Distributed collaborative inertia and primary frequency modulation control system and method for new energy station

Abstract

The application discloses a new energy station distributed collaborative inertia and primary frequency modulation control system, comprising a plurality of control subsystems and a communication subsystem; the control subsystem comprises: a frequency measurement unit, which measures the frequency change rate, frequency deviation and active power of the new energy station in real time; an inertia estimation unit, which estimates the equivalent inertia of the system in real time or under specific conditions; a distributed collaborative control unit, which obtains the state information of adjacent new energy stations and collaboratively calculates the frequency modulation power that each new energy station should bear based on a consistency iteration algorithm; an adaptive virtual inertia control unit, which dynamically adjusts the virtual inertia coefficient according to the frequency change rate; and a power regulation execution unit, which adjusts the active output of the new energy station based on the equipment state of the new energy station. The new energy station distributed inertia and primary frequency modulation control system without centralized scheduling of the main station and with adaptive collaborative capability realizes fast and coordinated frequency support of multiple stations and improves the frequency stability of the high-proportion new energy power grid.

Description

Technical Field

[0001] This invention relates to the field of distributed new energy power station inertia and primary frequency regulation control technology, specifically to a distributed cooperative inertia and primary frequency regulation control system and method for new energy power stations. Background Technology

[0002] Frequency stability in the power system is crucial for ensuring the safe operation of the power grid. Traditional power grids rely on the rotating inertia of synchronous generators and governors to achieve primary frequency regulation. New energy sources such as wind and solar power connect to the grid through power electronic devices, lacking natural inertia and typically operating in maximum power point tracking (MPPT) mode. New energy power plants generally lack active frequency support capabilities and cannot actively participate in frequency regulation, leading to decreased system inertia and insufficient primary frequency regulation capacity. As the penetration rate of new energy continues to increase, the system's equivalent inertia continues to decline, significantly increasing the risk of frequency instability and easily triggering power grid frequency stability problems.

[0003] To address the above problems, existing technologies have adopted the following solutions:

[0004] 1. New energy power plants reserve spare capacity to participate in frequency regulation, but this will result in a loss of power generation revenue;

[0005] 2. Configuring an energy storage system to provide frequency regulation services, but the cost is high;

[0006] 3. Virtual synchronous machine technology is used to simulate the characteristics of a synchronous machine in the inverter, but it cannot adapt to changes in the operating state of the power grid;

[0007] In addition, the above solutions also have the following problems:

[0008] 1. The regulation capacity of a single power station is limited, making it difficult to cope with large system disturbances;

[0009] 2. Lack of coordination can easily lead to regulatory conflicts or over-adjustment, causing a second drop in frequency;

[0010] 3. Relies on high-speed communication and master station scheduling, resulting in large response delays;

[0011] 4. Independent regulation at each station lacks coordination between stations, resulting in limited regulation effectiveness and the potential for oscillations. Summary of the Invention

[0012] In view of the above-mentioned deficiencies of the prior art, the purpose of this invention is to provide a distributed collaborative inertia and primary frequency regulation control system and method for new energy power plants. This system, which does not require centralized scheduling by the master station and has adaptive collaborative capabilities, enables rapid and coordinated frequency support for multiple power plants and improves the frequency stability of high-proportion new energy power grids.

[0013] The objective of this invention is achieved through the following technical solution:

[0014] The distributed cooperative inertia and primary frequency regulation control system for new energy power plants includes:

[0015] Several control subsystems are installed at the grid connection points of each new energy power station;

[0016] A communication subsystem that enables data interaction between several of the aforementioned control subsystems;

[0017] The control subsystem includes:

[0018] The frequency measurement unit measures the frequency change rate, frequency deviation, and active power output of the new energy power station in real time.

[0019] Inertia estimation unit, which estimates the equivalent inertia of the system in real time or under specific conditions.

[0020] The distributed collaborative control unit acquires the status information of adjacent new energy power stations and collaboratively calculates the frequency regulation power that each new energy power station should undertake based on a consensus iterative algorithm.

[0021] The adaptive virtual inertia control unit dynamically adjusts the virtual inertia coefficient according to the frequency change rate.

[0022] The power regulation execution unit adjusts the active power output of the new energy power station based on the status of the new energy power station equipment.

[0023] Furthermore, the inertia estimation unit uses the perturbation-based RoCoF method of the rocking equation and the recursive least squares system identification method to estimate the equivalent inertia of the system.

[0024] The equivalent inertia of the system is estimated using the RoCoF method after perturbation of the swing equation;

[0025] The equivalent inertia of the system obtained by RoCoF method after the perturbation of the swing equation is corrected by the recursive least squares system identification method;

[0026] The specific condition for the inertia estimation unit to estimate the equivalent inertia of the system is that the power deficit is greater than a predetermined value. The RoCoF method after the perturbation of the swing equation is used to quickly estimate the equivalent inertia of the system.

[0027] Furthermore, the frequency modulation power sharing ratio is selected as the consistency variable in the consensus iteration algorithm; the convergence of the consensus iteration algorithm adopts a dual convergence criterion of the state change of each new energy power station and the deviation of the state of each new energy power station from the mean.

[0028] Furthermore, the virtual inertia coefficients adjusted by the adaptive virtual inertia control unit include: the virtual inertia time constant and the frequency modulation coefficient.

[0029] Furthermore, the adaptive virtual inertia control unit uses a fuzzy RBF neural network adaptive algorithm to tune the time-varying virtual inertia and the primary frequency modulation coefficient.

[0030] Furthermore, the fuzzy RBF neural network adaptive algorithm is as follows:

[0031]

[0032] in, and For time-varying virtual inertia and primary frequency modulation coefficient.

[0033] Furthermore, the input layer of the fuzzy RBF neural network adaptive algorithm is:

[0034]

[0035] in, For frequency deviation, The rate of change of frequency, Efforts are being made to transform and support new energy power plants;

[0036] The hidden layer of the fuzzy RBF neural network adaptive algorithm uses radial basis function activation, as follows:

[0037]

[0038] in, As the center vector, Width;

[0039] The output layer of the fuzzy RBF neural network adaptive algorithm is:

[0040] ,

[0041] in, and The output weights of the RBF network are M, where M is the number of hidden layer nodes.

[0042] The fuzzy RBF neural network adaptive algorithm uses fuzzy rules to adaptively adjust the RBF weights, and the weight update law is as follows:

[0043]

[0044] in, For performance indicators, The learning rate (recommended range: 0.01~0.05) This is the momentum factor.

[0045] Furthermore, the equipment used by the power regulation execution unit to regulate the active power output of the new energy power station includes at least one of the following: wind turbine generators, energy storage systems, and photovoltaic power stations;

[0046] When the system faces short-term power surge demand, the power regulation execution unit regulates the energy storage system to undertake inertial response and support the initial stage of primary frequency regulation;

[0047] When the system faces the need to provide continuous power support, the power regulation execution unit adjusts the inertia support provided by the new energy power station.

[0048] When the system faces the need to provide continuous power support, the power regulation execution unit adjusts the load reduction of the new energy power station to provide support in the initial stage of primary frequency regulation.

[0049] A method for distributed cooperative inertia and primary frequency regulation control of new energy power plants includes the following steps:

[0050] Step S1: Check if the system frequency exceeds the limit. If the system frequency exceeds the limit, proceed to step S2; otherwise, proceed to step S11.

[0051] Step S2: Each control subsystem collects the frequency change rate, frequency deviation, and active power output of the new energy power stations it is connected to;

[0052] Step S3: Each control subsystem uses the RoCoF method after perturbation of the swing equation and the recursive least squares system identification method to estimate the equivalent inertia of the system.

[0053] Step S4: Based on the consensus iterative algorithm, calculate the final estimated system equivalent inertia by weighting the estimated system equivalent inertia of each control subsystem according to its rated capacity ratio;

[0054] Step S5: Each control subsystem collects information on the frequency regulation sharing ratio, up-regulation capacity, and down-regulation capacity status of adjacent renewable energy power stations;

[0055] Step S6: Calculate the frequency regulation power that each new energy power station should undertake based on the consensus iterative algorithm;

[0056] Step S7: Based on the fuzzy RBF neural network, calculate the virtual inertia coefficient and the primary frequency modulation coefficient according to the current frequency deviation, frequency change rate, and adjustable capacity.

[0057] Step S8: Generate frequency modulation power command;

[0058] Step S9: Control each device to execute according to the priority of the total frequency modulation command, and then check whether the system frequency exceeds the limit. If the system frequency exceeds the limit, proceed to step S10; otherwise, proceed to step S11.

[0059] Step S10: After adjusting the adjustment weights, proceed to step S6;

[0060] Step S11: Does the periodic system frequency exceed the limit? If the system frequency exceeds the limit, proceed to step S1; otherwise, proceed to step S12.

[0061] Step S12: Periodically assess the system status;

[0062] Step S13: Determine whether parameter readjustment is required. If so, proceed to step S14; otherwise, proceed to step S11.

[0063] Step S14: Online parameter learning updates the control model, proceed to step S11.

[0064] Furthermore, in step S9, the priorities are, in order, energy storage system, wind turbine, and photovoltaic power station.

[0065] Because of the adoption of the above technical solution, the present invention has the following advantages:

[0066] 1. This invention introduces a distributed collaborative control unit, which dynamically allocates frequency modulation power among adjacent power stations through a consensus iterative algorithm, aggregating the regulation capabilities of multiple power stations. This significantly improves the overall frequency modulation capacity. More importantly, the algorithm avoids regulation conflicts or over-adjustment problems caused by independent decision-making among multiple power stations, fundamentally suppressing the risk of secondary frequency drops.

[0067] 2. This invention is equipped with an inertia estimation unit (for real-time / condition-specific estimation of system equivalent inertia) and an adaptive virtual inertia control unit (for dynamically adjusting coefficients based on the frequency change rate RoCoF), enabling the control system to sense the power grid status. The system can optimize control intensity online based on the current inertia level and the rate of frequency degradation.

[0068] 3. This invention adopts a distributed architecture for low-speed state synchronization between stations, requiring only the exchange of a small amount of state information, without the need to upload massive amounts of data to the main station for centralized processing. This decentralized model pushes decision-making down to the local level, greatly shortening the control command path and fundamentally solving the latency problem caused by reliance on high-speed communication, while also reducing investment in infrastructure such as fiber optic communication.

[0069] 4. This invention utilizes power sharing among multiple power plants to dynamically distribute the total reserve capacity requirement across multiple power plants according to equipment status, avoiding large-scale power curtailment at a single power plant. Simultaneously, the consensus iterative algorithm prioritizes power plants with lower regulation costs, thereby improving the overall utilization rate of renewable energy while ensuring frequency regulation effectiveness, achieving an optimal trade-off between safety and economy.

[0070] 5. This invention adopts a modular design for the control subsystem, with independent yet logically unified control for each station. New stations only need to install the corresponding subsystem to quickly connect to the collaborative network (plug and play). Furthermore, this solution is explicitly compatible with existing AGC and virtual synchronizer technologies, allowing for smooth overlay and upgrades on existing new energy station automation systems, making it highly valuable for engineering promotion.

[0071] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0072] Figure 1 This is a structural diagram of the distributed collaborative inertia and primary frequency regulation control system for new energy power stations in the embodiment;

[0073] Figure 2 This is a control flowchart of the distributed collaborative inertia and primary frequency regulation control system for new energy power stations in the embodiment. Detailed Implementation

[0074] The present invention will be further described below with reference to the embodiments.

[0075] Example:

[0076] like Figure 1 As shown, the distributed cooperative inertia and primary frequency regulation control system for new energy power stations includes:

[0077] Several control subsystems are installed at the grid connection points of each new energy power station;

[0078] A communication subsystem that enables data interaction between several of the aforementioned control subsystems;

[0079] The control subsystem includes:

[0080] The frequency measurement unit measures the frequency change rate, frequency deviation, and active power output of the new energy power station in real time.

[0081] Inertia estimation unit, which estimates the equivalent inertia of the system in real time or under specific conditions.

[0082] The distributed collaborative control unit acquires the status information of adjacent new energy power stations and collaboratively calculates the frequency regulation power that each new energy power station should undertake based on a consensus iterative algorithm.

[0083] The adaptive virtual inertia control unit dynamically adjusts the virtual inertia coefficient according to the frequency change rate.

[0084] The power regulation execution unit adjusts the active power output of the new energy power station based on the status of the new energy power station equipment.

[0085] In this embodiment, the frequency measurement unit and the inertia estimation unit are integrated and installed at the grid connection point of each new energy power station. The hardware consists of a PMU frequency measurement module (including a GPS synchronization clock) and a data acquisition unit.

[0086] Input signals: Three-phase voltage at grid connection point (connected via voltage transformer PT, rated value 100V / √3), three-phase current (connected via current transformer CT, rated value 5A).

[0087] Internal function: Real-time frequency measurement Rate of change of frequency The station has made meritorious contributions. Online estimation of system equivalent inertia based on RoCoF method .

[0088] Output signal:

[0089] Frequency deviation, in Hz, update cycle 10ms;

[0090] : Frequency change rate, unit Hz / s, update period 10ms;

[0091] : Power output, unit MW, update cycle 10ms;

[0092] Inertia estimate, in seconds, update period 100ms;

[0093] Output interface: Uploaded to the distributed collaborative control module via optical fiber;

[0094] Signal accuracy: frequency ±0.005Hz, ±0.02Hz / s, active power ±0.5%.

[0095] A hybrid estimation scheme is adopted, which mainly uses the perturbation RoCoF method based on the swing equation and supplements it with the recursive least squares (RLS) system identification method.

[0096] The core formula of the RoCoF method is:

[0097]

[0098] This is the estimated equivalent inertia of the system, in seconds (s).

[0099] The sum of power increments for all participating frequency regulation stations, in megawatts (MW).

[0100] This is an estimate of load fluctuations, in megawatts (MW).

[0101] The frequency change rate is expressed in Hertz per second (Hz / s).

[0102] The innovation of this invention lies in fusing the inertia estimation results across various field stations using a consensus algorithm, thereby enabling each field station to obtain a unified system inertia estimate. This avoids discrepancies in frequency modulation actions caused by inconsistencies in the estimation results of different stations.

[0103] In this embodiment, the distributed collaborative control unit is installed on the local controller of each site (embedded industrial computer, such as ARM Cortex-M7 or equivalent processor).

[0104] Input signal (from Unit 1): , , (Update cycle 10ms).

[0105] Input signals (from the communication network): Other station status information is as follows:

[0106] Frequency modulation sharing ratio, per-unit value;

[0107] Increased capacity, unit: MW;

[0108] Reduced capacity, unit: MW.

[0109] Update cycle: 100ms.

[0110] Internal function: Runs a consensus iterative algorithm to calculate the local frequency modulation power sharing ratio. ; Calculate the total frequency regulation power that this station should bear. .

[0111] Output signal: (Target frequency modulation power, unit MW), update cycle 100ms.

[0112] Output interface: Sends data to the adaptive virtual inertia control algorithm module.

[0113] In this embodiment, the adaptive virtual inertia control unit is installed on the local controller of each site (with the same hardware as module two).

[0114] Input signal: (From Module 2) , (From Module 1) (Estimated inertia).

[0115] Internal function: Online tuning based on fuzzy RBF neural network and ; Calculate the virtual inertia response power and primary frequency modulation power The two are added together to obtain the total power command. .

[0116] Output signal: (Total power command, unit MW), update cycle 10ms (fast channel).

[0117] Output interface: Power regulation actuator is sent via industrial Ethernet.

[0118] In this embodiment, the power regulation execution unit is installed in the local controller (inverter controller, wind turbine main controller, energy storage EMS) of each power generation unit in each station.

[0119] Input signal: (From Module 3)

[0120] Internal functions: According to The symbol and amplitude of the signal, combined with the status of each device in the station (wind turbine speed, photovoltaic MPPT power, energy storage SOC), are used to allocate power commands to each device according to the priority of "energy storage takes the lead in short-term impact, and wind / solar takes the lead in continuous support".

[0121] Output signal: Target power command for each device (unit: MW), update cycle: 10ms;

[0122] Feedback interface: Actual adjustable capacity, (Increased capacity, unit MW) and (Reduced capacity, unit MW), update cycle 100ms, feedback to distributed collaborative control module.

[0123] In another embodiment, the inertia estimation unit uses a fusion of the perturbation-based RoCoF method of the rocking equation and the recursive least squares system identification method to estimate the equivalent inertia of the system; specifically:

[0124] 1. The perturbation-based RoCoF method based on the swing equation is used as the main method, supplemented by the recursive least squares (RLS) system identification method, and the two methods are integrated for estimation.

[0125] Main scheme: RoCoF method

[0126] Based on the swing equation of synchronous generator:

[0127]

[0128] in, For the system's equivalent inertia, RoCoF is the rate of change of frequency. This is the equivalent power deficit.

[0129] In actual calculations, the difference form is used:

[0130]

[0131] in, This is the sum of the power increments of all participating frequency modulation stations. This is an estimated value for load fluctuation. This represents the rate of change of frequency.

[0132] Auxiliary solution: RLS system identification method (real-time correction of the main solution results)

[0133] Discretized system model:

[0134] Where D is the damping coefficient, Pm is the mechanical power, Pe is the electromagnetic power, and Ts is the sampling period (10ms).

[0135] RLS estimation form:

[0136]

[0137] in ,

[0138] RLS update formula:

[0139]

[0140]

[0141]

[0142] in, It is the forgetting factor (typical value 0.98~0.995). For the gain vector, Covariance matrix

[0143] 2. The selection rules for the applicability of the main scheme and auxiliary scheme under different frequency practices are as follows:

[0144] Large disturbances (system trip, load trip) RoCoF method (main) optimal ±10% (typical value) Minor disturbances (small fluctuations in load) RLS identification method Applicable ±20% Continuous fluctuations (wind power fluctuations) RLS identification method Applicable ±25%

[0145] 3. The estimated triggering conditions are as follows;

[0146] Large disturbance trigger Upon detection of a significant power deficit, a single rapid estimation (using the RoCoF method) is immediately triggered. Continuous rolling estimation Normal operation RLS online identification is performed every 100ms (using the RLS method) to track slow changes in inertia.

[0147] 4. The estimated data to be measured and the methods for obtaining it are as follows:

[0148] Grid connection frequency Local PMU or frequency measurement module ±0.005Hz Rate of change of frequency Frequency differentiation or phase-locked loop estimation ±0.02Hz / s The station has made contributions Substation electricity meters or inverters report ±0.5% Connector power PMU at the beginning of the line ±0.5%

[0149] 5. This system estimates the overall equivalent inertia of the system, rather than the local inertia of a single site.

[0150] Consolidation processing for independent estimates from each site: After each site independently performs estimation using the same algorithm (RoCoF and RLS fusion), the estimation results are fused using a consensus algorithm.

[0151]

[0152] in This is the fusion step size for inertia estimation (set to 0.2). When the difference in estimation results between different stations is less than the convergence threshold... After 0.5s, the weighted average is used as the estimate of the overall system inertia. The weights are determined based on the proportion of each station's rated capacity. Ultimately, all stations achieve a consistent weighting. Estimated value.

[0153] In another embodiment, the frequency modulation power sharing ratio is selected as the consistency variable in the consensus iteration algorithm; the convergence of the consensus iteration algorithm adopts a dual convergence criterion of the state change of each new energy power station and the deviation of the state of each new energy power station from the mean, as follows:

[0154] 1. Select the frequency modulation power sharing ratio as a consistency variable:

[0155]

[0156] in:

[0157] The actual primary frequency regulation power undertaken by station i is expressed in megawatts (MW).

[0158] The rated capacity of station i is expressed in megawatts (MW).

[0159] This represents the frequency modulation power sharing ratio, which is a per-unit value and dimensionless.

[0160] Physical meaning: This represents the share of frequency regulation power undertaken by station i according to its rated capacity. Consistency iteration ensures that each station... Converging to the same value This aims to achieve the goal of fairly distributing frequency modulation power according to capacity ratio.

[0161] 2. The discrete-time consistency update equation of the consistency iterative algorithm is as follows:

[0162]

[0163] in:

[0164] k is the number of iterations, which is an integer and increments from 0.

[0165] The uniform step size (coupling gain) is dimensionless, and the recommended value range is 0.15 to 0.30.

[0166] The communication weight is dimensionless and satisfies... The present invention takes uniform weights ,in The number of elements in the neighbor set;

[0167] For the set of neighbors of station i;

[0168] As a reference proportional signal, determined by the system's total frequency modulation requirements. Total capacity of the cluster The ratio is determined;

[0169] This is the coefficient of the traction term, dimensionless, with a recommended value range of 0.05 to 0.10;

[0170] This indicates that belonging to the neighbor set Sum all neighboring nodes j.

[0171] 3. The connectivity requirement of the consensus iterative algorithm is: the communication graph must contain at least one spanning tree, that is, the information of each node can reach all other nodes through a directed path.

[0172] Laplace matrix:

[0173] L=DA

[0174] in, Let D be the adjacency matrix and D be the degree matrix. The degree matrix is ​​a diagonal matrix, and its diagonal elements are... Off-diagonal elements are 0.

[0175] Convergence condition:

[0176] ;

[0177] in Let L be the largest eigenvalue of the Laplace matrix L.

[0178] 4. The termination condition and convergence criterion of the consistency iterative algorithm are as follows:

[0179] This system employs a dual convergence criterion, and convergence is determined when both of the following conditions are met:

[0180] Main basis of judgment $\ $ ＜0.001 The change in the condition of each station is less than one-thousandth. Payment of judgment $\ $ ＜0.005 The deviation of each station's condition from the mean is less than 0.5%.

[0181] in The average frequency modulation sharing ratio of all stations is calculated using the following formula: , where n is the total number of stations.

[0182] Maximum iteration limit: Set Secondly, to prevent the algorithm from getting stuck in an infinite loop when communication is abnormal.

[0183] In this embodiment, the specific parameters of the consensus iteration algorithm are as follows:

[0184] Consistent step size 0.20 Fast convergence and no oscillation Communication weight <![CDATA[$1 / N i $]]> uniform weight Traction term coefficient 0.08 Tracking total frequency regulation demand Convergence principal criterion threshold 0.001 The change in state is less than one-thousandth Convergence auxiliary criterion threshold 0.005 Node deviation less than 0.5% Maximum number of iterations 30 Preventing infinite loops RBF learning rate 0.03 Weight update step size momentum factor 0.8 Accelerated convergence communication cycle 100ms Synchronized with the iteration cycle

[0185] In this embodiment, a simulation scenario is used to perform a one-time iterative calculation.

[0186] Simulation scenario: A regional power grid with 4 new energy power stations, total installed capacity of 800 MW (2 wind farms, 1 photovoltaic power station, and 1 energy storage power station), and a new energy penetration rate of 50%.

[0187] Simulation results: Under the above parameter configuration, the consensus iterative algorithm converges within 5 to 8 iterations (corresponding to 500 to 800 milliseconds), and the deviation of the lowest frequency point is controlled within ±0.15Hz.

[0188] In another embodiment, the virtual inertia coefficients adjusted by the adaptive virtual inertia control unit include: the virtual inertia time constant and the frequency modulation coefficient.

[0189] The adaptive virtual inertia control unit uses a fuzzy RBF neural network adaptive algorithm to tune the time-varying virtual inertia and the primary frequency modulation coefficient.

[0190] Frequency regulation at renewable energy power plants is significantly affected by nonlinear factors such as wind speed and irradiance. Fuzzy RBF networks can approximate complex nonlinear mappings without requiring precise mathematical models. RBF neural networks possess local response characteristics and fast learning speeds, making them suitable for millisecond-level frequency regulation requirements. The introduction of fuzzy rules provides interpretability, facilitating engineering debugging of parameter tuning.

[0191] The range (upper and lower limits) of the time-varying virtual inertia and the first-order frequency modulation coefficient, and the basis for their setting are as follows:

[0192] (Virtual inertia coefficient) 2s 15s 6~10s Referencing the synchronous generator's inertial time constant H = 2~10s; the lower limit ensures detectable inertial response, while the upper limit prevents secondary drops caused by excessive rotor speed or prolonged release of kinetic energy. (First frequency modulation coefficient) 5pu 50pu 15~30pu Referencing the governor droop coefficient of 5%, =20pu (per unit value), the lower limit ensures the frequency modulation effect, and the upper limit prevents oscillation caused by over-adjustment.

[0193] Upper / lower limit setting principles:

[0194] =2s: Ensures the effectiveness of inertia support (extremely low values ​​cannot produce a measurable power response).

[0195] =15s: Prevents excessive release of kinetic energy from the fan rotor, which could lead to excessively low speed, grid disconnection, or backflow of subsequent power.

[0196] =5pu: Corresponds to a 6% droop rate, ensuring basic performance of primary frequency modulation.

[0197] =50pu: Corresponds to a 2% droop rate, close to the control limit, preventing over-adjustment.

[0198] In another embodiment, the fuzzy RBF neural network adaptive algorithm is:

[0199]

[0200] in, and For time-varying virtual inertia and primary frequency modulation coefficient.

[0201] In another embodiment, the input layer of the fuzzy RBF neural network adaptive algorithm is:

[0202]

[0203] in, For frequency deviation, The rate of change of frequency, Efforts are being made to transform and support new energy power plants;

[0204] The hidden layer of the fuzzy RBF neural network adaptive algorithm uses radial basis function activation, as follows:

[0205]

[0206] in, As the center vector, Width;

[0207] The output layer of the fuzzy RBF neural network adaptive algorithm is:

[0208] ,

[0209] in, and The output weights of the RBF network are M, where M is the number of hidden layer nodes.

[0210] The fuzzy RBF neural network adaptive algorithm uses fuzzy rules to adaptively adjust the RBF weights, and the weight update law is as follows:

[0211]

[0212] in, For performance indicators, The learning rate (recommended range: 0.01~0.05) Momentum factor (recommended 0.8).

[0213] Fuzzy rule adaptation (adjusting RBF weights):

[0214] Fuzzy inference systems use error and error change rate The input is the weight adjustment amount, and the output is the weight adjustment amount. .

[0215] Fuzzy rule base (5 rules)

[0216] Rule 1 For negative and For negative For positive values ​​(the frequency is severely low and still decreasing, so the frequency modulation factor should be increased significantly). Rule 2 For negative small and For negative small For positive small Rule 3 zero and negative zero Zero Rule 4 For positive small and For positive small For negative small Rule 5 For the upright and For the sake of righteousness A negative value (frequency is significantly too high and still rising, requiring a substantial reduction in the frequency modulation factor)

[0217] In another embodiment, the triggering conditions and update frequency of the fuzzy RBF neural network adaptive algorithm are as follows:

[0218] Continuous adjustment Automatic execution in each control cycle RBF forward calculation is performed every 10ms, and weight update is performed every 100ms (synchronized with the consistency iteration cycle). Event trigger adjustment Frequency deviation or frequency change Perform parameter recalculation immediately

[0219] Time scale relationship:

[0220] Consistency iteration cycle: 100ms (slow channel);

[0221] Adaptive parameter update: 100ms (synchronized with consistency iteration to maintain consistent time scale);

[0222] RBF forward computation: 10ms (fast path, each computation is performed on the current data) and (Actual value).

[0223] In another embodiment, the independent adjustment of the time-varying virtual inertia and primary frequency regulation coefficient by each renewable energy power station can lead to uneven or over-regulation of frequency regulation resources. Therefore, the following coordination mechanism is adopted:

[0224] Consistency constraint propagation: Frequency modulation power sharing ratio output by the consensus algorithm As an additional input to the RBF network, it constrains the adjustment range of each station. The station and The adjustment amount is reduced proportionally.

[0225] Global performance metrics feedback: Each site will report local frequency deviations. Report, consensus algorithm calculates cluster average frequency deviation If the local deviation differs too much from the average value (exceeding 0.02Hz), the adjustment rate of the station parameters will be reduced (the adjustment amount will be halved).

[0226] Anti-oscillation design: The parameter adjustment of each station is limited to ±5% per step to prevent oscillation caused by simultaneous large adjustments of multiple stations.

[0227] In another embodiment, the device for the power regulation execution unit to regulate the active power output of the new energy power station includes at least one of the following: wind turbine, energy storage system, and photovoltaic power station;

[0228] The allocation principle and priority order of power commands at the site level among different devices are based on the principle of "energy storage taking priority to bear short-term power surges, and wind / solar taking responsibility for continuous power support", as follows:

[0229] 1 Energy storage system Inertia response (millisecond level) + initial stage of primary frequency modulation (seconds) Fastest response time, reaching rated power output in milliseconds. 2 wind turbine Inertia support (kinetic energy release, in seconds) Rotor kinetic energy can be released quickly, but its capacity is limited and there is a problem with speed recovery. 3 Photovoltaic power plant (load reduced) Frequency modulation duration (seconds) By reducing load and reserving backup capacity, the response is slower but sustainable.

[0230] Specific allocation formula:

[0231] Energy storage plays a crucial role in:

[0232] Wind turbine undertaking:

[0233] Photovoltaic undertaking:

[0234] In this embodiment, the wind turbine load reduction method is as follows:

[0235] Overspeed control (priority) Increase rotor speed to deviate from MPPT optimal speed, retaining 10-20% reserve capacity. No mechanical wear, fast response (milliseconds), but limited by the maximum rotational speed. Moderate wind speed operating conditions Pitch angle control (optional) Increasing the pitch angle reduces the wind energy capture factor Cp, thus preserving reserve capacity. It has a wide adjustable range, but a slow response time (on the order of seconds) and is subject to mechanical wear. High wind speed conditions

[0236] In this embodiment, the photovoltaic load reduction method is as follows:

[0237] Load reduction is achieved by adjusting the DC bus voltage to deviate from the MPPT voltage point:

[0238]

[0239] in To achieve a load reduction rate (typically 5-15%), the DC bus voltage is increased. Deviation Achieve. Load reduction rate. With voltage offset The relationship is determined by the PV characteristic curve of the photovoltaic array. > It operates in the descending segment on the right side of the PV curve. Follow The increase is due to the increase in demand. Reserve 5-10% of the capacity as a backup. =5~10%) can participate in one frequency regulation, and the impact on power generation loss is small (annual power generation loss <2%).

[0240] The division of labor between inertial response power (millisecond level) and primary frequency modulation power (second level) on the time scale is as follows:

[0241] 0~500ms (Initial inertial response) Energy storage system as the main Energy storage and discharge ≤2s (to prevent over-discharge) 500ms~2s (mid-to-late stage of inertial response) Energy storage + wind power release Energy storage is the primary energy source, supplemented by the kinetic energy of the wind turbine rotor. The fan will stop after 2-3 seconds. 2s~15s (one frequency adjustment) Energy storage + wind / solar load reduction Energy storage + reserved backup capacity Continue until frequency recovery

[0242] The mechanism for separating the inertial response power and primary frequency regulation power of energy storage:

[0243] The energy storage controller has two internal modules: a high-speed channel (inertia response, delay < 50ms, response time < 200ms) and a low-speed channel (primary frequency regulation, delay < 200ms, response time < 1s).

[0244] Using frequency deviation signal Power commands are generated simultaneously through parallel channels: ,in (Inertia response channel). (Primary frequency modulation channel);

[0245] The two modules jointly determine the total power command, which is then controlled by a limiting circuit. Ensure that the total output does not exceed the rated power of the energy storage.

[0246] In another embodiment, the energy storage SOC is too low or

[0247] SOC > 90% Overcharge area Charging power is limited to 0; discharging power is normal. Only discharging is allowed to participate in frequency modulation; if charging is required, it must be switched to bypass. 80% < SOC < 90% High SOC area Charging power drops to 50% of rated power; discharging power is normal. Prioritize discharging, limit charging 20%≤SOC≤80% Normal area Unlimited power Normal participation in frequency modulation 10%≤SOC<20% Low SOC area Discharge power drops to 50% of rated power; charging power is normal. Prioritizes charging for recovery, with limited discharging. SOC < 10% Over-release zone Discharge power is limited to 0; charging power is normal. Charging only, FM off.

[0248] The SOC recovery strategy is as follows:

[0249] When SOC < 20%, if the system is in a non-frequency modulation state, it is restored by charging with a small current of 0.1C to 0.2C.

[0250] If the system is in frequency regulation mode, reduce the power command allocation weight of this energy storage unit and transfer the power demand to other healthy units.

[0251] In another embodiment, the acquired status information of adjacent renewable energy power stations includes their adjustable capacity upper limit, adjustable capacity lower limit, frequency deviation, and frequency modulation power sharing ratio. per unit value ).

[0252] The method for real-time calculation of the adjustable capacity of renewable energy power plants is as follows:

[0253] 1. Wind farm:

[0254] Current active power calculation formula:

[0255]

[0256] in:

[0257] The density is expressed in kilograms per cubic meter (kg / m2), and is taken as 1.225 under standard atmospheric pressure.

[0258] The swept area of ​​the wind turbine is expressed in square meters (m²). 2 );

[0259] The wind energy utilization coefficient is dimensionless and ranges from 0 to 0.593 (Bez limit).

[0260] The tip speed ratio is dimensionless.

[0261] The pitch angle is expressed in degrees (°).

[0262] Wind speed, measured in meters per second (m / s).

[0263] Adjustable capacity calculation formula:

[0264]

[0265] in:

[0266] Rated power of wind turbine, in megawatts (MW);

[0267] Current active power, in megawatts (MW);

[0268] The increment of wind energy utilization coefficient that can be increased by overspeeding or pitch control is dimensionless;

[0269] Limited by the maximum speed limit (Overspeed protection action).

[0270] Formula for calculating adjustable capacity:

[0271]

[0272] in:

[0273] The minimum operating power is typically 10% of the rated power. ;

[0274] The reduction in wind energy utilization coefficient that can be reduced by pitch control is dimensionless;

[0275] Limited by minimum speed (To prevent stalling)

[0276] 2. Photoelectric field:

[0277] Current MPPT power calculation formula:

[0278]

[0279] in:

[0280] G represents irradiance, expressed in watts per square meter (W / m²). 2 )

[0281] A represents the total area of ​​the photovoltaic array, in square meters (m²). 2 )

[0282] Photovoltaic conversion efficiency, dimensionless;

[0283] Adjustable capacity

[0284]

[0285] (Photovoltaics typically operate at their maximum power point (MPPT), with no room for upward adjustment).

[0286] Adjustable capacity

[0287]

[0288] in The load reduction rate typically ranges from 5% to 15%.

[0289] 3. Energy storage system:

[0290] Charging power limit (lower capacity, power absorbed)

[0291]

[0292] Upper limit of discharge power (increase capacity, release power)

[0293]

[0294] in:

[0295] Rated energy storage capacity, expressed in megawatts (MW).

[0296] To determine the maximum permissible state of charge, we take 90%.

[0297] To ensure the minimum permissible state of charge, we take 20%;

[0298] The current state of charge is expressed as a percentage (%).

[0299] The rated capacity of the energy storage is expressed in megawatt-hours (MWh).

[0300] Frequency modulation duration, in hours (h).

[0301] 4. Hybrid type of station:

[0302] For hybrid power plants that simultaneously incorporate wind power, solar power, and energy storage, the aggregated adjustable capacity is added separately according to the power direction:

[0303]

[0304]

[0305] in, This represents the summation of all power generation units i within the station;

[0306] Considering the simultaneous rate correction:

[0307]

[0308] in, Simultaneity factor: 1.0 for energy storage, 0.8 for wind power, and 0.7 for photovoltaic power.

[0309] 5. The specific algorithms for the adjustable capacity of various types of stations are as follows:

[0310] wind farm Limited by rated power, smaller Limited by minimum speed, larger Increased power generation is limited, but load reduction can be achieved through a wide range of pitch adjustments. Photovoltaic power station Approximately 0 Limited by load reduction rate, medium MPPT has no room for adjustment. Energy storage power station Limited by SOC and discharge power Limited by SOC and charging power Independent in each direction, asymmetrical in most cases

[0311] The system processes the following separately: the capacity increase and decrease are calculated and reported separately, and then processed separately during the consistency iteration.

[0312] In another embodiment, the situation of adjacent renewable energy power stations is as follows:

[0313] 1. Communication between adjacent renewable energy power stations adopts a mesh topology, with each power station establishing communication links with multiple neighbors to provide redundant communication paths.

[0314] 2. Adjacent renewable energy power stations are determined using a comprehensive assessment of both electrical distance and communication signal quality:

[0315] electrical distance (per unit value) 60% The electrical coupling strength determines the degree of mutual influence in frequency modulation. Communication signal quality RSSI > -100dBm (communication reliability ≥ 90%) 40% Ensure the feasibility of communication link engineering

[0316] Overall score calculation formula:

[0317]

[0318] in:

[0319] The score is a comprehensive rating, ranging from 0 to 1.

[0320] The per-unit value of the electrical distance between node i and node j;

[0321] The signal strength indicator value received by node i from node j, in dBm;

[0322] Let be the magnitude of the equivalent mutual impedance between node i and node j. This is the reference impedance value;

[0323] Judgment rule: When If the value is greater than 0.5, then node i and node j are considered to be adjacent.

[0324] 3. The communication topology update rules when a site joins or leaves are as follows:

[0325] (1) After the new station LIU is powered on, it broadcasts a “join request” frame (including the node ID and rated capacity). (Information such as GPS coordinates);

[0326] (2) Once received, the existing node replies with an acknowledgment frame (ACK);

[0327] (3) The new LIU collects all confirmation frames and determines them according to the adjacent criteria. >0.5) Filter for formal neighbors;

[0328] (4) The new LIU sends a connection establishment request frame to the selected neighbor, and the neighbor replies with a connection establishment confirmation frame;

[0329] (5) The new LIU broadcasts a "Joining Complete" notification, updating neighbor information across the entire network;

[0330] Smooth access to new nodes: In the first 5 control cycles (500 milliseconds in total), the state variables gradually transition from measured values ​​to collaboratively calculated values, with a transition of 20% per cycle.

[0331] Existing station decommissioning inspection:

[0332] Heartbeat detection: Each LIU broadcasts a heartbeat frame every 100 milliseconds. If no heartbeat is received from a node for three consecutive heartbeat cycles (300 milliseconds), it is determined to exit.

[0333] Active Exit: When the station is scheduled for maintenance or shutdown, an active exit notification frame is sent.

[0334] Anomaly detection: Receiving anomaly data three times consecutively (e.g., A communication fault is determined when the adjustable capacity is negative.

[0335] Exit Processing: Each node removes the exiting node from its neighbor table and recalculates its communication weight. .

[0336] In another embodiment, to ensure the computational feasibility of the consistency iterative algorithm, the topology design in this example is as follows:

[0337] Minimum number of neighbors Each node must maintain at least 3 neighbors. There is still a backup path when any link fails. Suboptimal node chain replenishment When the number of neighbors is less than 3, actively request the second-best node to join. Preventing network splitting Dual-channel backup Simultaneously configure two communication channels: LoRa and PLC. Automatic switching when one channel fails

[0338] 4. New station discovery mechanism

[0339] Step ① <![CDATA[The new LIU sends a broadcast discovery frame after power-on (including the node ID, rated capacity P N , GPS coordinates)]]> ≤1 second Step ② LIU has already received and replied with an acknowledgment frame (including the node ID and current status). ≤500 milliseconds Step ③ New LIU collects all confirmation frames, press >0.5 Filter for formal neighbors ≤3 seconds Step 4 The new LIU sends a connection establishment request frame to the selected neighbor, and the neighbor replies with a connection establishment confirmation frame. ≤1 second Step 5 New LIU broadcast notification added, neighbor information updated across the network. ≤1 second

[0340] Total network setup time: less than 6.5 seconds

[0341] 5. Method for setting initial values ​​for consistency variables

[0342] FM sharing ratio Allocate capacity proportionally to allow for rapid integration into the cluster. FM power Preventing power surges during connection Virtual inertia coefficient Second Take the median value and adjust it after the adaptive algorithm takes over. First frequency modulation coefficient Take the median value and adjust it after the adaptive algorithm takes over.

[0343] 6. Detection mechanism for station shutdown or malfunction

[0344] Heart rate detection Each LIU broadcasts a heartbeat frame (including node ID, status, and timestamp) every 100 milliseconds. If no heartbeat is received from a certain node for three consecutive heartbeat cycles (300 milliseconds), it is determined to exit. Data anomaly detection Data received from a certain node exceeds the reasonable range If abnormal data is received three times in a row, the node is determined to have a communication failure. voluntary withdrawal notification When the station is scheduled for maintenance or shutdown, it sends an active exit notification frame. Update topology upon receiving notification frame

[0345] Exit Processing: After a node is determined to be exiting, the remaining nodes remove that node from their neighbor table and recalculate the communication weight. If a node has fewer than 3 neighbors, the suboptimal node chain replenishment mechanism is activated.

[0346] In another embodiment, communication anomalies between the various control subsystems are handled using the following strategy:

[0347] Single data frame timeout 50 milliseconds Applying the effective value of the previous period as a substitute Automatically resume when a new data packet is received in the next cycle. Continuous heartbeat timeout 300 milliseconds If a neighbor is found to be faulty, delete that neighbor and recalculate its weight; if the number of neighbors is less than 3, initiate chain replacement. After receiving the heartbeat from the node again, add it back to the neighbor table. complete communication blackout 600 milliseconds Switch to local independent FM mode Automatically switch back to distributed collaborative mode after detecting any neighbor's heartbeat.

[0348] Local independent frequency modulation mode

[0349] When communication is completely interrupted, each station switches to local independent frequency modulation mode, and the control equation is:

[0350]

[0351] in, (Fixed value), each station operates independently and does not rely on communication data.

[0352] At the same time, a communication recovery probe is attempted every 10 seconds (sending a heartbeat probe frame and listening for neighbor replies).

[0353] Data validation / outlier detection mechanism

[0354] Frequency deviation exceeds limit Discard the data and replace it with a valid value from the previous period. Adjustable capacity is negative or Determined to be invalid, discard. FM sharing ratio exceeded the limit Determined as an outlier, discarded.

[0355] Smooth transition mechanism

[0356] After communication is restored, the system switches from local independent frequency modulation mode back to distributed cooperative mode. For a smooth transition, the power command gradually transitions from local mode values ​​to cooperative mode values ​​over the first five control cycles (500 milliseconds in total):

[0357]

[0358] Where k = 1, 2, 3, 4, 5, is the number of transition steps.

[0359] like Figure 2 As shown, a distributed cooperative inertia and primary frequency regulation control method for new energy power stations includes the following steps:

[0360] Step S1: Check if the system frequency exceeds the limit. If the system frequency exceeds the limit, proceed to step S2; otherwise, proceed to step S11.

[0361] The frequency deviation threshold of this system is set to... (Corresponding to f≤49.98Hz or f≥50.02Hz).

[0362] The settings are based on the following: my country's GB / T15945 standard specifies that the normal range of power grid frequency is 49.8 Hz to 50.2 Hz. Using 0.02 Hz as the dead zone threshold avoids frequent frequency modulation triggering within the standard's allowable range; frequency measurement noise (accuracy ±0.005 Hz) and normal fluctuations (±0.01 Hz) are both less than 0.02 Hz, and the threshold design effectively avoids false triggering caused by measurement noise and normal load fluctuations; The system is only in monitoring mode and does not trigger frequency regulation, ensuring that the new energy power plants are not frequently disturbed by the small frequency fluctuations of the power grid.

[0363] Step S2: Each control subsystem collects the frequency change rate, frequency deviation, and active power output of the new energy power stations it is connected to;

[0364] Step S3: Each control subsystem uses the RoCoF method after perturbation of the swing equation and the recursive least squares system identification method to estimate the equivalent inertia of the system.

[0365] Step S4: Based on the consensus iterative algorithm, calculate the final estimated system equivalent inertia by weighting the estimated system equivalent inertia of each control subsystem according to its rated capacity ratio;

[0366] Step S5: Each control subsystem collects frequency modulation sharing ratio, up-regulation capacity, and down-regulation capacity status information of adjacent renewable energy power stations; specifically, each power station broadcasts its local status to neighboring nodes via a LoRa / PLC low-speed communication network. , , , Simultaneously, it receives status information from neighboring nodes; the communication data frame structure is 64 bytes, including node ID (2 bytes), frequency deviation (2 bytes), frequency modulation ratio (2 bytes), adjustable capacity (4 bytes x 2), timestamp (4 bytes), and CRC check (2 bytes); after receiving all valid neighbor data, it proceeds to step S6; if the communication times out (50 milliseconds), the valid value from the previous cycle is used instead;

[0367] Step S6: Based on the consensus iterative algorithm, calculate the frequency regulation power that each renewable energy power station should undertake; the details are as follows:

[0368] Execution logic: Consistent update equations for operation at each site:

[0369]

[0370] Transition condition: Iterative convergence ( and (or reaching the maximum number of iterations) Proceed to step S7;

[0371] Step S7: Based on the fuzzy RBF neural network, calculate the virtual inertia coefficient and the primary frequency modulation coefficient according to the current frequency deviation, frequency change rate, and adjustable capacity.

[0372] Step S8: Generate frequency modulation power command; details are as follows:

[0373] , , According to the upper and lower limits ( Second, Limit the amplitude;

[0374] Step S9: Control each device to execute according to the priority of the master frequency modulation command. The priority of the master frequency modulation command is as follows:

[0375] Priority 1: Energy storage system (inertia response + initial stage of primary frequency regulation), priority 2: Wind turbine (kinetic energy release), priority 3: Photovoltaic load shedding. Each device is executed after its own capacity is limited. After execution, the system frequency is checked to see if it exceeds the limit. If the system frequency exceeds the limit, proceed to step S10; otherwise, proceed to step S11.

[0376] Step S10: After adjusting the adjustment weights, proceed to step S6;

[0377] Step S11: Does the periodic system frequency exceed the limit? If the system frequency exceeds the limit, proceed to step S1; otherwise, proceed to step S12.

[0378] Before proceeding to step S12, the system exits frequency modulation mode. Specifically:

[0379] 1. The trigger condition for exiting is:

[0380] The system will only start and exit if both conditions are met simultaneously:

[0381] 1. (The frequency has been restored to the allowable steady-state range);

[0382] 2. And maintain this state for 3 seconds (the rate of frequency change tends to stabilize).

[0383] The exit is delayed by 200ms after it is triggered to prevent repeated switching caused by instantaneous frequency fluctuations.

[0384] 2. The specific form of the exit curve is a piecewise exponentially decreasing exit curve.

[0385]

[0386] in For the exit time constant, The duration of the exit.

[0387] Parameter design:

[0388] Quickly exit Phase 1 0~0.5s 0.2s Quickly exit 40% power Energy storage quickly withdraws part of the power Slowly exiting the second phase 0.5~2.5s 1.0s The remaining power slowly returns to zero. Wind / solar load reduction gradually recovers

[0389] 3. During the shutdown process, each station coordinates its shutdown pace through a consensus algorithm, rather than shutting down independently. If each station shuts down independently, slight differences in frequency measurements between stations (due to measurement noise, communication delays, etc.) may lead to the following problems: some stations may detect frequency recovery first and begin shutting down, while others may continue full-output frequency modulation, resulting in inconsistent rates of decrease in the total frequency modulation power of the cluster; stations that shut down too quickly may cause a secondary frequency drop, triggering other stations to re-enter the network, resulting in repeated adjustments (oscillations).

[0390] Each station coordinates its exit pace through a consensus algorithm, with the specific update equations as follows:

[0391]

[0392] in:

[0393] : Change in power output of power station i during the k-th iteration (unit: MW);

[0394] Rated capacity of station i (unit: MW);

[0395] Consistency step size, consistent with 'a' in the main algorithm (taken as 0.15~0.30);

[0396] Communication weights, in relation to the main algorithm same;

[0397] : The set of neighbors of station i;

[0398] : For those belonging to the neighbor set Summing all neighboring nodes j;

[0399] The above equations determine the exit rates of each station. (That is, the rate of change of power withdrawn per unit rated capacity) tends to be consistent. Each power station synchronously withdraws frequency power in proportion to its rated capacity.

[0400] For example, if power plant A has a rated capacity of 200MW and power plant B has a rated capacity of 100MW, then consistency coordination ensures that the decommissioning rate of A is always twice that of B. The final result is that A will be decommissioned. At the same time as MW, B withdrew. MW, both stations will be decommissioned simultaneously according to their capacity ratio.

[0401] The exit procedure is as follows:

[0402] After the exit trigger condition is met ( and (Continues for 3 seconds, then starts after a 200ms delay), the following actions will be executed simultaneously:

[0403] (1) Start and exit iteration: Each station starts the above consistency update equation and performs one iteration every control cycle (100ms).

[0404] (2) Calculate the current withdrawal amount: After each iteration, each station calculates the power reduction that should be achieved in this cycle. .

[0405] (3) Update power command: Each station subtracts the current power command from the current power command. This is issued and executed as a new power command.

[0406] (4) Continue iterating: Repeat the above process until all stations are completed. All converged to 0 (i.e., all power plants' power outputs recovered to pre-shutdown levels, with the total power withdrawn reaching a cumulative total of 0). , The cumulative value has reached the initial frequency modulation power of the station. )

[0407] The total duration of the exit is approximately 2.5 seconds (the total time of the first two stages of the segmented exponentially decreasing curve).

[0408] Comparison with other solutions

[0409] This invention Consistency Coordination Exit Each station will be rolled out simultaneously according to its capacity ratio to avoid a second frequency drop. Relies on communication and requires maintaining consistent iteration. Independent Exit Each station independently determines the timing and rate of withdrawal. It does not rely on communication and is simple and reliable. This can easily lead to a second drop in frequency and repeated adjustments.

[0410] The degradation strategy in case of communication failure is as follows:

[0411] If a complete communication interruption is detected during the exit process (no neighbor heartbeats received for 600 ms), each station automatically switches to local independent exit mode and exits independently according to its own exit curve (using a preset time constant). (The exponential decreases per second), no longer relying on consistency coordination. At this time, the exit rates of each site are not completely consistent, but this invention still guarantees a smooth (non-abrupt) exit for individual sites, only the overall exit pace of the cluster may vary slightly.

[0412] 4. Reinvestment strategy when the frequency deteriorates again during the exit process:

[0413] Detection mechanism:

[0414] Continuous monitoring of frequency deviation during exit process .like (If the frequency exceeds the limit again), immediately interrupt the exit process.

[0415] Re-engaging in action:

[0416] Freeze command: The current power command is frozen at the current value. (Unit: MW), stop decreasing.

[0417] Reverse increase: with Starting from this point, the power command is increased again according to the power increment curve of the initial frequency modulation. Increment Equal to the power increment at the first response But the starting point is Instead of initial value .

[0418] Smooth input: Power is smoothly increased through a first-order inertial element, with a time constant of 0.3 seconds. The transfer function of the first-order inertial element is G(s) = 1 / (1 + 0.3s), where s is the Laplace operator.

[0419] Timer reset: Exit the timer. Reset to zero.

[0420] Step S12: Periodically assess the system status;

[0421] Step S13: Determine whether parameter readjustment is required. If so, proceed to step S14; otherwise, proceed to step S11.

[0422] Step S14: Online parameter learning updates the control model, proceed to step S11.

[0423] In step S9, the priorities are, in order, energy storage system, wind turbine, and photovoltaic power station.

[0424] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distributed cooperative inertia and primary frequency regulation control system for new energy power plants, characterized in that, include: Several control subsystems are installed at the grid connection points of each new energy power station; A communication subsystem that enables data interaction between several of the aforementioned control subsystems; The control subsystem includes: The frequency measurement unit measures the frequency change rate, frequency deviation, and active power output of the new energy power station in real time. Inertia estimation unit, which estimates the equivalent inertia of the system in real time or under specific conditions. The distributed collaborative control unit acquires the status information of adjacent new energy power stations and collaboratively calculates the frequency regulation power that each new energy power station should undertake based on a consensus iterative algorithm. The adaptive virtual inertia control unit dynamically adjusts the virtual inertia coefficient according to the frequency change rate. The power regulation execution unit adjusts the active power output of the new energy power station based on the status of the new energy power station equipment.

2. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 1, characterized in that, The inertia estimation unit uses the perturbation-based RoCoF method and the recursive least squares system identification method to estimate the equivalent inertia of the system. The equivalent inertia of the system is estimated using the RoCoF method after perturbation of the swing equation; The equivalent inertia of the system obtained by RoCoF method after the perturbation of the swing equation is corrected by the recursive least squares system identification method; The specific condition for the inertia estimation unit to estimate the equivalent inertia of the system is that the power deficit is greater than a predetermined value. The RoCoF method after the perturbation of the swing equation is used to quickly estimate the equivalent inertia of the system.

3. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 1, characterized in that, In the consensus iteration algorithm, the frequency modulation power sharing ratio is selected as the consensus variable; the consensus iteration algorithm uses a dual convergence criterion of the state change of each new energy power station and the deviation of the state of each new energy power station from the mean.

4. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 1, characterized in that, The virtual inertia coefficients adjusted by the adaptive virtual inertia control unit include the virtual inertia time constant and the frequency modulation coefficient.

5. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 1, characterized in that, The adaptive virtual inertia control unit uses a fuzzy RBF neural network adaptive algorithm to tune the time-varying virtual inertia and the primary frequency modulation coefficient.

6. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 5, characterized in that, The fuzzy RBF neural network adaptive algorithm is as follows: in, and For time-varying virtual inertia and primary frequency modulation coefficient.

7. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 5 or 6, characterized in that, The input layer of the fuzzy RBF neural network adaptive algorithm is: in, For frequency deviation, The rate of change of frequency, Efforts are being made to transform and support new energy power plants; The hidden layer of the fuzzy RBF neural network adaptive algorithm uses radial basis function activation, as follows: in, As the center vector, Width; The output layer of the fuzzy RBF neural network adaptive algorithm is: ， in, and The output weights of the RBF network are M, where M is the number of hidden layer nodes. The fuzzy RBF neural network adaptive algorithm uses fuzzy rules to adaptively adjust the RBF weights, and the weight update law is as follows: in, For performance indicators, The learning rate (recommended range: 0.01~0.05) This is the momentum factor.

8. The distributed cooperative inertia and primary frequency regulation control system for new energy power stations according to claim 1, characterized in that, The equipment used by the power regulation execution unit to regulate the active power output of new energy power plants includes at least one of the following: wind turbine generators, energy storage systems, and photovoltaic power plants; When the system faces short-term power surge demand, the power regulation execution unit regulates the energy storage system to undertake inertial response and support the initial stage of primary frequency regulation; When the system faces the need to provide continuous power support, the power regulation execution unit adjusts the inertia support provided by the new energy power station. When the system faces the need to provide continuous power support, the power regulation execution unit adjusts the load reduction of the new energy power station to provide support in the initial stage of primary frequency regulation.

9. A method for distributed cooperative inertia and primary frequency regulation control of new energy power stations, characterized in that, Includes the following steps: Step S1: Check if the system frequency exceeds the limit. If the system frequency exceeds the limit, proceed to step S2; otherwise, proceed to step S11. Step S2: Each control subsystem collects the frequency change rate, frequency deviation, and active power output of the new energy power stations it is connected to; Step S3: Each control subsystem uses the RoCoF method after perturbation of the swing equation and the recursive least squares system identification method to estimate the equivalent inertia of the system. Step S4: Based on the consensus iterative algorithm, calculate the final estimated system equivalent inertia by weighting the estimated system equivalent inertia of each control subsystem according to its rated capacity ratio; Step S5: Each control subsystem collects information on the frequency regulation sharing ratio, up-regulation capacity, and down-regulation capacity status of adjacent renewable energy power stations; Step S6: Calculate the frequency regulation power that each new energy power station should undertake based on the consensus iterative algorithm; Step S7: Based on the fuzzy RBF neural network, calculate the virtual inertia coefficient and the primary frequency modulation coefficient according to the current frequency deviation, frequency change rate, and adjustable capacity. Step S8: Generate frequency modulation power command; Step S9: Control each device to execute according to the priority of the total frequency modulation command, and then check whether the system frequency exceeds the limit. If the system frequency exceeds the limit, proceed to step S10; otherwise, proceed to step S11. Step S10: After adjusting the adjustment weights, proceed to step S6; Step S11: Does the periodic system frequency exceed the limit? If the system frequency exceeds the limit, proceed to step S1; otherwise, proceed to step S12. Step S12: Periodically assess the system status; Step S13: Determine whether parameter readjustment is required. If so, proceed to step S14; otherwise, proceed to step S11. Step S14: Online parameter learning updates the control model, proceed to step S11.

10. The distributed cooperative inertia and primary frequency regulation control method for new energy power stations according to claim 9, characterized in that, In step S9, the priorities are, in order, energy storage system, wind turbine, and photovoltaic power station.

Images