Inverter single machine electromechanical modeling parameter identification method and device

Abstract

The invention discloses an inverter single machine electromechanical modeling parameter identification method and device, and the method comprises the steps: obtaining the actual measurement data or semi-physical test data of the fault ride-through of an inverter of a target model; processing the measured data or the semi-physical test data according to a preset voltage threshold to generate a traversing data matrix; establishing a corresponding control strategy according to the traversing data matrix, and generating a to-be-identified control parameter vector; constructing a least square method objective function, and fitting a predicted value under the control strategy with a real output value to obtain an initial control parameter vector; introducing an elastic network regularization item into the least square method objective function for optimization; and solving and iterating the optimized least square method objective function by adopting a near-end gradient descent method, and outputting a final control parameter vector. And the modeling parameters are identified through the optimized least square method, so that the identification accuracy of the key control parameters is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic inverters, in particular to an inverter single machine electromechanical modeling parameter identification method and device. BACKGROUND

[0002] The photovoltaic power station needs to carry out power quality test in the process of grid connection, including active / reactive power control capability test, primary frequency modulation and inertia response test, inverter and SVG (static var generator) performance test, fault ride-through capability verification, and power station whole station simulation modeling, etc., wherein the fault ride-through capability verification needs to compare the inverter single machine measured data or semi-physical data with the electromechanical model data, and the power station whole station simulation modeling needs to build the electromechanical model of the whole photovoltaic power station on the basis of single machine modeling.

[0003] As the confidentiality requirements of manufacturers on control strategies and parameters are gradually increasing, the accurate modeling of inverter single machine is also gradually increasing in difficulty, which affects the authenticity of the comparison between the fault ride-through measured data or semi-physical data and the modeling parameters. The existing common scheme is based on the fault ride-through measured data or semi-physical data, and uses the least square method to complete the control logic parameter identification during the inverter ride-through, and then uses the identified control logic parameters to complete the inverter single machine fault ride-through model construction and comparison with the fault ride-through measured data or semi-physical data, to verify whether the error results meet the requirements of the photovoltaic system technology related standards. However, the common scheme has the following deficiencies: the least square method and part of the optimization algorithm still have the problems of overfitting or insufficient analysis ability for specific "ill-conditioned" data, and therefore, there is an urgent need for a method capable of improving the accuracy of single machine electromechanical modeling parameter identification. SUMMARY

[0004] Embodiments of the present application provide an inverter single machine electromechanical modeling parameter identification method and device to improve the accuracy of modeling parameter identification and the ability to process specific data.

[0005] To solve the above technical problems, embodiments of the present application disclose the following technical solutions: In a first aspect, the present application provides an inverter single machine electromechanical modeling parameter identification method, comprising: obtaining measured data or semi-physical test data of fault ride-through of a target type inverter; processing the measured data or semi-physical test data according to a preset voltage threshold to generate a ride-through data matrix; establishing a corresponding control strategy according to the ride-through data matrix to generate a to-be-identified control parameter vector; constructing a least square method objective function to fit the predicted value under the control strategy with the real output value to obtain an initial control parameter vector; introducing an elastic network regularization term into the least square method objective function for optimization; and using a proximal gradient descent method to solve the least square method objective function after iteration optimization to output a final control parameter vector.

[0006] Further, in the above method, the measured data or semi-physical test data includes low voltage ride through data and high voltage ride through data, the preset voltage threshold includes a low voltage ride through threshold and a high voltage ride through threshold, and the ride through data matrix includes a low ride through data matrix and a high ride through data matrix.

[0007] In the above further solution, preferably, the low voltage ride through threshold is preset as 0.9 p.u., and the high voltage ride through threshold is preset as 1.1 p.u.

[0008] Preferably, the low voltage ride through data includes three-phase symmetrical faults and two-phase asymmetrical faults, and the positive sequence voltage during the faults is 0 p.u., 0.2 p.u., 0.35 p.u., 0.5 p.u., and 0.75 p.u., respectively; and the high voltage ride through data includes three-phase symmetrical faults and two-phase asymmetrical faults, and the positive sequence voltage during the faults is 1.2 p.u., 1.25 p.u., and 1.30 p.u., respectively.

[0009] Further, the measured data or semi-physical test data is stored in a six-column matrix form, which includes a time label, a positive sequence voltage, an active power, a reactive power, an active current, and a reactive current, respectively.

[0010] Further, the elastic network regularization term includes L1 regularization and L2 regularization, and the weight ratio of the L1 regularization and the L2 regularization is dynamically adjusted according to the voltage ride through working condition.

[0011] Still further, the proximal gradient descent method includes the following steps: initializing a control parameter vector, calculating a fixed term gradient, adopting a soft threshold function for the L1 regularization part, performing spherical projection for the L2 regularization part, updating the parameter vector and performing termination judgment.

[0012] Further, the data acquisition frequency of the measured data or semi-physical test data is not less than 1000 Hz, and all are in per unit format.

[0013] In a second aspect, the application provides an inverter single machine electromechanical modeling parameter identification device, which includes: a data acquisition module, configured to acquire fault ride through measured data or semi-physical test data of a photovoltaic inverter; a data processing module, configured to process the measured data according to a preset voltage threshold to generate a ride through data matrix; a control strategy module, configured to construct a corresponding control strategy according to the ride through data matrix to generate a to-be-identified control parameter vector; a parameter fitting module, configured to construct a least square method objective function and fit a predicted value and an actual output value to generate an initial control parameter vector; and a parameter optimization module, configured to introduce an elastic network regularization term into the least square method objective function, and iteratively optimize through a proximal gradient descent method to output a final control parameter vector.

[0014] Further, the parameter optimization module includes a soft threshold operation unit for processing L1 regularization and a spherical projection unit for processing L2 regularization.

[0015] The above technical solutions have at least the following advantages or beneficial effects: by introducing the least square method of elastic network optimization in the inverter single machine mechanical modeling parameter identification, the overfitting and ill-conditioned data sensitivity problems prone to occur in the traditional least square method can be effectively avoided, and the accurate identification of key control parameters is realized.

[0016] In the above technical solutions, by distinguishing the high and low voltage ride-through conditions and dynamically adjusting the regularization ratio, the accuracy and stability of parameter identification are ensured, the adaptability of data under different fault ride-through conditions is improved, reliable data support is provided for the whole station modeling and network verification of photovoltaic power stations, and the process, convenience and high robustness of the modeling parameter identification method are improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The modeling parameter identification method provided by the present application is shown in the schematic diagram of the basic process; Figure 2 The example proximal gradient descent solving flowchart provided by the present application is shown in the schematic diagram of the example proximal gradient descent solving flowchart; Figure 3 The least square method low voltage ride-through parameter identification result provided by the present application is shown in the schematic diagram of the least square method low voltage ride-through parameter identification result, wherein Figure 3 - (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 3 - (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 4 The optimized low voltage ride-through parameter identification result provided by the present application is shown in the schematic diagram of the optimized low voltage ride-through parameter identification result, wherein Figure 4 - (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 4 - (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 5 The least square method high voltage ride-through parameter identification result provided by the present application is shown in the schematic diagram of the least square method high voltage ride-through parameter identification result, wherein Figure 5 - (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 5 - (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 6 The optimized high voltage ride-through parameter identification result provided by the present application is shown in the schematic diagram of the optimized high voltage ride-through parameter identification result, wherein Figure 6 - (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 6 - (b) is a voltage drop degree-active power scatter plot during fault ride-through. Detailed Implementation

[0018] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, further clarifies the application. It should be understood that the specific embodiments described in this specification are merely for explaining the application and are not intended to limit it. Those skilled in the art should recognize that the application can be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary detail from hindering the description of this application.

[0019] In the description of this application, it should be noted that, unless otherwise expressly defined, the terms "comprising" and "including" should be interpreted broadly, meaning that the listed elements or steps may be some or all, and may also include other elements or steps not listed; the terms "based on" and "according to" should be understood as "at least partially based on" or "at least partially based on"; the terms "step" and "process" may be performed sequentially, or may be performed in parallel under certain conditions, or in a different order. Those skilled in the art can understand the specific meaning of the above terms in the method of this application according to the specific application scenario.

[0020] In the description of this application, it should also be noted that "semi-physical testing" refers to obtaining the operating data of the inverter under specific operating conditions through HIL (Hardware-in-the-Loop) testing or a combination of some real equipment and simulation platforms.

[0021] To address the problems of inaccurate control parameter acquisition, overfitting susceptibility of traditional least squares methods, and sensitivity to ill-conditioned data in existing inverter single-unit electromechanical modeling, this application proposes a parameter identification method for inverter single-unit electromechanical modeling using elastic network optimization of least squares. This method acquires and processes measured data of the inverter under fault ride-through conditions, establishes a control strategy and generates a vector of control parameters to be identified, combines least squares fitting with elastic network regularization optimization, and iteratively solves the problem using the proximal gradient descent method to ultimately obtain control parameters with higher accuracy and stronger robustness. The specific steps and implementation process of this invention are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 A schematic diagram illustrating the basic process of the modeling parameter identification method provided in this application. Figure 1As shown, the present application provides an exemplary inverter single machine mechanical modeling parameter identification method, including the following steps: S1: obtaining the measured data or semi-physical test data of the fault ride-through of the target type inverter; S2: processing the measured data or semi-physical test data according to the preset voltage threshold to generate a ride-through data matrix; S3: establishing a corresponding control strategy according to the ride-through data matrix to generate a to-be-identified control parameter vector; S4: constructing a least squares objective function, fitting the predicted value under the control strategy with the real output value, and obtaining an initial control parameter vector; S5: introducing an elastic network regularization term into the least squares objective function for optimization; S6: solving and iteratively optimizing the least squares objective function by using a proximal gradient descent method, and outputting a final control parameter vector.

[0023] The present application provides the above method steps, and the following data forms and fault conditions are taken as examples: the measured data or semi-physical test data of the fault ride-through of the target type inverter includes low voltage ride-through data and high voltage ride-through data, and the low voltage ride-through data and the high voltage ride-through data are both stored in the form of a six-column matrix, and each column of the matrix is respectively a time label, a positive sequence voltage, an active power, a reactive power, an active current and a reactive current; the preset voltage threshold includes a low voltage ride-through threshold and a high voltage ride-through threshold; the ride-through data matrix includes a low ride-through data matrix and a high ride-through data matrix; the data acquisition frequency of the measured data or semi-physical test data is not less than 1000 Hz, and both are in the per unit format; the low voltage ride-through data includes three-phase symmetric faults and two-phase asymmetric faults, and the positive sequence voltage during the fault is 0 p.u., 0.2 p.u., 0.35 p.u., 0.5 p.u. and 0.75 p.u. respectively; the high voltage ride-through data includes three-phase symmetric faults and two-phase asymmetric faults, and the positive sequence voltage during the fault is 1.2 p.u., 1.25 p.u. and 1.30 p.u. respectively.

[0024] In the above step S1, taking the three-phase 0 p.u. ride-through original data matrix A as an example of the obtained measured data or semi-physical test data, there can be: A= , Wherein, the first column is the time label, the second column of the matrix is the positive sequence voltage data, the third column is the active power data, the fourth column is the reactive power data, the fifth column is the active current data, and the sixth column is the reactive current data; there are n sampling points in total.

[0025] In the above step S2, the preset low voltage ride-through threshold is 0.9 p.u., and the preset high voltage ride-through threshold is 1.1 p.u., which can be adjusted according to the actual situation. The exemplary low ride-through data matrix has the following format: wherein, according to different ride-through conditions, the data of each row means as follows: is the steady-state positive-sequence voltage per unit before three-phase symmetric 0 p.u. ride-through, is the steady-state active power per unit before three-phase symmetric 0 p.u. ride-through, is the steady-state reactive power per unit before three-phase symmetric 0 p.u. ride-through; is the steady-state positive-sequence voltage per unit during three-phase symmetric 0 p.u. ride-through, is the steady-state active power per unit during three-phase symmetric 0 p.u. ride-through, is the steady-state reactive power per unit during three-phase symmetric 0 p.u. ride-through, and so on. is the steady-state positive-sequence voltage per unit before two-phase 0 p.u. ride-through, is the steady-state active power per unit before two-phase 0 p.u. ride-through, is the steady-state reactive power per unit before two-phase 0 p.u. ride-through; is the steady-state positive-sequence voltage per unit during two-phase 0 p.u. ride-through, is the steady-state active power per unit during two-phase 0 p.u. ride-through, is the steady-state reactive power per unit during two-phase 0 p.u. ride-through, and so on. The format is as follows: wherein, according to different ride-through conditions, the data of each row means as follows: is the steady-state positive-sequence voltage per unit before three-phase symmetric 1.2 p.u. ride-through, is the steady-state active power per unit before three-phase symmetric 1.2 p.u. ride-through, is the steady-state reactive power per unit before three-phase symmetric 1.2 p.u. ride-through; is the steady-state positive-sequence voltage per unit during three-phase symmetric 1.2 p.u. ride-through, is the steady-state active power per unit during three-phase symmetric 1.2 p.u. ride-through, is the steady-state reactive power per unit during three-phase symmetric 1.2 p.u. ride-through, and so on. is the steady-state positive-sequence voltage per unit before two-phase 1.2 p.u. ride-through, is the steady-state active power per unit before two-phase 1.2 p.u. ride-through, is the steady-state reactive power per unit before two-phase 1.2 p.u. ride-through; is the steady-state positive-sequence voltage per unit during two-phase 1.2 p.u. ride-through, This represents the per-unit value of the steady-state active power during the two-phase 1.2 pu ride-through process. This is the per-unit value of steady-state reactive power during the two-phase 1.2 pu ride-through process, and so on.

[0026] This manual also provides an explanation of how the above-mentioned low-speed data matrix is ​​calculated. Using the above-mentioned low-speed data matrix... Taking any row as an example, first compare the positive sequence voltage value in the second column of matrix A in step S1 with the low voltage crossing threshold. Find the first and last ones that are below the low voltage ride-through threshold. The positive sequence voltage value is recorded as and Then there is to determine to The positive sequence voltage before crossing. to This represents the positive sequence voltage during the crossing process. to This is the positive sequence voltage after crossing. to The active power before crossing, to The active power during the crossing. to The active power after crossing to The reactive power before crossing to This refers to the reactive power during the crossing process. to The reactive power after crossing; secondly, the initial search interval length is set. If it exists Simultaneously satisfying: , , Then record , , They are respectively , , If it does not exist If the above conditions are met simultaneously, then the length of the initial search interval should be reduced. Record the result until it is satisfied. , , They are respectively , , In the above formula, Set the deviation value; set the initial search interval length. If it exists Meanwhile, the following conditions are met: , , , the initial search interval length is recorded as , , , respectively , , ; if there is no , the initial search interval length is reduced until the above conditions are met, and the initial search interval length at this time is recorded as , , , respectively , , ; in the above formula, is the deviation setting value; the low-penetration data matrix is calculated in the same way as the rest of the rows of data, which will not be repeated here; the high-penetration data matrix is calculated in the same way as the low-penetration data matrix, which will not be repeated here.

[0027] The present application provides the following exemplary control strategy according to the above step S3, and generates the to-be-identified control parameter vector according to the exemplary control strategy.

[0028] Taking low penetration as an example, the corresponding control strategy is set as: ; .

[0029] In the above formula, is the steady-state reactive current given value during fault penetration, is the reactive control voltage coefficient, is the reactive control current coefficient, is the steady-state positive sequence voltage value during fault penetration, is the steady-state value of reactive current before fault penetration; is the reactive current deviation value during fault penetration; is the steady-state active current given value during fault penetration, is the active control voltage coefficient, is the active control current coefficient, is the steady-state positive sequence voltage value during fault penetration, is the steady-state value of active current before fault penetration; is the active current deviation value during fault penetration. Taking three-phase symmetric 0 p.u. penetration as an example, at this time in the above objective function; ; ; ; A set of to-be-identified control parameter vectors under each working condition of low voltage ride through is identified through a corresponding low ride through control strategy, including reactive power control voltage coefficient , reactive power control current coefficient , reactive power current deviation value during fault ride through ; active power control voltage coefficient , active power control current coefficient , active power current deviation value during fault ride through .

[0030] Taking high ride through as an example, the corresponding control strategy is set as: ; .

[0031] In the above formula, is a given value of steady-state reactive current during fault ride through, is a reactive power control voltage coefficient, is a reactive power control current coefficient, is a steady-state positive sequence voltage value during fault ride through, is a steady-state value of reactive current before fault ride through; is a reactive power current deviation value during fault ride through; is a given value of steady-state active current during fault ride through, is an active power control voltage coefficient, is an active power control current coefficient, is a steady-state positive sequence voltage value during fault ride through, is a steady-state value of active current before fault ride through; is an active power current deviation value during fault ride through. Taking three-phase symmetric 1.20 p.u. ride through as an example, at this time ; ; ; ; A set of to-be-identified control parameter vectors under each working condition of high voltage ride through is identified through a corresponding high ride through control strategy, including reactive power control voltage coefficient , reactive power control current coefficient , reactive power current deviation value during fault ride through ; active power control voltage coefficient , active power control current coefficient , active power current deviation value during fault ride through .

[0032] The present application provides the following exemplary least square method objective function according to the above step S4, and obtains an initial control parameter vector.

[0033] The above example of low penetration, high penetration control strategy is set as an example, the least square method objective function, input, output, as follows: ; , Wherein, is the real output vector corresponding to the low penetration control strategy obtained in step S3, and has: , Where are the steady-state given values of reactive power and active current during low voltage fault ride-through under three-phase symmetrical fault, and the steady-state given values of reactive power and active current during low voltage fault ride-through under two-phase fault.

[0034] According to step S3, it can be obtained that: ; ; ; .

[0035] is the input vector, which can be represented as: , In the above formula, are the steady-state values of positive sequence voltage, reactive current and active current during low voltage fault ride-through under three-phase symmetrical fault, and the steady-state values of positive sequence voltage, reactive current and active current before fault ride-through under two-phase fault.

[0036] According to step S3, it can be obtained that: ; ; ; ; ; .

[0037] is the control parameter vector of low voltage ride-through, which can be represented as: , corresponds to a set of reactive power control voltage coefficients under each condition of low voltage ride-through in step S3 , reactive power control current coefficient , reactive current deviation value during fault ride-through ; active control voltage coefficient , active control current coefficient , active current deviation value during fault ride-through .

[0038] is the real output vector of low voltage ride through; is the control parameter vector of a set of low voltage ride through , the predicted fitting value vector and the real output value vector The least square error of the real output value vector is the control parameter vector of the low voltage ride through is the identified initial low ride control parameter; Similarly, is the real output vector corresponding to the high ride control strategy obtained in step S3, and has: , wherein are the steady-state given values of reactive and active currents during high voltage fault ride-through under different working conditions of three-phase symmetry and two-phase fault.

[0039] According to step S3, it can be obtained that: ; ; ; .

[0040] is the input vector, which can be represented as:

[0041] , wherein are the steady-state values of positive sequence voltage during high voltage fault ride-through under different working conditions of three-phase symmetry and two-phase fault, and the steady-state values of reactive and active currents before fault ride-through under different working conditions of three-phase symmetry and two-phase fault.

[0042] According to step S3, it can be obtained that: ; ; ; ; ; .

[0043] For the control parameter vector of low voltage ride through, it can be expressed as: A set of reactive power control voltage coefficients corresponding to each working condition of high voltage ride through in step S3 , reactive power control current coefficients , reactive current deviation values during fault ride through ; active power control voltage coefficients , active power control current coefficients , active current deviation values during fault ride through .

[0044] For the real output vector of high voltage ride through; For a set of control parameter vectors of high voltage ride through , the least square error between the predicted fitting value vector and the real output value vector ; In order to minimize the above least square error, the control coefficient vector of high voltage ride through is obtained.

[0045] The application gives an exemplary elastic network regularization term introduction process according to the above step S5.

[0046] Taking the least square method objective function in the above step S4 as an example, the following exemplary elastic network is used for optimization: . , In the above formula, , are the overall regularization strengths of the low voltage ride through and high voltage ride through parameter identification functions respectively, , are the weight proportion adjustment coefficients of L1 regularization and L2 regularization of the low voltage ride through and high voltage ride through parameter identification functions respectively; , are the L1 regularization penalty terms of the low voltage ride through and high voltage ride through parameter identification functions respectively; and are the L2 regularization penalty terms of the low voltage ride through and high voltage ride through parameter identification functions respectively.

[0047] The application gives an exemplary proximal gradient descent method to solve the above exemplary optimized least square objective function according to the above step S6. Figure 2 The application gives an exemplary proximal gradient descent solving flowchart. Combined withFigure 2 The exemplary proximal gradient descent solving step of the embodiment includes: S601: initialize the first group With , set the learning rate (step size) a.

[0048] S602: set the fixed term gradient calculation method of low and high penetration of least squares, in the specific way as follows: ; , In the above formula, , are the transposes of , respectively.

[0049] S603: set the gradient of the regularization term of low and high voltage penetration, wherein the L1 regularization part and the L2 regularization part are as follows, Low penetration L1 part: ; Low penetration L2 part: ; High penetration L1 part: ; High penetration L2 part: ; In the above formula, is a basic function for judging the positive and negative of a value.

[0050] S604: carry out proximal operation on L1 regularization and L2 regularization, for low and high penetration L1 regularization, use soft threshold function, as follows, ; , In the above formula, denotes projection, is the input of the proximal operator, which is used to process the non-differentiable part; for low and high penetration L2 regularization, project onto the sphere, as follows: ; .

[0051] S605: update and , the specific formula is as follows: ; , In the above formula, and is the result of the hth low-voltage ride-through, high-voltage ride-through parameter iteration, With is the result of the (h+1)th low-voltage ride-through, high-voltage ride-through parameter iteration.

[0052] S606: termination determination is performed, specifically, for low-voltage ride-through, if , it indicates that the iteration is completed, and the final control parameter vector, is output, which is the final parameter; otherwise, the iteration continues according to S605; similarly, for high-voltage ride-through, if , it indicates that the iteration is completed, and the final control parameter vector, is output, which is the final parameter; otherwise, the iteration continues according to S605. Wherein, is the termination iteration allowable deviation.

[0053] To illustrate the effect of the inverter single-machine mechanical modeling parameter identification method provided in the present application on modeling parameter identification, the present application provides the results after the execution of the above exemplary steps.

[0054] Figure 3 is the least squares low-voltage ride-through parameter identification result diagram provided in the present application, wherein Figure 3 (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 3 (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 4 is the optimized low-voltage ride-through parameter identification result diagram provided in the present application, wherein Figure 4 (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 4 (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 5 is the least squares high-voltage ride-through parameter identification result diagram provided in the present application, wherein Figure 5 (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 5 (b) is a voltage drop degree-active power scatter plot during fault ride-through; Figure 6 is the optimized high-voltage ride-through parameter identification result diagram provided in the present application, wherein Figure 6 (a) is a voltage drop degree-reactive power scatter plot during fault ride-through, Figure 6 (b) is a voltage drop degree-active power scatter plot during fault ride-through.

[0055] In combination with Figure 3 and Figure 4 illustration. As Figure 3As shown, the time-per-unit scatter plot display of the inverter's reactive / active power identification results under low-voltage ride-through conditions using the traditional least squares method shows a large fitting deviation in some time periods, indicating significant overfitting or fluctuations; for example... Figure 4 As shown, the optimized method provided in this application is used for low voltage ride-through parameter identification, in conjunction with... Figure 3 Under the same working conditions and coordinate system, the optimized identification scatter points are closer to the actual measured data than... Figure 3 The deviation was significantly reduced, indicating that the recognition accuracy and stability were significantly improved after the introduction of elastic network optimization.

[0056] Combination Figure 5 and Figure 6 Explanation. For example... Figure 5 As shown, the traditional least squares method was used to identify high-voltage ride-through parameters. The scatter plots of the results show that, under high-voltage ride-through conditions, the reactive / active power identification results have a significant fitting error compared to the measured values, especially during voltage surges, where the model predictions deviate considerably from the measured values. Figure 6 As shown, the optimized method provided in this application is used for high voltage ride-through parameter identification, in conjunction with... Figure 5 Under the same high-voltage ride-through conditions, the optimized identification results are in high agreement with the measured data, the error is significantly reduced, and the fitting process is smoother, indicating that the method provided in this application can still maintain good robustness when processing ill-conditioned high-voltage ride-through data.

[0057] This application also provides a device for identifying electromechanical modeling parameters of a single inverter unit. The device includes: a data acquisition module for acquiring fault ride-through measured data or semi-physical test data of a photovoltaic inverter; a data processing module for processing the measured data according to a preset voltage threshold to generate a ride-through data matrix; a control strategy module for constructing a corresponding control strategy based on the ride-through data matrix to generate a vector of control parameters to be identified; a parameter fitting module for constructing a least-squares objective function and fitting predicted values ​​with actual output values ​​to generate an initial control parameter vector; and a parameter optimization module for introducing an elastic network regularization term into the least-squares objective function and iteratively optimizing it using a proximal gradient descent method to output the final control parameter vector. Further, the parameter optimization module includes a soft thresholding unit for processing L1 regularization and a spherical projection unit for processing L2 regularization.

[0058] In summary, by introducing the least square method of elastic network optimization, combined with the proximal gradient iteration, the high-precision identification of the inverter single-machine fault ride-through control parameter is realized, the method can effectively overcome the overfitting and sensitivity to ill-conditioned data of the traditional least square method, ensure the accuracy and stability of the parameter result, and the corresponding device can realize the rapid landing and execution of the method, improve the process, convenience and robustness of the whole station modeling and grid verification of the photovoltaic power station.

[0059] The above steps provide an introduction, which is only used to help understand the method, structure and core idea of the application. It should be noted that the embodiments described in the application are only used to illustrate the principles and effects of the method, and do not constitute a limitation on the protection scope of the application. Those skilled in the art can make various equivalent replacements and improvements to the execution order of the steps, the combination mode of the steps or the implementation means without departing from the essence of the method, and these equivalent schemes shall fall within the protection scope of the application.

Claims

1. A method for identifying machine electrical modeling parameters of an inverter unit, characterized in that, The method comprises the following steps: acquiring measured data or semi-physical test data of fault ride-through of a target model inverter; processing the measured data or semi-physical test data according to a preset voltage threshold to generate a ride-through data matrix; establishing a corresponding control strategy according to the ride-through data matrix to generate a to-be-identified control parameter vector; constructing a least square method objective function to fit a predicted value under the control strategy with an actual output value to acquire an initial control parameter vector; introducing an elastic network regularization term into the least square method objective function for optimization; solving and iteratively optimizing the least square method objective function by using a proximal gradient descent method to output a final control parameter vector.

2. The method of claim 1, wherein the measured data or semi-physical test data comprises low-voltage ride-through data and high-voltage ride-through data; the preset voltage threshold comprises a low-voltage ride-through threshold and a high-voltage ride-through threshold; the ride-through data matrix comprises a low-ride data matrix and a high-ride data matrix.

3. The method of claim 2, wherein, The low-voltage ride-through threshold is preset to 0.9 p.u., and the high-voltage ride-through threshold is preset to 1.1 p.u.

4. The method of claim 2, wherein the low-voltage ride-through data comprises three-phase symmetrical faults and two-phase asymmetrical faults, and the positive sequence voltage during the faults is 0 p.u., 0.2 p.u., 0.35 p.u., 0.5 p.u., and 0.75 p.u., respectively; the high-voltage ride-through data comprises three-phase symmetrical faults and two-phase asymmetrical faults, and the positive sequence voltage during the faults is 1.2 p.u., 1.25 p.u., and 1.30 p.u., respectively.

5. The method of any one of claims 1-4, wherein, The measured data or semi-physical test data is stored in a six-column matrix form, and the six columns are respectively time labels, positive sequence voltages, active powers, reactive powers, active currents, and reactive currents.

6. The method of any one of claims 1-4, wherein, The elastic network regularization term comprises L1 regularization and L2 regularization, and the weight ratio of the L1 regularization and the L2 regularization is dynamically adjusted according to the voltage ride-through working condition.

7. The method of claim 6, wherein, The proximal gradient descent method comprises the following steps: initializing the control parameter vector, calculating a fixed term gradient, adopting a soft threshold function for the L1 regularization part, performing spherical projection on the L2 regularization part, updating the parameter vector, and performing termination determination.

8. The method of any one of claims 1-4, wherein, The data acquisition frequency of the measured data or semi-physical test data is not less than 1000 Hz, and all the data are in per unit format.

9. An apparatus for identifying parameters of an electromechanical model of an inverter unit, characterized by The method comprises the following steps: a data acquisition module for acquiring fault ride-through measured data or semi-physical test data of a photovoltaic inverter; a data processing module for processing the measured data according to a preset voltage threshold to generate a ride-through data matrix; a control strategy module for establishing a corresponding control strategy according to the ride-through data matrix to generate a to-be-identified control parameter vector; a parameter fitting module for constructing a least square method objective function and fitting a predicted value with an actual output value to generate an initial control parameter vector; a parameter optimization module for introducing an elastic network regularization term into the least square method objective function and iteratively optimizing the least square method objective function by using a proximal gradient descent method to output a final control parameter vector.

10. The apparatus of claim 9, wherein, The parameter optimization module includes a soft threshold operation unit for processing L1 regularization and a spherical projection unit for processing L2 regularization. The parameter optimization module includes a soft threshold operation unit for processing L1 regularization and a spherical projection unit for processing L2 regularization.

Images