New energy control parameter adaptive identification method based on double-layer particle swarm optimization

Abstract

The application provides a new energy control parameter self-adaptive identification method based on double-layer particle swarm optimization, and belongs to the technical field of power system simulation analysis. Typical control strategies of heterogeneous new energy units are comprehensively used to construct a strategy database and perform verification; test data are acquired for processing and feature extraction, so that high and low threshold values, current initial states and test feature data are obtained; a fitness function is constructed based on the strategy database, a discrete particle swarm algorithm is used in the global optimization layer to select an optimal strategy combination from the strategy database, and in the local optimization layer, parameters of the selected strategy are continuously optimized in combination with a momentum factor and a learning rate. The application uses actual unit test data characteristic values as inputs of the double-layer particle swarm algorithm, determines an optimal strategy-parameter pair through a total error index weighting, verifies the accuracy of the strategy-parameter pair, and effectively improves the efficiency and precision of modeling of heterogeneous new energy equipment.

Description

Technical Field

[0001] This invention relates to an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization, belonging to the field of power system simulation and analysis technology. Background Technology

[0002] With the rapid growth of installed capacity of new energy sources such as wind power and photovoltaics, the penetration rate of new energy equipment in the power system continues to rise. Their large-scale grid connection makes the accurate identification of control strategies and parameters crucial to ensuring the stable operation of the system. The importance of new energy simulation modeling for the analysis of power system operation modes is also becoming increasingly prominent.

[0003] In the process of building structured models for new energy sources, establishing standardized simulation models can provide convenient and rapid verification methods and simulation analysis tools for the analysis of grid connection characteristics, verification of control strategies, and optimization and improvement of new energy sources. It can also provide theoretical and technical support for large-scale wind power grid connection planning and design, operation control, and fault analysis. The construction of structured models for new energy sources encompasses the construction of standardized technical processes for new energy unit modeling, research on algorithms for identifying key control parameters of units, and testing and verification of new energy unit models.

[0004] However, current work faces numerous challenges. Existing parameter identification methods mostly focus on parameter optimization for single device types, improving convergence efficiency through enhanced intelligent algorithms. For example, Chinese invention patent application CN117335486A discloses a photovoltaic inverter parameter identification method and system based on chaotic particle swarm optimization. It significantly improves identification accuracy in photovoltaic scenarios by constructing a dual-loop control model for photovoltaic inverters and employing a chaotic disturbance and fitness variance determination mechanism to optimize parameter search. However, the patent model architecture is fixed to the control strategy of specific devices (such as photovoltaic inverters) and cannot be compatible with the strategy library of diverse new energy devices such as wind turbines; moreover, the optimization process only uses a single particle swarm framework to handle hybrid optimization problems, making it difficult to balance global exploration and local development; in addition, its verification mechanism relies on static calculation of current error and lacks multi-dimensional closed-loop verification based on dynamic response characteristics, resulting in insufficient generalization of the model in transient scenarios such as actual grid fault ride-through.

[0005] In summary, there is an urgent need for a new energy control parameter identification method that supports multi-strategy compatibility, hierarchical collaborative optimization, and dynamic feature verification. This method would overcome key technical bottlenecks such as poor adaptability of heterogeneous equipment, low efficiency of strategy-parameter coupling optimization, and disconnect between models and actual operating conditions, providing theoretical support and engineering tools for the stable control of high-proportion new energy power grids. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization.

[0007] The technical solution of the present invention is as follows:

[0008] On one hand, this invention provides an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization, the method comprising:

[0009] By integrating the typical control strategies of four types of heterogeneous new energy units, namely wind power, photovoltaic, energy storage and SVG, a strategy database is constructed that includes a mathematical model of voltage-current response characteristics with multiple different typical control strategies, the value range of parameters associated with each control strategy and the initial parameter values. The mathematical model of voltage-current response characteristics is in the form of a piecewise function.

[0010] Obtain actual unit test data, extract features from the test data, and obtain test feature data;

[0011] A two-layer particle swarm optimization framework, comprising a global optimization layer and a local optimization layer, is constructed based on a policy database, and a hierarchical fitness function is set.

[0012] The test feature data is input into a two-layer particle swarm optimization framework, which outputs the optimal policy-parameter pair. The global optimization layer uses a discrete particle swarm algorithm, which selects the optimal policy combination using the policy error function as the fitness function. The local optimization layer uses a continuous particle swarm algorithm, which optimizes the associated parameters of the optimal policy combination using the parameter error function as the fitness function.

[0013] Preferably, the strategy database includes a strategy sub-library of SVG, which includes SVG reactive power control strategies, specifically including voltage-controlled current with superimposed initial reactive current component, specified reactive current control, specified reactive power control, SVG current remaining unchanged, SVG reactive current changing to a specified value based on the pre-fault condition, and voltage-controlled current without superimposed initial reactive current component.

[0014] Preferably, the strategy database includes a strategy sub-library for wind power and photovoltaic power, which includes wind and solar reactive power control strategies and wind and solar active power control strategies. The wind and solar reactive power control strategies include specified reactive current control and specified reactive power control; the wind and solar active power control strategies include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

[0015] Preferably, the strategy database includes a strategy sub-library for energy storage, which includes energy storage reactive power control strategies and energy storage active power control strategies under charging and discharging modes. The energy storage reactive power control strategies include specified reactive current control and specified reactive power control. The energy storage active power control strategies under charging and discharging modes include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

[0016] Preferably, the method further includes:

[0017] After constructing the strategy database, the accuracy of the strategy database is verified through a simulation platform to ensure that the simulation data and the dynamic response characteristic curves corresponding to the voltage-current response characteristic mathematical model in the strategy database are completely matched.

[0018] Preferably, the method further includes:

[0019] When extracting features from actual unit test data, a stepped voltage curve is plotted simultaneously. When a sudden change in current is detected, the voltage critical value of the stepped voltage curve is used as the high and low breakdown threshold. The high and low breakdown threshold is used to define the boundary of the piecewise function interval of the mathematical model of voltage-current response characteristics in the strategy database.

[0020] Preferably, the method further includes:

[0021] When extracting features from the test data, the fault ride-through test data under different initial reactive power values ​​are analyzed to determine whether the reactive current changes during the fault are consistent, and the current superposition mode is determined based on the correlation. Specifically:

[0022] By comparing the trend and magnitude of reactive current changes over time with different initial reactive values, we can determine whether the changes are consistent and calculate the correlation coefficient between reactive current changes under different initial reactive values.

[0023] When the correlation coefficient of reactive current changes with different initial reactive values ​​during the fault period is greater than the preset correlation coefficient threshold, it is determined to be an incremental mathematical model.

[0024] When the correlation coefficient of reactive current change with different initial reactive values ​​during the fault period is less than or equal to the preset correlation coefficient threshold, it is determined to be an absolute value mathematical model.

[0025] The current superposition mode is used to control the optimal strategy combination selection process of the global optimization layer, and only activates the corresponding type of mathematical model in the strategy database.

[0026] Preferably, the feature extraction of the test data adopts the fault partitioning method, which divides the test data into low voltage ride-through zone, steady state zone and high voltage ride-through zone. The effective values ​​of active current, reactive current and voltage are extracted in each zone, and the extracted values ​​are used as test feature data.

[0027] Preferably, the method further includes constructing an association parameter set by taking the value range of each control strategy association parameter and the initial parameter value. .

[0028] Preferably, the particle update in the discrete particle swarm optimization algorithm during the iteration process is expressed by the following formula:

[0029] ;

[0030] In the formula, express Global velocity at any given moment; express Global velocity at any given moment; Inertial weight; and For learning factors; and In order to be in Random numbers within; This represents the optimal position in the particle's history. The globally optimal position; This represents the particle's current position.

[0031] During particle update, the Discrete Particle Swarm Optimization (DPSO) algorithm evaluates different policy combinations in the policy database using a policy error function. This policy error function measures the quality of the current policy combination, and the optimal policy combination is obtained based on this function, expressed by the following formula:

[0032] ;

[0033] In the formula, The optimal strategy combination; The initial sample size for evaluating the effectiveness of the strategy; For the first The predicted current value corresponding to each sample; For the first The effective current value corresponding to each sample; Indicated in the policy database Find the strategy that minimizes the policy error function among all strategy combinations. ; For strategy combination The corresponding strategy error function.

[0034] Preferably, the optimization process of the continuous particle swarm optimization algorithm is as follows:

[0035] The local optimization layer corrects particle velocities in real time based on parameter error feedback, expressed by the following formula:

[0036] ;

[0037] In the formula, for Local velocity at a given moment; for Local velocity at a given moment; The learning rate; Momentum factor; For the optimal strategy combination, the associated parameters The corresponding parameter error function; The optimal strategy combination is associated with parameters;

[0038] The parameter error is expressed by the formula:

[0039] ;

[0040] In the formula, To comprehensively evaluate the number of data samples; for The effective value of the current at any given time; for The effective value of the voltage at any given time;

[0041] By optimizing the correlation parameters of the selected optimal strategy combination based on the parameter error, the optimal correlation parameters of the optimal strategy combination are obtained, which can be expressed by the formula:

[0042] ;

[0043] In the formula, Associating optimal parameters with the optimal strategy combination; This represents the unit control model, based on the optimal strategy combination. Optimal strategy combination associated parameters and external input at time Calculate the unit The predicted current value at time; Indicates in the associated parameter set Find the correlation parameter that minimizes the parameter error function value. .

[0044] Preferably, the optimal policy-parameter pair is output, specifically:

[0045] In the two-layer particle swarm optimization framework, the global optimization layer performs a preset number of iterations, and after the policy optimization iteration of the global optimization layer, the parameter optimization iteration of the local optimization layer is performed. Through the two-layer collaborative optimization iteration, the total error index of each policy-parameter pair is calculated, expressed by the formula:

[0046] ;

[0047] In the formula, This is the total error index; For the optimal strategy combination The corresponding policy error function; Associating optimal parameters with the optimal strategy combination The corresponding parameter error function;

[0048] When the total error index is less than the preset error threshold, the optimization is considered complete, and the current policy-parameter pair is output as the optimal policy-parameter pair; otherwise, the optimization is restarted.

[0049] Preferably, the method further includes constructing a simulation model based on the optimal strategy-parameter pair, extracting features from the simulation data generated by the simulation model to obtain simulation feature data, and performing error verification on the test feature data and simulation feature data, wherein the error verification adopts mean square error (MSE), mean absolute error (MAE), or relative error.

[0050] Preferably, the method further includes visualizing the error verification results.

[0051] In another aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in the present invention.

[0052] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in the present invention.

[0053] The present invention has the following beneficial effects:

[0054] 1. This invention is an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization. It integrates control strategies from multiple manufacturers to construct a strategy library containing a reactive-active coupling model and defines the dynamic range of each strategy parameter. It combines a two-layer particle swarm algorithm to achieve hierarchical collaboration between strategy selection and parameter optimization. The global optimization layer uses a discrete particle swarm algorithm to quickly screen suitable control strategies (such as reactive power control and low-voltage current limiting combination), while the local optimization layer dynamically corrects the parameter search direction through momentum factor and learning rate. This solves the local optimum problem caused by strong coupling between strategy and parameters in single optimization, is compatible with the differentiated mathematical models of strategies from different manufacturers, significantly improves the global parameter adaptation capability of new energy equipment under complex operating conditions, and supports the construction of a unified optimization platform for heterogeneous equipment.

[0055] 2. This invention is an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization. It constructs a multi-dimensional feature vector by extracting dynamic response features from measured data, uses it as the optimization target to drive the parameter identification process, and designs a total error index for iterative optimization and threshold determination. Furthermore, it generates a simulation model based on the optimal strategy-parameter pair, and realizes dynamic error verification through feature comparison, forming a closed loop of "feature extraction-optimization-verification". Through the above design, this invention avoids the risk of model decoupling from actual operating conditions in traditional open-loop identification, and supports dynamic expansion of strategy library and parameter set, ensuring the long-term adaptability of the algorithm to new equipment and complex power grid environment.

[0056] 3. This invention is an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization. It proposes to automatically identify the mathematical model type (incremental vs. absolute value) based on test data. By analyzing the correlation of reactive current changes during the fault under different initial reactive values, the superposition mode is determined, and the corresponding model in the strategy database is dynamically activated accordingly. This solves the key problem of model type uncertainty in practical applications and significantly improves the automation, accuracy and robustness of identification.

[0057] 4. This invention is an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization. By using the fault partitioning method (low-voltage region, steady-state region, high-voltage region), the effective values ​​of active current, reactive current, and voltage in each region are extracted as features, focusing on the core information reflecting the dynamic response of the system and more effectively characterizing the control characteristics.

[0058] 5. This invention is an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization. By plotting stepped voltage curves and determining the voltage critical value as the high-low current threshold when the current changes abruptly, it is used to define the segmented interval boundary of the strategy model, thereby improving the adaptive capability of the method. Attached Figure Description

[0059] Figure 1This is a flowchart of the method of the present invention;

[0060] Figures 2-9 These are simulated wind turbine fault-period response characteristic curves under different control strategies according to embodiments of the present invention. Figure 2 The figure shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 1. Figure 3 The figure shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 2. Figure 4 The figure shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 3. Figure 5 The image shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 4. Figure 6 The image shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 5. Figure 7 The figure shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 6. Figure 8 The figure shows the simulated fault-period response characteristic curves of the wind turbine under Strategy 7. Figure 9 The simulated wind turbine fault period response characteristic curves under Strategy 8 are shown.

[0061] Figures 10-11 This refers to the determination of the high and low penetration thresholds in this embodiment of the invention. Figure 10 This is a graph showing the results of the low penetration threshold determination. Figure 11 This is a graph showing the results of the high penetration threshold determination;

[0062] Figure 12 This is a diagram for determining whether there is superimposed initial reactive current in an embodiment of the present invention;

[0063] Figure 13 This is a flowchart of the two-layer particle swarm optimization framework according to an embodiment of the present invention;

[0064] Figure 14 This is a graph showing the error verification results of test feature data and simulation feature data corresponding to different control strategies under different operating conditions in this embodiment of the invention. Figure 14 (a) is a graph showing the results of reactive power control error verification under low operating conditions. Figure 14 (b) is a graph showing the verification results of reactive power control error under large operating conditions. Figure 14 (c) is a graph showing the verification results of active power control error under small operating conditions. Figure 14 (d) is a graph showing the verification results of active power control error under large operating conditions. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0067] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0068] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0069] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0070] Example 1:

[0071] See Figure 1 This embodiment provides an adaptive identification method for new energy control parameters based on two-layer particle swarm optimization, the method comprising:

[0072] S1. Collect all controller-related information for different types of heterogeneous new energy units such as wind power, photovoltaic, energy storage, and SVG. The controller-related information includes typical control strategy documents, control parameter manuals, and actual measured data provided by various equipment manufacturers under different operating conditions. Different types of heterogeneous new energy units possess unique control strategies due to differences in their energy characteristics, working principles, and application scenarios. Taking wind power as an example, the study collects control strategy information for various wind turbine units in wind farms under typical operating conditions such as normal power generation, sudden wind speed changes, and grid faults. Taking photovoltaics as an example, the study focuses on acquiring relevant data on the control strategies of photovoltaic inverters in photovoltaic power plants under different operating conditions such as light intensity, temperature changes, and day-night cycles. Taking energy storage as an example, the study collects controller data for energy storage systems under various operating conditions such as charging, discharging, energy regulation, and participation in grid frequency regulation. The control strategy of energy storage systems needs to take into account factors such as battery life, charging and discharging efficiency, and grid demand. Taking SVG as an example, the study acquires control strategy information for SVG in different application scenarios such as improving grid power factor, stabilizing voltage, and suppressing harmonics to meet the grid's requirements for dynamic reactive power compensation.

[0073] Based on the collected controller-related data of the aforementioned heterogeneous new energy units, a comprehensive strategy database will be established. This database will contain various mathematical models of voltage-current response characteristics for typical control strategies of different types of units, such as common voltage control, current control, and low-through / high-through logic. These mathematical models accurately describe the electrical response characteristics of the equipment under each control strategy in the form of formulas.

[0074] Furthermore, in this embodiment, the strategy database includes a strategy sub-library for SVG, a strategy sub-library for wind power and photovoltaics, and a strategy sub-library for energy storage, wherein:

[0075] The SVG strategy sub-library includes SVG reactive power control strategies, specifically including voltage-controlled current with superimposed initial reactive current component, specified reactive current control, specified reactive power control, SVG current remaining constant, SVG reactive current changing to a specified value based on pre-fault values, and voltage-controlled current without superimposed initial reactive current component. Among these, "SVG reactive current changing to a specified value based on pre-fault values" means that the SVG (Static Var Generator) reactive current will use the value of the reactive current before the system fault as a reference, and then generate a pre-set, clearly defined change. Under normal power system operation, the SVG will... The SVG outputs a certain amount of reactive current; this current value is the "pre-fault baseline." In power system analysis, the system's operating state changes significantly before and after a fault. Using the pre-fault state as a reference allows for a clearer assessment of the fault's impact on the SVG's reactive current output and subsequent control strategies. For example, if the SVG's reactive current output was 50A before the fault, this 50A is the starting point for subsequent changes. The specified value is a fixed value pre-set by relevant control strategies or operators based on specific power system requirements and operational objectives. This specified value can be increased or decreased. It is important to note that as a reactive power compensation device, the SVG's control only involves reactive current and power regulation and does not include active power control functions.

[0076] The strategy sub-library for wind and solar power includes wind and solar reactive power control strategies and wind and solar active power control strategies. The wind and solar reactive power control strategies include specified reactive current control and specified reactive power control; the wind and solar active power control strategies include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

[0077] The strategy database includes a strategy sub-library for energy storage, which includes energy storage reactive power control strategies and energy storage active power control strategies under charging and discharging modes. The energy storage reactive power control strategies include specified reactive current control and specified reactive power control. The energy storage active power control strategies under charging and discharging modes include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

[0078] To better understand this embodiment, we will now use a strategy sub-library for wind power and photovoltaics as an example. Each strategy can be expressed in formula form using a corresponding voltage-current response characteristic mathematical model. The voltage-current response characteristic mathematical model is in piecewise function form, specifically:

[0079] The specified reactive power strategy is expressed by the following formula:

[0080] ;

[0081] In the formula, To specify reactive power; This is the reactive power proportionality coefficient; This represents the initial reactive power. This is the reactive power setpoint;

[0082] The specified reactive current strategy is expressed by the following formula:

[0083] ;

[0084] In the formula, To specify the reactive current; This is the voltage proportionality coefficient; Set voltage value; for The effective value of the voltage at any given time; This is the initial reactive current; This is the reactive current setpoint;

[0085] The specified active power strategy is expressed by the following formula:

[0086] ;

[0087] In the formula, To specify active power; This is the active power proportionality coefficient; This represents the initial active power. This is the active power setpoint;

[0088] The specified active current strategy is expressed by the following formula:

[0089] ;

[0090] In the formula, To specify the active current; This is the active current proportionality coefficient; This is the initial active current; This is the active current setpoint;

[0091] Under the rule of using the pre-fault current value, the active current is specified to be equal to the initial active current. Therefore, the active current control based on the pre-fault current is expressed by the formula:

[0092] ;

[0093] The low-voltage-limited active current strategy is expressed by the following formula:

[0094] ;

[0095] In the formula, The low-voltage limit corresponding to the maximum permissible active current increment is used to limit the specified active current within this voltage range. This is the high voltage threshold voltage; This is the low-voltage threshold voltage; This is the maximum active current;

[0096] In the control process of new energy equipment (such as wind turbine generators), active and reactive power control are not independent of each other, but are subject to various constraints, such as active power limits, reactive power limits, total current limits, active current limits, and reactive current limits. For example, the total current capacity of the equipment is limited. When the reactive current increases, the active current needs to be reduced accordingly to ensure that the total current does not exceed the limit value. At the same time, active power control is also affected by reactive power control. This means that in actual operation, active power control strategies or reactive power control strategies cannot be considered simply in isolation, but their interrelationships must be comprehensively considered.

[0097] To ensure safe and efficient equipment operation under various operating conditions, reactive power control strategies and active power control strategies need to be combined. Different combinations can adapt to different grid conditions and equipment states. For example, when the grid voltage is low, a specified reactive current strategy may be used to provide sufficient reactive power support, while a low-voltage-limited active current strategy may be combined to limit active power output to ensure stable equipment operation. Conversely, when the grid voltage is normal and maximum active power output is required, a combination of a specified active power strategy and an appropriate reactive power control strategy can be used.

[0098] Therefore, in this embodiment, reactive power control strategies and active power control strategies are combined into multiple sets of strategies. For example, in this embodiment, 8 combinations can be formed by 2 reactive power strategies and 4 active power strategies. The above combinations can establish a strategy database. Expressed as a formula:

[0099] ;

[0100] In the formula, As a strategy combination, in this embodiment ;

[0101] Preferably, to more accurately describe and execute these strategies, the strategy database also includes a set of associated parameters for each strategy, with defined value ranges. By setting these associated parameter sets, the strategies can be flexibly adjusted according to actual operating conditions, ensuring stable and efficient operation of new energy equipment under different working conditions. The associated parameter sets are constructed based on the corresponding mathematical models of the voltage-current response characteristics for each strategy. .

[0102] Taking a specified active current control strategy as an example, the formula is:

[0103] ;

[0104] The mathematical form can then be simplified to a linear form. ,in correspond It has a clear physical meaning (reflecting the regulatory effect of voltage changes on active current), while the constant term... This integrates the initial current correction term and the active current reference value. During the actual parameter identification process, The slope of the response curve can be directly identified, and a value range of [0.5, 1.5] pu can be set to match the device's dynamic response capability. However, All three parameters are included in the constant term in the simplified model. This leads to a redundancy problem in parameter degrees of freedom—that is, different combinations of parameters may produce the same result. To address this issue, this embodiment establishes constraints based on historical device operating data. It is limited to the average active current of the cycle preceding the fault. Constrained within 90%-110% of the rated current, while setting... The rate of change does not exceed 0.2 pu per second. This processing method effectively reduces the degree of freedom in parameter identification while preserving the physical meaning of the parameters.

[0105] To verify the effectiveness and adaptability of the aforementioned strategy library in actual wind turbine failure scenarios, this embodiment uses a simulation platform to verify the accuracy of the strategy database. Based on the eight combinations of strategy databases constructed above, wind turbine failure scenarios are simulated in the simulation environment, yielding the following results: Figures 2-9The simulated wind turbine fault response characteristic curve shown corresponds to two reactive power control strategies and four active power control strategy combinations under eight different strategies. This curve visually demonstrates the control effect of different strategy combinations during wind turbine faults. Figure 2 Strategy 1, which combines specified reactive power control with specified active power control. Figure 3 Strategy 2, which combines specified reactive power control with specified active current control. Figure 4 Strategy 3, which adds active current control before crossing to the specified reactive power control, Figure 5 Strategy 4, which combines reactive power control with low-voltage-limited active current control, Figure 6 Strategy 5, which combines specified reactive current control with specified active current control. Figure 7 Strategy 6, which combines specified reactive current control with specified active current control. Figure 8 Strategy 7: To specify reactive current control plus active current control before crossing. Figure 9 To specify strategy 8, which combines reactive current control with low-voltage limit active current control, ensure that the dynamic response characteristic curves corresponding to the voltage-current response characteristic mathematical model in the strategy database are completely matched between the simulation data and the data.

[0106] S2, such as Figure 13 As shown, a two-layer particle swarm optimization framework, comprising a global optimization layer and a local optimization layer, is constructed based on a policy database, and a hierarchical fitness function is set. This framework utilizes a two-layer particle swarm algorithm to achieve adaptive policy selection and parameter identification. The specific execution process of the two-layer particle swarm algorithm is as follows:

[0107] S21, Data Preprocessing Stage:

[0108] The data in the strategy database is processed, including data cleaning (e.g., removing outliers and filling in missing values) and data standardization (e.g., unifying the data scale; different data in the strategy database may have different units and ranges of values. For example, voltage data may range from hundreds of volts to thousands of volts, while current data may range from a few amperes to tens of amperes. To prevent certain data from dominating the algorithm due to their large range, it is necessary to standardize such data, such as Z-score standardization).

[0109] The initial number of particles and iterations are set for the two-layer particle swarm optimization (PSO) algorithm, and a penalty term is determined. The initial number of particles determines the number of samples searched in the search space, the number of iterations specifies the maximum number of search steps for the algorithm, and the penalty term is set based on various constraints in new energy equipment control (such as active power limits, reactive power limits, total current limits, etc.) to ensure that the two-layer PSO algorithm does not violate these constraints during the search process.

[0110] S22, Optimization Phase:

[0111] After completing the above data preprocessing, based on the pre-designed error evaluation function, collaborative optimization is performed between the global optimization layer and the local optimization layer, with the global optimization layer serving as the upper layer and the local optimization layer as the lower layer.

[0112] S221. The global optimization layer uses the Discrete Particle Swarm Optimization (PSO) algorithm to select the optimal policy combination from the policy database. This layer uses a strategy with decreasing inertial weights to search the feasible region of the parameters. During the iterative process of the PSO algorithm, its core particle update formula is used to guide the particles to find a better solution in the search space. The particle update of the global optimization layer is expressed by the formula:

[0113] ;

[0114] In the formula, express Global velocity at any given moment; express Global velocity at any given moment; Inertial weight; and This is the learning factor, which is usually set to a fixed value; and In order to be in Random numbers within; This represents the optimal position in the particle's history. The globally optimal position; This represents the particle's current position.

[0115] In this process, the Discrete Particle Swarm Optimization (DPSO) algorithm uses the policy error function as the fitness function to evaluate different policy combinations in the policy database. The policy error function measures the quality of the current policy combination, and the optimal policy combination is obtained based on the policy error function, expressed by the formula:

[0116] ;

[0117] In the formula, The optimal strategy combination; The initial sample size for evaluating the effectiveness of the strategy; For the first The predicted current value corresponding to each sample; For the first The effective current value corresponding to each sample; Indicated in the policy database The goal is to find the strategy that minimizes the strategy error function among all strategy combinations, which is the strategy combination for each particle. According to strategy combination Corresponding policy error function ,Right now The fitness value is calculated; a smaller fitness value indicates a smaller error and better performance for the strategy combination. In practice, this needs to be determined based on the specific strategy combination. The wind power generation system was simulated to obtain the predicted current value. and the effective value of the current Compare the results and calculate the sum of squares of the errors.

[0118] Although the policy error function can initially assess the merits of policy combinations, iterative searches are required to find the true optimal solution due to factors such as the size of the policy database and initial information, ultimately determining the optimal combination of reactive and active power control policies. The global optimization layer is the key step in the entire optimization process; it filters possible optimal policy combinations from the large search space of the policy database, providing a foundation for subsequent local dynamic adjustments.

[0119] S222. The local optimization layer uses a continuous particle swarm optimization algorithm to continuously optimize the parameter set associated with the optimal strategy combination. This layer corrects the particle velocity in real time based on parameter error feedback, expressed by the following formula:

[0120] ;

[0121] In the formula, for Local velocity at a given moment; for Local velocity at a given moment; The learning rate; Momentum factor; For the optimal strategy combination, the associated parameters The corresponding parameter error function; The optimal strategy combination is associated with parameters;

[0122] The local optimization layer optimizes the association parameters of the optimal strategy combination using a parameter error function as the fitness function. The parameter error function measures the difference between the actual operating effect and the target state under the optimal strategy combination association parameters, and is expressed by the formula:

[0123] ;

[0124] In the formula, In order to comprehensively evaluate the number of data samples, in this embodiment, The number of samples collected at different time points; for The effective value of the current at any given time;

[0125] By incorporating the parameter error function, improved PSO optimization is performed on the association parameters of the selected optimal strategy combination. This continuous optimization yields the optimal association parameters for the optimal strategy combination, expressed as a formula:

[0126] ;

[0127] In the formula, Associating optimal parameters with the optimal strategy combination; This represents the wind turbine control model, based on the optimal strategy combination. Optimal strategy combination associated parameters and external input at time The predicted current value of the wind turbine at that moment is calculated.

[0128] The local optimization layer focuses on fine-tuning the associated parameters of the optimal strategy combination selected by the global optimization layer. Based on real-time parameter error feedback, it adaptively adjusts the particle velocity, thereby optimizing the associated parameters. Through continuous iterative optimization, it achieves accurate identification of the optimal associated parameters, enabling the control strategy to better adapt to the actual situation and reach the optimal strategy-parameter pair.

[0129] S3. Obtain actual unit test data, extract features from the actual unit test data to obtain test feature data, and input the test feature data into a two-layer particle swarm optimization framework to obtain the optimal policy-parameter pair. Specifically:

[0130] S31. In this embodiment, the actual unit is a wind turbine. Test data of the wind turbine is obtained through semi-physical testing or on-site testing. In semi-physical testing, part of the system uses actual physical equipment, while the other part is simulated by computer simulation models. For example, when testing a wind turbine, an actual generator and controller can be used, while the power grid part uses a simulation model to simulate different fault conditions. The operating data of the unit under different fault ride-through conditions obtained in this way is the semi-physical test data. On-site testing is to conduct fault ride-through testing on the unit at the actual power production site.

[0131] The actual unit test data is expressed by the following formula:

[0132] ;

[0133] In the formula, This refers to actual unit test data, specifically wind turbine test data in this embodiment. for The effective value of active current at any given time; for The effective value of reactive current at time t.

[0134] Feature extraction is performed on the actual unit test data to remove redundant information that does not directly affect the analysis of the unit's fault ride-through performance. Only the key data that best reflects the fault ride-through characteristics of the unit is retained. Specifically, in this embodiment, a fault partitioning method is used to divide the actual unit test data into a low-voltage ride-through region, a steady-state region, and a high-voltage ride-through region. The effective values ​​of active current, reactive current, and voltage are extracted from each region, and the extracted values ​​are used as test feature data. ;

[0135] S32. Input the test feature data into the algorithm. Under the collaborative optimization framework of the global optimization layer and the local optimization layer, the upper layer performs a preset number of iterations. In this implementation, the preset number of iterations is 6. In each upper layer iteration, the strategy combination is adjusted or the search direction is changed to expand the search space and find a better strategy combination.

[0136] Preferably, each time the upper loop is executed, the lower loop will also be used to optimize the combination of related parameters.

[0137] Through continuous collaborative optimization loops between upper and lower layers, the total error index of each policy-parameter pair is calculated, expressed by the formula:

[0138] ;

[0139] In the formula, This is the total error index; For the optimal strategy combination The corresponding policy error function; Associating optimal parameters with the optimal strategy combination The corresponding parameter error function;

[0140] Preferably, in order to maintain consistency in magnitude, this embodiment also provides another formula for calculating the total error index:

[0141] ;

[0142] ;

[0143] In the formula, and The weighting coefficient for the total error index;

[0144] When the total error index is less than a preset error threshold, the optimization is considered complete, and the current policy-parameter pair is output as the optimal policy-parameter pair; otherwise, the optimization is restarted. For example, in this embodiment, the preset error threshold is 2%. Once the optimization is complete, the optimal strategy-parameter pair is obtained.

[0145] S33. Preferably, this embodiment further includes simultaneously plotting a stepped voltage curve when extracting features from actual unit test data, and when a sudden current change is detected, such as... Figures 10-11 As shown, the voltage threshold of the stepped voltage curve is used as the high-low pass threshold. Figure 10 For low penetration threshold determination, Figure 11 To determine the high penetration threshold, the piecewise function interval boundary of the voltage-current response characteristic mathematical model in the strategy database is defined by the high and low penetration thresholds.

[0146] S34. This embodiment also includes, when extracting features from the test data, using fault ride-through test data under different reactive power initial values ​​(such as...) Figure 12 As shown in the figure, first analyze whether the reactive current changes are consistent with different initial reactive values ​​during the fault period. That is, by comparing the trend and magnitude of the reactive current changes over time with different initial reactive values, we can determine whether the changes are consistent.

[0147] If a direct comparison alone is insufficient for accurate determination, correlation analysis can be used to precisely identify the current superposition mode. This involves calculating the correlation coefficient between reactive current changes under different initial reactive current values ​​and making a determination based on the magnitude of the correlation coefficient.

[0148] When the correlation coefficient of reactive current changes with different initial reactive values ​​during the fault period is greater than the preset correlation coefficient threshold, it indicates that the reactive current changes are highly consistent and is judged to be an incremental mathematical model.

[0149] When the correlation coefficient of reactive current changes with different initial reactive values ​​during the fault period is less than or equal to the preset correlation coefficient threshold, it indicates that the consistency of reactive current changes is low and is judged as an absolute value type mathematical model.

[0150] The current superposition mode is used to control the optimal strategy combination selection process of the global optimization layer, and only activates the corresponding type of mathematical model in the strategy database.

[0151] S4. To verify the effectiveness and accuracy of the method in this embodiment, a simulation model is constructed using the optimal strategy-parameter pair obtained in step S3, and the simulation data generated by the simulation model is processed. Feature extraction is performed to obtain simulation feature data. .

[0152] Error verification is performed on the test feature data and simulation feature data. The error verification uses mean squared error (MSE), mean absolute error (MAE), or relative error. In this embodiment, relative error is used for error verification to obtain the error verification results. The error verification results are then visualized, including plotting line graphs, bar charts, and other graphs to intuitively show the comparison between the test feature data and simulation feature data, as well as the error distribution. In this embodiment, for example... Figure 14 As shown, the actual data was tested under two operating conditions: a small operating condition and a large operating condition. The small operating condition refers to a given initial active power of 0.2Pn and a large operating condition refers to a given initial active power of 0.9Pn, where Pn is the rated power. Figure 14 (a) and Figure 14 (c) forms a group, consisting of reactive power control identification and verification under low-load conditions and active power control identification and verification. Figure 14 (b) and Figure 14 (d) is a group consisting of reactive power control identification and verification under large operating conditions and active power control identification and verification. By comparing the differences between test data and simulation data, the results are visualized. The point error of the simulation feature data tracking the test feature data is within 0.02, which proves the effectiveness of the method in this embodiment and verifies the accuracy of the simulation model.

[0153] It is worth noting that the above-mentioned control strategy identification methods and steps are universal. SVG simulation modeling only requires referring to the above steps to identify reactive power control strategies and parameters; energy storage simulation modeling needs to consider active and reactive power control strategies under charging and discharging modes respectively. The mathematical model expressions of the strategies under different modes are consistent with the control strategies of wind power and photovoltaics, and do not affect the identification steps and methods. Therefore, the parameter identification process for SVG and energy storage will not be repeated.

[0154] Example 2:

[0155] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in any embodiment of the present invention.

[0156] Example 3:

[0157] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in any embodiment of the present invention.

[0158] It is worth noting that the electronic device and computer-readable storage medium described in this invention are based on the same inventive concept as the method described in Embodiment 1 of this invention, and will not be described again here.

[0159] In this embodiment of the invention, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0160] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0161] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0162] In several embodiments provided by this invention, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0163] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An adaptive identification method for new energy control parameters based on two-layer particle swarm optimization, characterized in that, The method includes: By integrating the typical control strategies of four types of heterogeneous new energy units, namely wind power, photovoltaic, energy storage and SVG, a strategy database is constructed that includes a mathematical model of voltage-current response characteristics with multiple different typical control strategies, the value range of parameters associated with each control strategy and the initial parameter values. The mathematical model of voltage-current response characteristics is in the form of a piecewise function. Obtain actual unit test data, extract features from the test data, and obtain test feature data; A two-layer particle swarm optimization framework, comprising a global optimization layer and a local optimization layer, is constructed based on a policy database, and a hierarchical fitness function is set. The test feature data is input into a two-layer particle swarm optimization framework, which outputs the optimal policy-parameter pair. The global optimization layer employs a discrete particle swarm optimization algorithm, using the policy error function as the fitness function to select the optimal policy combination. The policy error function is expressed by the following formula: ； In the formula, The optimal strategy combination; The initial sample size for evaluating the effectiveness of the strategy; For the first The predicted current value corresponding to each sample; For the first The effective current value corresponding to each sample; Indicated in the policy database Find the strategy that minimizes the policy error function among all strategy combinations. ; For strategy combination The corresponding policy error function; the local optimization layer uses a continuous particle swarm optimization algorithm, with the parameter error function as the fitness function to optimize the association parameters of the optimal policy combination, expressed by the formula: ； In the formula, Associating optimal parameters with the optimal strategy combination; This represents the unit control model, based on the optimal strategy combination. Optimal strategy combination associated parameters and external input at time Calculate the unit The predicted current value at time; for The effective value of the current at any given time; To comprehensively evaluate the number of data samples; Indicates in the associated parameter set Find the correlation parameter that minimizes the parameter error function value. .

2. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The strategy database contains a strategy sub-library of SVG, which includes SVG reactive power control strategies, specifically including voltage-controlled current with superimposed initial reactive current component, specified reactive current control, specified reactive power control, SVG current remaining unchanged, SVG reactive current changing to a specified value based on the pre-fault condition, and voltage-controlled current without superimposed initial reactive current component.

3. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The strategy database includes strategy sub-libraries for wind power and photovoltaics, which include wind and solar reactive power control strategies and wind and solar active power control strategies. The wind and solar reactive power control strategies include specified reactive current control and specified reactive power control; the wind and solar active power control strategies include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

4. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The strategy database includes a strategy sub-library for energy storage, which includes energy storage reactive power control strategies and energy storage active power control strategies under charging and discharging modes. The energy storage reactive power control strategies include specified reactive current control and specified reactive power control. The energy storage active power control strategies under charging and discharging modes include specified active power control, specified active current control, active current control based on pre-fault conditions, and low-voltage limit active current control.

5. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The method further includes: After constructing the strategy database, the accuracy of the strategy database is verified through a simulation platform to ensure that the simulation data and the dynamic response characteristic curves corresponding to the voltage-current response characteristic mathematical model in the strategy database are completely matched.

6. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The method further includes: When extracting features from actual unit test data, a stepped voltage curve is plotted simultaneously. When a sudden change in current is detected, the voltage critical value of the stepped voltage curve is used as the high and low breakdown threshold. The high and low breakdown threshold is used to define the boundary of the piecewise function interval of the mathematical model of voltage-current response characteristics in the strategy database.

7. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 6, characterized in that, The method further includes: When extracting features from the test data, the fault ride-through test data under different initial reactive power values ​​are analyzed to determine whether the reactive current changes during the fault are consistent, and the current superposition mode is determined based on the correlation. Specifically: By comparing the trend and magnitude of reactive current changes over time with different initial reactive values, we can determine whether the changes are consistent and calculate the correlation coefficient between reactive current changes under different initial reactive values. When the correlation coefficient of reactive current changes with different initial reactive values ​​during the fault period is greater than the preset correlation coefficient threshold, it is determined to be an incremental mathematical model. When the correlation coefficient of reactive current change with different initial reactive values ​​during the fault period is less than or equal to the preset correlation coefficient threshold, it is determined to be an absolute value mathematical model. The current superposition mode is used to control the optimal strategy combination selection process of the global optimization layer, and only activates the corresponding type of mathematical model in the strategy database.

8. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The feature extraction of the test data adopts the fault partitioning method, which divides the test data into low voltage ride-through zone, steady state zone and high voltage ride-through zone. The effective values ​​of active current, reactive current and voltage are extracted in each zone and the extracted values ​​are used as test feature data.

9. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The method further includes constructing an association parameter set by combining the value ranges and initial parameter values ​​of the associated parameters of each control strategy. .

10. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The particle update in the iterative process of the discrete particle swarm optimization algorithm is expressed by the following formula: ； In the formula, express Global velocity at any given moment; express Global velocity at any given moment; Inertial weight; and For learning factors; and In order to be in Random numbers within; This represents the optimal position in the particle's history. The globally optimal position; This represents the particle's current position. During the particle update process, the Discrete Particle Swarm Optimization (DPSO) algorithm evaluates different policy combinations in the policy database using a policy error function. The policy error function measures the quality of the current policy combination, and the optimal policy combination is obtained based on the policy error function.

11. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The optimization process of the continuous particle swarm optimization algorithm is as follows: The local optimization layer corrects particle velocities in real time based on parameter error feedback, expressed by the following formula: ； In the formula, for Local velocity at a given moment; for Local velocity at a given moment; The learning rate; Momentum factor; For the optimal strategy combination, the associated parameters The corresponding parameter error function; The optimal strategy combination is associated with parameters; The parameter error is expressed by the formula: ； In the formula, for The effective value of the voltage at any given time; The parameters associated with the selected optimal strategy combination are optimized by taking into account the parameter error, so as to obtain the optimal strategy combination associated with the optimal parameters.

12. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, Output the optimal policy-parameter pair, specifically: In the two-layer particle swarm optimization framework, the global optimization layer performs a preset number of iterations, and after the policy optimization iteration of the global optimization layer, the parameter optimization iteration of the local optimization layer is performed. Through the two-layer collaborative optimization iteration, the total error index of each policy-parameter pair is calculated, expressed by the formula: ； In the formula, This is the total error index; For the optimal strategy combination The corresponding policy error function; Associating optimal parameters with the optimal strategy combination The corresponding parameter error function; When the total error index is less than the preset error threshold, the optimization is considered complete, and the current policy-parameter pair is output as the optimal policy-parameter pair; otherwise, the optimization is restarted.

13. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The method further includes constructing a simulation model based on the optimal strategy-parameter pair, extracting features from the simulation data generated by the simulation model to obtain simulation feature data, and performing error verification on the test feature data and simulation feature data. The error verification adopts mean square error (MSE), mean absolute error (MAE), or relative error.

14. The adaptive identification method for new energy control parameters based on two-layer particle swarm optimization according to claim 1, characterized in that, The method also includes visualizing the error verification results.

15. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in any one of claims 1 to 14.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the adaptive identification method for new energy control parameters based on two-layer particle swarm optimization as described in any one of claims 1 to 14.

Images