] 2.1 Virtual Synchronous Power Angle-Frequency Model

[0109] The phase angle of a grid-type converter is obtained by integrating the internal angular frequency:

[0110] (1)

[0111] in, The rate of change of the work angle, The power angle of the grid-type voltage source relative to the system reference phase angle, in rad; The internal angular frequency of the network structure is rad / s; The rated angular frequency is rad / s.

[0112] Active power-frequency dynamics using virtual synchronization are

[0113] (2)

[0114] In the formula, The rate of change of internal angular frequency. The equivalent virtual inertia coefficient is pu·s² / rad; Given the active power, pu; The actual active power output on the AC side is pu; The virtual damping coefficient is expressed as pu·s / rad. If a virtual inertia constant is used... If the expression indicates that the two satisfy the condition, then the two conditions are met.

[0115] (3)

[0116] in, Let be the virtual inertia constant, s; Rated capacity for grid-type devices; This is the baseline value for system capacity.

[0117] 2.2 Voltage-Reactive Dynamics

[0118] The grid-type voltage amplitude adopts reactive power-voltage droop and considers the limited bandwidth of the controller:

[0119] (4)

[0120] in, The rate of change of the equivalent internal potential amplitude in a grid-type structure. pu represents the equivalent internal potential amplitude of the grid-type structure. This is the voltage reference value; This represents the actual reactive power output. Given reactive power; It is the reactive-voltage droop coefficient, with the dimension pu voltage / pu reactive, which can be written as pu / pu in the per-unit system; Let be the equivalent time constant of the outer voltage loop, in seconds.

[0121] 2.3 Network Power Transfer Model

[0122] The grid-type power supply is passed through the equivalent reactance Connected to voltage amplitude For a system bus with a phase angle of 0, neglecting resistance, we have:

[0123] (5)

[0124] (6)

[0125] in, pu; is the equivalent reactance between the strong voltage support point and the grid-connected power supply. pu represents the voltage amplitude at the grid connection point / equivalent grid voltage. It defines the maximum transmittable active power.

[0126] (7)

[0127] but The malfunction caused Reduced or network cut-off When it increases, It will suddenly drop, thus generating acceleration power. This is the basic mechanism by which grid-connected power supplies experience transient loss of synchronization.

[0128] 2.4 Converter Current Constraints and Feasible Power Domain

[0129] In voltage orientation In the coordinate system, neglecting filter loss, we have

[0130] (8)

[0131] in, , These represent the active and reactive axis currents, pu; under the notation conventions of this invention, the injected capacitive reactive power during voltage dips corresponds to... The converter must meet the upper limit of device current.

[0132] (9)

[0133] in, The short-time allowable current limit is pu. Therefore, given the grid connection point voltage... Below, the instantaneous power domain achievable by the grid-type device is

[0134] (10)

[0135] This formula illustrates that the deeper the fault, The lower, even If the active power sustaining frequency and reactive power boosting voltage need to be increased simultaneously during a fault, the two will inevitably compete for the same current circle radius.

[0136] 2.5 Active power reserve and energy constraint

[0137] The active power up-regulation of grid-connected wind, solar, or energy storage systems is constrained by available power and energy resources. Instantaneous active power reserve is defined.

[0138] (11)

[0139] in, This represents the maximum active power currently available. This represents the active power at the operating point before the disturbance. For energy storage, the SOC is dynamically written as...

[0140] (12)

[0141] in, The rate of change of the energy storage state of charge with respect to time. It is in a charged state; The per-unit value of active power for energy storage to discharge into the system; Rated energy, MWh; For discharge efficiency. Given the minimum permissible state of charge. It can release energy as

[0142] (13)

[0143] To ensure a period of support Internal power is sustainable and must meet the following requirements.

[0144] (14)

[0145] in, pu is the additional active power support relative to the operating point before the disturbance.

[0146] The stabilization mechanism of three-dimensional grid-type new energy participating in the three lines of defense

[0147] 3.1 Frequency Support Mechanism

[0148] The motion of the system's frequency center can be represented by an equivalent oscillation equation:

[0149] (15)

[0150] in, Let s be the system's equivalent inertia constant. For system frequency; This refers to the power imbalance, expressed in MW. If a power deficit occurs in the initial stage of an accident... Ignore governor lag. Definition Let RoCoF be the initial rate of change of frequency after the accident.

[0151] (16)

[0152] Networked devices, through virtual inertia and rapid energy storage power injection, effectively reduce the net power deficit after an accident, thus significantly lowering RoCoF. It is important to emphasize that virtual inertia itself is not an energy source; if... or However, the virtual inertia setting is only valid for a short time, subsequently causing a secondary frequency drop. Therefore, this invention incorporates both energy margin and frequency margin into a unified safety margin.

[0153] 3.2 Voltage Support Mechanism

[0154] In weak networks, the sensitivity of node voltage to reactive power is approximately positively correlated with the network reactance:

[0155] (17)

[0156] in, Thevenin equivalent reactance at the node; The voltage before the disturbance. The grid-type device can be quickly adjusted. But when In such cases, adding reactive power must sacrifice active power. When the fault is severe, a fixed "reactive power priority" may exacerbate the frequency deficit; a fixed "active power priority" may cause the voltage to remain below the safety margin for an extended period. Therefore, priority must be determined based on the current level of risk.

[0157] 3.3 Energy Mechanism of Transient Synchronous Stability

[0158] Assume the power angle curve after the fault can still be approximated as The stable equilibrium point and the unstable equilibrium point are respectively

[0159] (18)

[0160] when At that time, a potential energy function relative to the stable equilibrium point can be constructed.

[0161] (19)

[0162] The equivalent kinetic energy of the network type is

[0163] (20)

[0164] Critical energy barrier is

[0165] (twenty one)

[0166] If the current total transient energy

[0167] (twenty two)

[0168] satisfy In this simplified model, the trajectory remains within the attraction domain; if it approaches or exceeds the critical energy, it needs to be reduced. Increase damping or change the network structure to reduce acceleration energy. This energy perspective is more effective in reflecting the dynamic synchronization risk of network-type devices than simply monitoring the instantaneous power angle.

[0169] 3.4 Coupling Instability Mechanism Caused by Current Limiting

[0170] When there is no traffic restriction, and It can be adjusted separately through the frequency loop and voltage loop; after current limiting, the control input is compressed from the two-dimensional plane to the disk. If the original command current Located outside the circle, it must be projected onto the feasible region. Traditional radial limiting is equivalent to fixed holding. The proportion cannot be adjusted based on system frequency and voltage risk to change the support focus; fixing active power priority or reactive power priority is equivalent to permanently placing one stability risk above the other. Therefore, this invention directly maps "system risk" to projected cost weights.

[0171] 4. Dynamic Network Security Margin Construction

[0172] 4.1 Frequency safety margin

[0173] Let the lower and upper limits of the allowed frequency be respectively , The rated frequency is The frequency gain margin is defined as follows:

[0174] (twenty three)

[0175] in, Indicates that it is located at the rated frequency. This indicates that the frequency safety boundary has been reached. This indicates an out-of-bounds error. The RoCoF margin is then redefined.

[0176] (twenty four)

[0177] in, The maximum allowable rate of frequency change is expressed in Hz / s. To avoid situations where the frequency has not yet exceeded the limit but RoCoF has already reached an extreme level, the frequency dimension comprehensive margin is written as...

[0178] (25)

[0179] in For the smooth minimum operator defined later, This is the smoothing coefficient.

[0180] 4.2 Voltage safety margin

[0181] Assume the allowable voltage range is ,but

[0182] (26)

[0183] For multi-node systems, take the set of critical buses. The worst-case scenario:

[0184] (27)

[0185] in, This refers to the system voltage safety margin, which is the most unfavorable voltage margin among all critical buses. This represents the voltage safety margin of the kth critical bus.

[0186] 4.3 Synchronous stability margin

[0187] Based on the transient energy relationship in Section 3.3, the synchronization energy margin is defined.

[0188] (28)

[0189] in, To prevent small positive numbers with a denominator of zero. This indicates that an energy barrier still exists. Corresponding critical trajectory. This indicates that the reduced-order model has crossed the synchronous stability boundary.

[0190] 4.4 Current Margin

[0191] The current margin of a network-type device is defined as

[0192] (29)

[0193] If short-term heat accumulation is taken into account, the heat margin can be further defined.

[0194] (30)

[0195] in, This refers to the equivalent junction temperature or thermal state of power devices. To be the maximum allowed, This is the rated / reference thermal state. The reduced-order simulation of this invention does not explicitly simulate the thermal network, but it can be used in engineering applications. Replace simple .

[0196] 4.5 Energy and Active Reserve Margin

[0197] Given a prediction support window The predicted additional energy required is

[0198] (31)

[0199] Define energy margin and power reserve margin

[0200] (32)

[0201] in, It is a small positive number; The maximum additional active power required within the prediction window. The energy dimension comprehensive margin is defined as...

[0202] (33)

[0203] This allows us to distinguish between two physically completely different constraints: "having energy but insufficient power" and "having sufficient instantaneous power but insufficient energy".

[0204] 4.6 Smooth Minimum and Dynamic Network Safety Margin

[0205] right Individual normalized margin Define the weighted smoothing minimum operator.

[0206] (34)

[0207] in, and ; For smoothing parameters. When Sometimes,

[0208] (35)

[0209] Therefore, the current dynamic network security margin is defined.

[0210] (36)

[0211] This index has three characteristics: first, it is dimensionless, which can unify different physical quantities; second, it is dominated by the weakest stable link, avoiding the simple weighted average from masking the over-limit of a single item; and third, it is continuously differentiable, making it suitable for predictive control and gradient optimization.

[0212] 4.7 Short-term forecasting and uncertainty contraction

[0213] At the current sampling time t, the following operational information is acquired: power system frequency, key bus voltage, power angle and internal angular frequency of grid-type devices, equivalent internal electromotive force, energy storage state of charge, and active, reactive, and current data. A state vector is constructed using dynamic variables such as system frequency, power angle, voltage, and energy storage state of charge, and control inputs and external disturbances are incorporated into the reduced-order state equations as follows:

[0214] (37)

[0215] in, The system state vector at time t includes frequency, power angle of the network device, internal angular frequency, voltage, and energy storage state of charge. The first derivative of the state vector with respect to time; The control input vector includes control quantities such as network-type control commands, machine and load shedding, and network switching; The disturbance vector includes disturbances caused by faults, load changes, and available power from renewable energy sources. It is a reduced-order state change function composed of the dynamic equations of each subsystem.

[0216] Based on the current state and the rate of change of state, short-time second-order extrapolation is used within the short-time prediction window, resulting in:

[0217] (38)

[0218] in: From time Departure, Time The predicted state vector; The short-term forecast duration is (s). The first-order state derivative at time t; is the second-order state derivative at time t; coefficient 1 / 2 is the constant coefficient of the second-order Taylor extrapolation.

[0219] Based on formula (37), it can be further written as:

[0220]

[0221] The meaning of (t) can be found in the character definition below formula (37), where (t) represents the corresponding quantity at time t.

[0222] Substituting the predicted state into the same evaluation function as the current dynamic network security margin, we obtain the predicted security margin. , The predicted safety margin (dimensionless) before uncertainty contraction has been performed. For dynamic network security margin evaluation function; subscript This indicates a prediction. During the calculation, the margins for each dimension are calculated based on the frequency, voltage, synchronization, current, energy state, and required operating parameters corresponding to the prediction time, and then a fusion calculation with the same current safety margin is performed.

[0223] Considering state estimation error, available renewable energy prediction error, and model mismatch, a conservative prediction margin is defined.

[0224] (39)

[0225] in, The margin prediction standard deviation or the uncertainty scale obtained from sensitivity propagation; The confidence contraction coefficient. When the error is approximately Gaussian and At that time, it corresponds to a lower bound of approximately 95% one-sided confidence.

[0226] If first-order covariance propagation is used, let the state covariance be... ,but

[0227] (40)

[0228] in, This is an additional term for model uncertainty. To predict the standard deviation of the safety margin, To predict the variance of the safety margin, Let M be the column vector of the gradient of the safety margin with respect to the state vector x. Let be the state estimation error covariance matrix.

[0229] 5. Adaptive Cooperative Strategy for Three Lines of Defense Based on Security Margin Prediction (see...) Figure 1 )

[0230] 5.1 Defense Line Switching Criteria

[0231] Set the threshold for the first / second line of defense. Second / Third Line of Defense Boundary Threshold ,satisfy To prevent control mode chattering, a recovery threshold is also set. , and duration The pattern criterion is...

[0232] (41)

[0233] in, These represent the first, second, and third lines of defense, respectively. Only when the margin recovers to or above the corresponding recovery threshold and remains above it... Only then can the control mode allow retreat to lower-level defenses.

[0234] 5.2 First Line of Defense: Optimization of Network Support Capacity

[0235] The goal of the first line of defense is to maintain forecast margins by jointly optimizing active power reserve, dynamic reactive power, current margin, energy storage SOC, and network operation modes. That's all. Scrolling optimization can be implemented:

[0236] (42)

[0237] in, For the construction of a network-type device assembly, The current state of charge of the g-th energy storage unit is... This is the reference value for the state of charge of the g-th energy storage unit;

[0238] satisfy

[0239] (43)

[0240] (44)

[0241] (45)

[0242] in, As the first line of defense, control variables; , The first The active and reactive power support capacity reserved in the Taiwan-based grid device; For the first The Taiwan-based grid device is currently outputting no reactive power. For the first Dynamic reactive power limit of Taiwan-based grid devices; , , and As cost weight; These are soft-constraint slack variables; The recommended upper limit of current during normal operation is to provide a margin for emergency support.

[0243] From a mechanistic perspective, the first line of defense does not pursue maximum network parameters, but rather aims to "increase the margin of the weakest dimension." For example, when At the minimum, priority should be given to restoring SOC or increasing wind and solar load reduction reserves; when At a minimum, the dynamic reactive power capacity of key nodes and the capacity of grid voltage sources should be increased; when To minimize this, the active power of the cross section should be reduced, the equivalent reactance decreased, or the damping increased.

[0244] 5.3 Second Line of Defense: Risk-Weighted Current Circular Projection

[0245] 5.3.1 Unlimited support command

[0246] Constructing the active current command required for frequency support

[0247] (46)

[0248] in, The equivalent gain for primary frequency modulation / droop is pu / Hz; The gain for RoCoF is pu·s / Hz.

[0249] The reactive current command required for voltage support is

[0250] (47)

[0251] in, The reactive power gain is the voltage deviation. This is the gain for the rate of change of voltage; This is the reactive operating point before the disturbance.

[0252] 5.3.2 Risk Weights

[0253] Define frequency risk and voltage risk

[0254] (48)

[0255] in, , Let be the risk sensitivity coefficient. The higher the risk, the greater the penalty should be imposed on the current tracking error; therefore, let .

[0256] (49)

[0257] in, Based on weights, This is a risk amplification factor. Note. , It is not a fixed "active power priority / reactive power priority", but is determined by the real-time weak points of the system.

[0258] 5.3.3 Current Limiting Optimization Model

[0259] when When, solve

[0260] (50)

[0261] satisfy

[0262] (51)

[0263] Constructing the Lagrange function

[0264] (52)

[0265] in, It is a KKT multiplier. Based on the stationary point condition...

[0266] (53)

[0267] (54)

[0268] Analytical structure for obtaining optimal current

[0269] (55)

[0270] If the original command is located inside the current circle, then If it lies outside the circle, then the complementary relaxation condition is given.

[0271] (56)

[0272] The left side of the equation is about It is strictly monotonically decreasing, therefore it has a unique root, which can be found online using 2 to 5 binary search or Newton iterations. When, it degenerates into a traditional radial circular projection; when When, the algorithm automatically tends towards active power priority; when When the reactive power is prioritized, it automatically tends to prioritize reactive power. Therefore, RWCCP is a unified generalization of the three traditional strategies. RWCCP refers to Risk-Weighted Current-Circle Projection.

[0273] 5.3.4 Secondary Energy Constraints and Anti-Saturation

[0274] Calculated After that, active power reserve and energy constraints also need to be met:

[0275] (57)

[0276] When the aforementioned saturation occurs, the difference is fed back to the frequency outer loop integrator to implement inverse integral saturation.

[0277] (58)

[0278] in, for Rate of change over time For frequency control integral state; ; It is an anti-saturation gain. This design is used to suppress secondary impacts caused by excessive commands accumulated in the controller after the energy storage / wind / solar energy is depleted.

[0279] 5.4 The Third Line of Defense: Active Resolution of Network Sufficiency Constraints

[0280] when Furthermore, the second line of defense could not be completed within the stipulated time. When the internal safety margin is restored, the third line of defense is activated. Let the busbar set be... The candidate isolated network set is binary variables Indicates busbar Assigned to an isolated network Define the line. Variables of whether to remove The following proactive unblocking objectives can be established:

[0281] (59)

[0282] in, The cost of cutting off the line; For the first The individual network has an imbalance in its contribution; Let i be the load shedding amount for node i; , As weight; For the set of candidate resection routes; The unit cost for load shedding at node i. Each bus belongs to only one isolated network:

[0283] (60)

[0284] in, For 0-1 variables, This indicates that bus i is assigned to isolated network k. This is a set of candidate isolated network partitions.

[0285] Each isolated network must meet the minimum network capacity constraint.

[0286] (61)

[0287] in, The rated capacity of the g-th grid-connecting device, For 0-1 variables, This indicates that the g-th network-building device is assigned to the isolated network k. The minimum grid capacity required to maintain voltage / frequency for an isolated grid k For the construction of a network-type device assembly;

[0288] Dynamic reactive power support constraints

[0289] (62)

[0290] in, This is the dynamic non-functional capacity that the g-th network-building device can provide after being assigned to the isolated network k; if it is not assigned, this value is 0. This indicates the dynamic reactive power requirement of the isolated network k.

[0291] Energy support constraints

[0292] (63)

[0293] in, The g-th grid energy storage unit can release energy. The energy required to continuously support the isolated network;

[0294] and the work balance constraint after demultiplexing

[0295] (64)

[0296] in, The total active power generated within the isolated grid k. The total active power of the load within the isolated network k. For isolated networks, active power can be increased for backup. The function can be adjusted down within the isolated network k. This represents the maximum load that can be switched on an isolated network.

[0297] Therefore, the unblocking partitioning method of this method does not simply pursue the minimum power of the tie line, but first ensures that "each remaining isolated network can form and maintain its voltage and frequency on its own".

[0298] 5.5 Derivation of Minimum Adaptive Load Reduction

[0299] After disconnection or high-power deficit, set the frequency braking time. Maximum allowable frequency drop Integrating the frequency-equivalent oscillation equation, we have:

[0300] (65)

[0301] If the average grid support within the window is used Approximately, to ensure the frequency is not lower than Minimum required load reduction estimate

[0302] (66)

[0303] Compared with fixed-cycle low-frequency load shedding, this formula explicitly considers system inertia and current available active power support for the grid, which can avoid excessive load shedding when there is still a large amount of energy storage available, and also avoid insufficient load shedding when grid resources are close to being exhausted.

[0304] 5.6 Recovery and Black Boot

[0305] After the third line of defense is activated, priority should be given to establishing local bus voltage using grid-type energy storage with sufficient SOC, strong short-circuit support capability, and proximity to critical loads. The recommended restoration sequence is: establish voltage / frequency reference using grid-type energy storage → restore power to the power plant → gradually deploy wind and solar power → restore critical loads → start conventional power supply → expand the isolated grid → synchronize and connect. Before synchronizing two isolated grids, the following conditions should be met:

[0306] (67)

[0307] in, , , These are the allowed frequency difference, voltage difference, and phase angle difference for parallel operation, respectively.

[0308] 6. Algorithm Implementation Process and Tuning Method

[0309] 6.1 Online Algorithm Flow

[0310] Each control cycle performs the following calculations: state estimation → five-dimensional margin calculation → short-time prediction and uncertainty reduction → defense mode determination → corresponding control optimization → current / power / energy constraint verification → control issuance → closed-loop correction for the next cycle. The pseudocode is as follows:

[0311] Algorithm 1: Three-line defense coordinated control driven by dynamic network security margin

[0312] 1) Collection and the condition of key cross sections;

[0313] 2) Calculation and ;

[0314] 3) Prediction ,calculate ;

[0315] 4) If Implement the first line of defense for backup / reactive power / SOC optimization;

[0316] 5) If ,calculate Solve for RWCCP and perform fast support;

[0317] 6) If If the confirmation time is exceeded, proactive unblocking and adaptive load reduction will be performed;

[0318] 7) Inspection SOC, energy, and thermal constraints are used to implement anti-saturation.

[0319] 8) When the margin exceeds the recovery threshold and continues to do so They are then allowed to retreat to a lower-level defensive line.

[0320] 6.2 Key Parameter Tuning Principles

[0321] The margin should not be too small, otherwise the overall margin will degenerate into the average value; nor should it be too large, otherwise it will approach the hard minimum and increase measurement noise. In engineering practice, each margin can be initially set at a safe operating point of approximately 0.5 to 1, and then the minimum term can be adjusted accordingly. The contribution of over 70% was used for offline tuning. Prediction window The response time of the network's rapid control and safety automatic devices should be covered, but it should not be too long to avoid amplifying model errors; the electromechanical-control coordination layer can be on the order of 50~300ms, and a longer rolling window can be used for the station's energy control.

[0322] Threshold recommendations to meet . This corresponds to the boundary condition where "there are still controllable degrees of freedom, and discrete excision is not required"; This corresponds to the boundary where "continuous control can no longer guarantee safety, and discrete topology / load measures are required." Specific values ​​should be calibrated through N-1 / N-2 scans, severe fault sets, and hardware-in-the-loop testing, rather than being directly applied across systems.

[0323] 7. Simulation System and Scene Design

[0324] 7.1 Simulation Model Description

[0325] To verify the method's mechanism and relative effectiveness, this invention establishes a reduced-order electromechanical-control coupled time-domain model with an integration step size of 1 ms and a total simulation time of 6 s. The model explicitly includes synchronous equivalent inertia, GFM virtual synchronous dynamics, first-order voltage recovery, GFM active / reactive current limiting, releasable energy, and low-frequency load shedding logic; network faults are represented by piecewise changes in equivalent transmission capacity, voltage source strength, and power deficit. This example is used to verify the control mechanism and the relative performance of the strategy, but it cannot replace the hardware-in-the-loop verification required in actual engineering projects using PSCAD / EMTDC, RTDS, or manufacturer-manufactured controllers. Key parameters are shown in Table 1.

[0326] Table 1 Main parameters for reduced-order time-domain simulation

[0327]

[0328] 7.2 Comparison Strategy

[0329] Three strategies are set up: A) Traditional strategy, which only retains the conventional primary frequency modulation, fixed voltage recovery and fixed threshold low-frequency load reduction; B) Fixed parameter network construction strategy, where GFM provides fixed ratio frequency / voltage support and does not change the priority when current limiting; C) Adaptive strategy proposed in this invention, which enables dynamic network construction safety margin, predictive defense line switching, RWCCP current limiting allocation and third line of defense adaptive load reduction.

[0330] 7.3 Scene Setup

[0331] Scenario 1: Composite Disturbance Scenario. A critical line fault occurs and is disconnected in 1.0s, causing a sudden voltage drop and active power deficit; subsequently, the system maintains weak network operation to examine the frequency-voltage composite risk and current limiting distribution capability under the second line of defense.

[0332] Scenario 2: Extreme Multiple Fault Scenario. Building upon Scenario 1, this scenario further superimposes a larger power deficit and network strength degradation, rendering continuous control unable to fully restore the safety margin and triggering a third line of defense. This is used to verify the effectiveness of adaptive load shedding and backup control.

[0333] Evaluation metrics include the lowest frequency point Maximum absolute RoCoF, minimum voltage , GFM maximum current, final load shedding and minimum dynamic safety margin.

[0334] 8. Simulation Results and Analysis

[0335] 8.1 Frequency Response of Composite Disturbance

[0336] See Figure 3Simulation results show that the lowest frequency achieved by the traditional strategy is 48.800Hz; the fixed-parameter network strategy improves it to 49.033Hz; and the proposed adaptive strategy further improves it to 49.607Hz. This represents an improvement of 0.807Hz compared to the traditional strategy and 0.573Hz compared to the fixed-parameter network strategy. This indicates that when frequency risk becomes the main weak point, RWCCP will automatically increase. This allows for the allocation of more limited current to active power support, while energy constraints prevent unsustainable over-output.

[0337] 8.2 Combined Disturbance Voltage Response

[0338] See Figure 4 The lowest voltage point of the traditional strategy is 0.723 pu, the fixed-parameter network strategy is 0.808 pu, and the proposed strategy is 0.833 pu. This means the proposed method improves upon the traditional strategy by approximately 0.111 pu. The results demonstrate that risk weighting is not simply a matter of sacrificing reactive power for frequency; as voltage risk increases... Automatically raise, make By maintaining a sufficient proportion, both the frequency and the lowest voltage point can be improved simultaneously.

[0339] 8.3 Dynamic safety margin and defense triggering

[0340] Figure 5 The current safety margin is given. With forecast margin The changes are as follows. After a fault, multiple components of frequency, voltage, current, and energy decrease simultaneously. The smoothing minimum index is dominated by the weakest component and quickly enters the second line of defense region. Under extreme disturbances, the prediction margin enters the third line of defense region. The control logic no longer attempts to infinitely increase the continuous current support, but instead shifts to discrete safety measures such as load reduction / de-stationing. This design, which "acknowledges the boundaries of continuous control," can avoid the thermal risks, energy depletion, and control desynchronization caused by the GFM being stuck at the current limit for a long time.

[0341] 8.4 Extreme Multiple Faults and the Third Line of Defense

[0342] See Figure 6 In extreme scenarios, the lowest frequencies for the traditional strategy, the fixed-parameter network strategy, and the proposed strategy are 48.277Hz, 48.175Hz, and 48.666Hz, respectively. The fixed-parameter network strategy is slightly lower than the traditional strategy in this scenario because its current limiting and energy management do not change with risk, indicating that "configuring GFM" does not necessarily mean "more stable under all severe faults." The proposed strategy, through the connection of risk-weighted support in the second line of defense and load reduction in the third line of defense, increases the lowest frequency by 0.389Hz (relative to the traditional strategy) and 0.490Hz (relative to the fixed network).

[0343] See Figure 7 The traditional strategy resulted in a load reduction of 0.28 pu, the fixed network structure strategy a load reduction of 0.20 pu, and the proposed strategy a load reduction of 0.17 pu. This represents a 39.3% reduction in load reduction compared to the traditional strategy and a 15.0% reduction compared to the fixed network structure strategy. This demonstrates that the aforementioned minimum load reduction concept based on inertia and actual network support capacity can reduce the conservatism of fixed-cycle load reduction.

[0344] 8.5 Summary Indicators

[0345] Table 2 Main simulation indicators for two types of fault scenarios

[0346]

[0347] It should be noted that the maximum RoCoF of the three strategies at the moment of failure is mainly determined by the initial power deficit and equivalent inertia of the fault. Therefore, the maximum instantaneous RoCoF in Table 2 are similar. The advantage of grid-based control is more reflected in the continuous power support and improvement of the minimum frequency point for tens to hundreds of milliseconds after the fault. The maximum current of the proposed method in the extreme fault scenario reaches 1.20 pu, which just touches the set limit, indicating that the control is fully utilized but does not exceed the physical constraints of the converter.

[0348] 8.6 Current Upper Limit Sensitivity

[0349] See Figure 8 Sensitivity results indicate that only improving Improving all performance metrics is not necessarily a one-size-fits-all approach. As the current limit increases, the controller may bear more of the active power load, leading to a change in the reactive power ratio; therefore, the final voltage minimum depends on risk weighting, network strength, and active-reactive coupling. In engineering, "increasing the overcurrent factor" should not be the sole means of improving stability; instead, a combined optimization approach should be adopted. , Virtual impedance, energy reserve, and defense threshold.

[0350] 8.7 Energy Capacity Sensitivity

[0351] See Figure 9 As the amount of releasable energy increases, grid-type energy storage can maintain additional active power support for a longer period, reducing the load shear required for the third line of defense. However, once the energy exceeds the support window required for an accident, the marginal benefit tends to saturate. This illustrates that grid-type energy storage capacity should be described using both "power" and "energy," and simply providing a virtual inertia constant, rated MW, or overcurrent multiple is insufficient to characterize its true stable support capacity.

[0352] 9 Project Implementation

[0353] First, during the planning phase, a regional network capacity ledger is established, incorporating the rated capacity, short-time current limit, virtual inertia range, dynamic non-functional capacity, active power reserve, energy storage SOC, and continuous support time of each GFM power station into the scheduling model. Second, during the operation phase, the scheduling master station calculates system-level... and The station-side calculates local margins and executes millisecond-level RWCCP, forming a hierarchical architecture of "master station assesses risk—station executes quickly." Furthermore, protection and stability control must consider the characteristics of GFM fault current limitation and phase control, avoiding settings based on the assumption of large short-circuit current of synchronous machines. Finally, before project commissioning, at least four levels of verification must be completed: offline EMT batch fault scanning, controller hardware-in-the-loop testing, station-level closed-loop testing, and regional stability control joint commissioning.

[0354] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. A three-line adaptive collaborative prevention and control method for a grid-type new energy power system, characterized in that, include: Acquire the frequency, rate of change of frequency, key bus voltage, power angle of grid-connected devices, internal angular frequency, dq axis current, upper limit of converter current, state of charge of energy storage and available active power data of the power system including synchronous generators, grid-connected new energy, grid-connected new energy / grid-connected energy storage and AC network during operation. Calculate the frequency amplitude margin and frequency change rate margin based on the obtained frequency and frequency change rate, respectively; calculate the voltage margin based on the critical bus voltage; calculate the synchronization stability margin based on the power angle, kinetic energy and critical energy of the grid-type device; calculate the current margin based on the dq axis current and the upper limit of the converter current; and calculate the energy margin based on the available energy of energy storage and active reserve. The frequency amplitude margin, frequency change rate margin, voltage margin, synchronization stability margin, current margin, and energy margin are fused using the smoothing minimum operator to obtain the current dynamic network safety margin. Based on the system state, short-time prediction is used to calculate the prediction safety margin within the future prediction window. The prediction safety margin is then conservatively reduced by combining the state estimation error, available power prediction error, or model mismatch uncertainty to obtain the conservative prediction margin. The conservative prediction margin is used to determine the defense mode of the line of defense based on the first threshold, the second threshold and the hysteresis recovery threshold: when the conservative prediction margin is not lower than the first threshold, it enters the first defense mode; when the conservative prediction margin is lower than the first threshold but not lower than the second threshold, it enters the second defense mode; when the conservative prediction margin is lower than the second threshold, it enters the third defense mode. In the first line of defense mode, the active power reserve, reactive power reserve, energy storage charge status and operation mode of the grid-type device are optimized according to the conservative forecast margin constraint, and preventive control commands are generated. In the second line of defense mode, the active and reactive current tracking weights are determined based on frequency risk and voltage risk. The initial dq axis current command is projected and solved within the converter current circle constraint to generate the current command of the grid-type converter after current limiting. The energy storage energy and available active power are verified and anti-saturation processing is performed. In the third line of defense mode, active disconnection and partitioning are carried out based on the grid capacity, dynamic reactive power, releaseable energy and active power balance constraints. Adaptive load reduction is calculated according to system inertia, power deficit and active power that the grid can support, and disconnection and load reduction instructions are generated. The generated preventive control commands, converter current commands, disconnection commands, and load reduction commands are sent to grid-type devices, energy storage devices, and safety automatic devices for execution.

2. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The frequency amplitude margin is calculated using the following formula: The frequency change rate margin is calculated using the following formula: Using smooth minimum operator pairs and The fusion is used to obtain a comprehensive frequency margin, which is used to determine the current dynamic network security margin and subsequent defense mode. For a single busbar, voltage margin The minimum voltage margin of each bus in the set of critical buses is taken as the system voltage margin. in, For system frequency, For the rated frequency, , For frequency safety lower and upper limits, The rate of change of frequency, To allow the maximum rate of change of frequency, For the critical bus voltage, These are the lower and upper limits for voltage safety.

3. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The synchronization stability margin is calculated based on the potential energy, kinetic energy, and critical energy of the unstable equilibrium point of the network-type device, and is taken as follows: in, The current transient total energy is obtained by adding the kinetic and potential energy of the network-type device. The critical energy barrier corresponding to the unstable equilibrium point. To prevent zero decimals; The current margin is calculated using the following formula: in, These are the active shaft and reactive shaft currents of the grid-type converter, respectively. This is the upper limit of the short-time current of the converter; The energy margin is calculated based on the ratio of the energy required by the prediction window to the energy that can be released from the storage, and the ratio of active power reserve to the maximum additional active power required by the prediction. The results are then combined into a comprehensive energy margin using a smoothing minimum operator.

4. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The current dynamic network security margin adopts a weighted smooth minimum operator: in, For comprehensive frequency margin, For system voltage margin, To ensure synchronous stability margin, For current margin, To consider the overall energy margin, the weighted smoothing minimum operator is used. for: Where n represents the number of margin components, in this case n=5. It is any margin component to be merged. The normalized weights for each margin component, and ; For smoothing parameters; The predicted safety margin is calculated by substituting the future state obtained from the system state equation or second-order extrapolation into the same margin function. The conservative predicted margin... , To predict margin, Determined by state covariance through margin gradient propagation and / or model uncertainty. is the confidence contraction coefficient.

5. The three-line adaptive and coordinated prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The defense mode determination is set with a first threshold. Second threshold Hysteresis recovery threshold and continuous confirmation time Hysteresis recovery threshold includes first backoff threshold. Second backoff threshold ,and , , ; When the conservative forecast margin is no less than Maintain or enter the first line of defense mode at all times; When the conservative forecast margin is lower than and not less than It then enters the second line of defense mode; When the conservative forecast margin is lower than Enter the third line of defense mode; When in the second line of defense mode and the conservative forecast margin recovers to The above and continue for no less than When the time comes, revert to the first line of defense mode; when in the third line of defense mode and the conservative forecast margin recovers to The above and continue for no less than At that time, retreat to the second line of defense mode.

6. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The preventative optimization under the first line of defense mode constructs an objective function based on the active power reserve, reactive power reserve, energy storage state of charge deviation from the reference value, and slack variables of the grid-type device, and satisfies: The conservative forecast margin plus slack variables should not be lower than the first threshold. The active power reserve of each network-type device shall not exceed its maximum adjustable reserve, and the sum of reactive power reserve and current reactive power output shall not exceed the dynamic reactive power limit. The energy storage state of charge is maintained above the minimum allowable value, the normal operating current command of the grid-type device does not exceed the reserved limit current, and the reserved limit current is less than the short-time current limit of the converter. Adjust the active power reserve, reactive power reserve, energy storage charging / discharging status, or network operation mode according to the current minimum dimensional margin.

7. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, In the second line of defense mode, the initial active current command is generated based on the frequency difference and / or the rate of frequency change, and the initial reactive current command is generated based on the voltage deviation and / or the rate of voltage change; when the magnitude of the initial dq-axis current command exceeds the short-time current limit of the converter... At that time, risk based on frequency and voltage risk Determine the weights: in, For comprehensive frequency margin, For voltage margin, , For risk sensitivity coefficient, Based on weights, This is a risk amplification factor. These are the active axis current tracking error weights and the reactive axis current tracking error weights, respectively. by , Construct a weighted objective for dq-axis current tracking error, in The constraint-based current-limiting current calculation command is as follows: These are the initial active current command and the initial reactive current command, respectively.

8. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 7, characterized in that, The current-limiting command is used to construct the Lagrange function: From the partial derivative stationary point condition, we obtain: in, These are the active shaft and reactive shaft currents of the grid-type converter, respectively. For Lagrange multipliers; When the initial command is located within the current circle When the initial command is outside the current circle, press to make the reconstructed current satisfy... non-negative Solve this problem; The obtained optimal active axis current Optimal reactive axis current The active and reactive current commands are converted into those of the grid-type converter, and saturation is limited according to the available active power, the energy stored and released, and the predetermined support time. When saturation occurs, the difference between the actual support and the commanded support is fed back to the frequency outer loop integral state for inverse integral saturation.

9. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The active disconnection under the third line of defense mode constructs an optimization problem using busbar partitioning variables, line disconnection variables, and network-type device attribution variables, and satisfies: Each bus belongs to only one isolated network partition; The total rated capacity of network-type devices deployed in each isolated network partition shall not be less than the minimum network capacity requirement of that partition. The total dynamic reactive power of the network-type devices in each isolated network partition shall not be less than the dynamic reactive power demand of that partition. The total energy that can be released by grid-type energy storage in each isolated grid zone is no less than the energy demand that the zone can continuously support. The active power generated, active power of the load, active power reserve of the grid, and the remaining load after load shedding in each isolated grid zone meet the active power balance constraints. Based on the cost of line disconnection, the active power imbalance and load reduction of each zone, construct the target and output the disconnection scheme and load shedding scheme.

10. The three-line adaptive collaborative prevention and control method for grid-type new energy power systems according to claim 1, characterized in that, The adaptive load reduction is calculated based on the system's equivalent inertia, system capacity baseline, rated frequency, allowable frequency drop, predicted power deficit, and average active power support of the grid. in, The shortfall was due to meritorious service in the accident. To predict the average active power support of the internal structure of the mesh-type device, Let be the system's equivalent inertia constant. This is the baseline value for system capacity. To allow frequency drops, For the rated frequency, For the third line of defense frequency targets, For frequency braking time window; Will As the minimum load reduction command, it is issued to the load reduction device; before disconnection or reconnection, parallel verification is performed to ensure that the frequency difference, voltage difference and power angle difference between islanded grids do not exceed the allowable values ​​for parallel connection during the same period.