DFIG wind power plant dynamic equivalence method, system and equipment based on protection action state recognition

Through the improved light gradient hoist model and misjudgment penalty factor matrix identification Crowbar action state, combined with the two-stage grouping method and capacity weighting method, the problem of network off-network unit identification under severe failure of DFIG wind farm is solved, and the equivalent model is improved Accuracy and universality.

CN118917193AActive Publication Date: 2024-11-08WUHAN UNIV
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202410993971.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-08
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

The existing DFIG wind farm equivalent method fails to effectively consider the situation of some fans disconnecting the network under serious failures, resulting in low equivalence accuracy and the inaccuracy of the disconnection unit being unable to accurately identify the disconnection unit.

Method used

The data-driven classification model is used to identify the Crowbar action state. By constructing an improved light gradient hoist model, combining the misjudgment penalty factor matrix, the recognition accuracy of the Crowbar action state is improved, and the two-stage grouping method and capacity weighting method are proposed. , obtain a wind farm equivalent model with no more than 4 machines.

Benefits of technology

It realizes accurate identification of off-grid units in wind farms, improves the universality and accuracy of the equivalent model, is suitable for more practical scenarios, and reduces the misjudgment rate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917193A_ABST
    Figure CN118917193A_ABST
Patent Text Reader

Abstract

The invention provides a DFIG wind power plant dynamic equivalence method, system and device based on protection action state recognition. By setting different fault degrees and wind speeds, fan protection action conditions in different operation scenes are simulated, and data supplied to the LightGBM classifier for learning are obtained; a fan operation state criterion is provided, and fan protection action states under different faults are rapidly judged; constructing an improved LightGBM fan protection action recognition model, and proposing a misjudgment penalty factor matrix to improve the classification precision; when the method is applied, the state quantity of the fan is input into the improved LightGBM classification model, so that a grouping result can be quickly and accurately obtained, and a wind power plant equivalent model is obtained through parameter calculation; and finally, the accuracy and the universality of the method provided by the invention are verified based on Simulink. According to the method, the partial off-network operation condition of the wind power plant under the fault is considered, identification and classification are carried out based on data driving, the speed is high, the accuracy is high, and the highest universality is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of doubly-fed wind farm equivalence, and in particular to a DFIG wind farm dynamic equivalence method, system and equipment based on protection action state identification. Background Art

[0002] The installed capacity of wind power in my country is increasing year by year. The impact of large-scale wind turbines on the stable operation of the power system is becoming increasingly important. It is urgent to establish a model that can accurately describe the dynamic characteristics of wind farm operation. Since wind farms are often composed of dozens or even hundreds of wind turbines, it is of great significance to establish a reliable and accurate wind farm equivalent model in order to reduce the complexity and calculation amount of the simulation model and improve the simulation accuracy at the same time.

[0003] The current equivalence methods for doubly fed wind turbine (DFIG) wind farms mainly include single-machine equivalence method and multi-machine equivalence method. The single-machine equivalence method is to equate the entire wind farm to one unit, which reduces the amount of calculation but has low accuracy; the multi-machine equivalence method studies the differences in the operating state quantities of different wind turbines during steady state and fault periods, such as active power, wind speed, pitch angle, etc., and groups units with similar states into a group, thereby obtaining a multi-machine equivalence model, which has higher accuracy than single-machine equivalence, but the selection of state quantities is crucial.

[0004] Although domestic and foreign scholars have conducted a lot of research on the equivalence of DFIG wind farms, the existing equivalence methods have not considered the situation where some wind turbines in the entire wind farm are disconnected from the grid under severe faults. The partial disconnection of wind turbines will cause the external power of the wind farm to change. Therefore, it is very important to identify and equate the off-grid units. The action characteristics of the crowbar are often the first grouping indicator to distinguish the differences between units during faults, but the existing crowbar identification method has low accuracy and cannot guarantee the equivalence accuracy. Summary of the invention

[0005] The example of the present invention provides a DFIG wind farm dynamic equivalent method, system and device based on protection action state identification, which identifies the action state of Crowbar based on a data-driven classification model, and can ensure a high recognition accuracy rate; considering the scenario where some wind turbines in the wind farm are off-grid, the off-grid units are identified, so that the equivalent model can adapt to more actual scenarios and has the highest universality.

[0006] To achieve the above object, the technical solution of the method of the present invention is:

[0007] A DFIG wind farm dynamic equivalent method based on protection action state identification, the method comprising the following steps:

[0008] Data acquisition steps: Build a wind farm system with multiple DFIGs. Each wind turbine in the wind farm system is equipped with low voltage ride-through LVRT control, Crowbar protection, and low voltage grid-tripping protection. Set different fault levels and wind speeds to simulate the protection actions inside the wind turbines under different operating scenarios. Collect the power, voltage, and wind speed of each wind turbine to construct a sample data set. After normalizing and processing the sample data set, divide it into a training set and a test set.

[0009] Operation status judgment step: setting the fan operation status judgment criteria and marking the fan operation status label on the fan unit based on the fan operation status judgment criteria;

[0010] Model training step: using the data in the training set constructed in the data acquisition step as model input, using the fan operation status label in the operation status judgment step as output, constructing a fan protection action recognition classification model based on the improved light gradient boosting machine, and improving the model based on the misjudgment penalty factor matrix to obtain an improved fan protection status classification model, training the improved fan protection status classification model, and using the data in the test set obtained in the data acquisition step to evaluate the classification effect;

[0011] Wind farm equivalent steps: obtain the power, voltage, and wind speed of each wind turbine in the actual wind farm and input them into the improved wind turbine protection status classification model, and group the wind turbines with the same action status into one group; propose a two-stage grouping method to further divide the Crowbar non-action unit into two units according to the wind speed range; calculate the power and capacity of the equivalent unit based on the capacity weighted method, calculate the line impedance based on the equal voltage loss method, and calculate the equivalent wind speed based on the relationship curve between power and wind speed, and obtain the equivalent model of a wind farm with no more than 4 units;

[0012] Simulation verification steps: Build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification.

[0013] Furthermore, the data acquisition steps are specifically as follows:

[0014] By setting different short-circuit resistance values ​​and simulating different fault degrees, the voltage at the wind farm grid-connected point drops to 0.1pu to 0.5pu. By setting different wind speeds, the protection action inside the wind turbine under different operating scenarios is simulated. Five characteristic quantities, namely steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed, are collected for each wind turbine to construct a sample data set, which is then divided into training set and test set after normalization and processing.

[0015] Furthermore, the operation status determination step is specifically as follows:

[0016] Fan operating status criteria:

[0017]

[0018] In the formula, i r represents the rotor current, U min Represents the lowest voltage, NAND represents the unit where Crowbar does not operate; YAND represents the unit where Crowbar operates but is not disconnected from the grid; YAYD represents the unit where Crowbar operates and is disconnected from the grid;

[0019] The Crowbar units are divided into two groups according to their Crowbar action and inaction. The Crowbar inaction units are labeled NAND as 0. The Crowbar action units include the offline units and the non-offline units. The non-offline units are labeled YAND as 1, and the offline units are labeled YAYD as 2.

[0020] Furthermore, the model training steps are specifically as follows:

[0021] Based on the LightGBM machine learning framework, the histogram algorithm and the Leaf-wise decision tree growth strategy with depth limit are introduced to build a wind turbine protection action recognition and classification model based on the improved light gradient boosting machine;

[0022] The misjudgment penalty factor matrix is ​​introduced to improve the model and obtain the improved wind turbine protection status classification model;

[0023] The five characteristic vectors of steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed are used as the input of the improved wind turbine protection status classification model, and the labels of the wind turbine units according to the wind turbine operation status criterion are used as the output corresponding to each group of inputs;

[0024] Train the improved wind turbine protection status classification model and use the test set to evaluate the classification effect;

[0025] Furthermore, the misjudgment penalty factor matrix is ​​as follows:

[0026]

[0027] Among them, m 01 Represents the MPF when the actual value is 0 but predicted to be 1, m 10 Represents the MPF when the actual value is 1 but predicted to be 0, m 02 Represents the MPF when the actual value is 0 but predicted to be 2, m 20 Represents the MPF when the actual value is 2 but the prediction is 0, m 12 Represents the MPF when the actual value is 1 but predicted to be 2, m 21 Represents the MPF when the actual value is 2 but the prediction is 1,

[0028] Set m 01 , m 02, m 10 , m 12 is 1.5; set m 20 The range is 2.5-3, m 21 The range is 2-2.5.

[0029] The optimal parameters are selected through grid search to minimize the loss function and the lowest misjudgment rate.

[0030] Furthermore, the misjudgment penalty factor matrix is ​​introduced to improve the model, and the improved wind turbine protection status classification model is obtained, including:

[0031] The loss function of LightGBM is:

[0032]

[0033] In the formula, N represents the number of decision trees, F m-1 (x i ; A m-1 ) is the first m-1 trees for x i The predicted value of A m-1 is the parameter, y i is the actual value, F m-1 (x i ; A m-1 ) is the predicted value, L(y i , F m-1 (x i ; A m-1 )) is the error function between the true value and the predicted value, and the logarithmic function is selected here;

[0034] The loss function expression after adding MPFM is as follows:

[0035]

[0036] Where, L modified Represents the improved loss function.

[0037] Furthermore, the wind farm equivalent steps are specifically as follows:

[0038] Obtain the pre-fault power, pre-fault steady-state voltage, minimum voltage during the fault, steady-state voltage during the fault, and wind speed data of each wind turbine in the actual wind farm, and input them into the improved wind turbine protection status classification model;

[0039] The improved wind turbine protection status classification model divides wind turbines with the same operating status into a group, and further divides the Crowbar non-operating units into two units according to the wind speed range;

[0040] Calculate the power and capacity of equivalent units based on the capacity weighting method:

[0041]

[0042] In the formula, S equ , P equ Represents the capacity and active power of the equivalent unit, S i , P i represents the capacity and active power of each unit, and M represents the number of action groups belonging to this type;

[0043] At the same time, the equivalent line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed;

[0044] Based on the above steps, an equivalent model of a wind farm with no more than 4 turbines is obtained.

[0045] Furthermore, the specific steps of the simulation verification based on the MATLAB / Simulink simulation platform are as follows:

[0046] A detailed wind farm is established based on MATLAB / Simulink, and a large amount of training data is obtained through a large number of operating scenario settings to train and improve the LightGBM model; some scenarios are selected to perform dynamic protection of the wind farm and verify the rationality of the equivalent results.

[0047] On the other hand, the present invention provides a DFIG wind farm dynamic equivalent system based on protection action state identification, comprising:

[0048] Data acquisition module: It is used to build a wind farm system with multiple DFIGs. Each wind turbine in the wind farm system is equipped with low voltage ride-through LVRT control, Crowbar protection, and low voltage grid-tripping protection. Different fault levels and wind speeds are set to simulate the protection actions inside the wind turbines under different operating scenarios. The power, voltage, and wind speed of each wind turbine are collected to construct a sample data set. The sample data set is normalized and processed and then divided into a training set and a test set.

[0049] Operation status judgment module: used to set the fan operation status judgment criteria and label the fan operation status of the fan unit based on the fan operation status judgment criteria;

[0050] Model training module: it is used to use the data in the training set constructed in the data acquisition step as model input, and the fan operation status label in the operation status judgment step as output, to build a fan protection action recognition classification model based on the improved light gradient boosting machine, and to improve the model based on the misjudgment penalty factor matrix to obtain an improved fan protection status classification model, train the improved fan protection status classification model, and evaluate the classification effect with the data in the test set obtained in the data acquisition step;

[0051] Wind farm equivalent module: It is used to obtain the power, voltage, and wind speed of each wind turbine in the actual wind farm and input them into the improved wind turbine protection status classification model, and group wind turbines with the same action status into one group; a two-stage grouping method is proposed to further divide the Crowbar non-action unit into two units according to the wind speed range; the power and capacity of the equivalent unit are calculated based on the capacity weighted method, the line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed, so as to obtain an equivalent model of a wind farm with no more than 4 units;

[0052] Simulation verification module: It is used to build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification.

[0053] On the other hand, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned DFIG wind farm dynamic equivalent method based on protection action state identification.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention fully considers more scenarios during the operation of the wind farm, that is, the scenario where some wind turbines in the wind farm are disconnected from the grid after a fault. The proposed method can accurately identify the offline units, which is taken into consideration by all existing methods; at the same time, the proposed equivalent model can also be applied to the scenario where no wind turbines are offline, so it has the highest versatility.

[0056] 2. The present invention is based on data-driven classification function and proposes MPFM matrix to improve classification accuracy and reduce misjudgment rate, thereby improving the recognition accuracy of Crowbar action state and further improving the equivalent effect of wind farm equivalent model. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 This is a dynamic equivalent flow chart of a DFIG wind farm based on improved LightGBM protection action state identification according to an embodiment of the present invention.

[0059] Figure 2 It is a detailed model of a 24-machine DFIG wind farm based on MATLAB / Simulink.

[0060] Figure 3 It is to improve the confusion matrix of the classification effect of LightGBM classifier.

[0061] Figure 4 This is the rendering of the equivalent model when some wind turbines are off-grid.

[0062] Figure 5 This is the rendering of the equivalent model in the scenario where no wind turbine is off-grid.

[0063] Figure 6 The invention discloses a DFIG wind farm dynamic equivalent system based on improved LightGBM protection action state recognition. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Example 1

[0066] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0067] like Figure 1 As shown, a DFIG wind farm dynamic equivalent method based on protection action state identification is provided in an embodiment of the present invention, and the method includes:

[0068] Step S101, data acquisition:

[0069] A wind farm system with multiple DFIGs is constructed, and each wind turbine is equipped with low voltage ride-through (LVRT) control, Crowbar protection, and low voltage grid disconnection protection. Among them, LVRT control refers to the control strategy that switches from fault steady-state control to LVRT control when the wind turbine detects a voltage drop, giving priority to increasing reactive power and appropriately reducing active power. qr Control, the active power P is only affected by i dr control, so i qr ,i dr They are determined by the following formulas:

[0070]

[0071] In the formula, i qr represents the rotor q-axis current, i dr represents the rotor d-axis current, the superscripts bf, df, and af represent the state quantities before, during, and after the fault, respectively, and uds represents the d-axis component of the stator voltage, ω s represents the synchronous angular velocity, L m Represents the magnetizing inductance, L s represents the stator flux, k1 represents the dynamic reactive current proportional coefficient, which is 0.2. Represents the stator voltage before the fault, I N is the rated current of the fan, Represents the active power before the fault, I rmax represents the maximum rotor current, which is taken as 1.2, k2 represents the active power recovery coefficient, which is taken as 0.2, t1 represents the fault clearing moment, and t represents time.

[0072] Crowbar is a resistor connected to the rotor side. When it is detected that the rotor current is greater than the threshold value of 1.5, Crowbar is turned on to short-circuit the rotor side to prevent damage to the power electronic equipment.

[0073] The fan end is equipped with low-voltage cut-off protection according to national standards. The criteria are as follows:

[0074]

[0075] In the formula, u i Represents the phase voltage at the fan end, t is the time, when any phase voltage is lower than 0.2, the fan is immediately cut off.

[0076] By setting different short-circuit resistance values, different fault levels are simulated, so that the voltage at the wind farm grid connection point drops to between 0.1pu and 0.5pu. By setting different wind speeds, including different wind directions and wind speeds set according to the wake effect and random wind speed scenarios, the protection action inside the wind turbine under different operating scenarios is simulated. Five feature quantities of each wind turbine, namely steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed, are collected to construct a sample data set, which is divided into training set and test set after normalization and processing.

[0077] Step S102, running status determination:

[0078] The transient response of the Crowbar action is very different from that of the non-actuated unit, and the Crowbar action depends on the magnitude of the current on the wind turbine rotor side. The Crowbar will act when the rotor side current is greater than 1.5:

[0079]

[0080] In the formula, i r represents the rotor current, P bf is the steady-state power before the fault, U bf is the steady-state voltage before the fault, is the steady-state voltage during the fault, and v is the wind speed.

[0081] At the same time, this embodiment proposes to consider the equivalent value in the scenario of wind turbine off-grid, and the wind turbine off-grid is related to the lowest voltage of the grid connection point during the fault period.

[0082] Therefore, the following criteria for fan operation status are proposed:

[0083]

[0084] In the formula, NAND represents the unit without Crowbar action, and it is labeled as 0; YAND represents the unit with Crowbar action but not offline, and it is labeled as 1; YAYD represents the unit with Crowbar action and offline, and it is labeled as 2, so as to obtain the sample output.

[0085] Step S103, model training:

[0086] LightGBM is a machine learning framework based on Gradient Boosting Decision Tree (GBDT), which can be used to solve classification problems. GBDT builds multiple decision trees N and increases the number of decision trees iteratively until the accuracy improvement is less than the threshold. Based on GBDT, LightGBM enhances the robustness to noise by introducing a histogram algorithm and a Leaf-wise decision tree growth strategy with depth restrictions, while ensuring good evaluation accuracy and training speed.

[0087] The loss function of LightGBM is:

[0088]

[0089] In the formula, N represents the number of decision trees, F m-1 (x i ; A m-1 ) is the first m-1 trees for x i The predicted value of A m-1 is the parameter, y i is the actual value, F m-1 (x i ; A m-1 ) is the predicted value, L(y i , F m-1 (x i ; A m-1 )) is the error function between the true value and the predicted value, and the logarithmic function is selected here.

[0090] It should be noted that in actual scenarios, there are few scenarios where wind turbines are partially off-grid, so the amount of YAYD data will be much smaller than the amount of NAND and YAND data. However, the partial off-grid wind turbine has the most important impact on the equivalent effect of the wind farm. If the off-grid wind turbine is misjudged, the accuracy of the equivalent effect will be significantly reduced. Based on this, the present invention proposes a misjudgment penalty factor matrix (MPFM):

[0091]

[0092] Among them, m 01 Represents the MPF when the actual value is 0 but predicted to be 1, m 10 Represents MPF when the actual value is 1 but predicted to be 0. The other letters have similar meanings. Since NAND and YAND have a large amount of data, their misjudgment has less impact on the equivalent result than YAYD’s misjudgment, so m is set 01 , m 02 , m 10 , m 12 is 1.5; in order to improve the accuracy of the model's judgment of YAYD, m 20 , m 21 Set a higher penalty, especially considering that the impact of judging YAYD as NAND is the most obvious, so set m 20 The range is 2.5-3, m 21 The range of is 2-2.5, and the optimal parameters are selected by grid search to minimize the loss function and the lowest misjudgment rate.

[0093] The loss function expression after adding MPFM is as follows:

[0094]

[0095] The input of the improved LightGBM model is the five features of steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed obtained in step S101. The output corresponding to each set of input is 0, 1, 2 marked in step S102. By introducing MPFM, the classification performance is further improved. The wind turbine protection status classification model is trained and the classification effect is evaluated using the test set;

[0096] Step S104, wind farm equivalents:

[0097] In practical applications, the pre-fault power, pre-fault steady-state voltage, minimum voltage during the fault process, steady-state voltage during the fault, and wind speed data of each wind turbine in the actual wind farm are obtained and input into the improved LightGBM classification model;

[0098] The classification model will group wind turbines with the same operating status into one group. To improve the accuracy of the equivalent model, a two-stage grouping method is proposed to further divide the Crowbar non-operating units into two units according to the wind speed range.

[0099] Calculate the power and capacity of equivalent units based on the capacity weighting method:

[0100]

[0101] In the formula, S equ , P equ Represents the capacity and active power of the equivalent unit, S i , P i Represents the capacity and active power of each unit, and M represents the number of action groups belonging to this type.

[0102] At the same time, the equivalent line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed.

[0103] Based on the above steps, an equivalent model of a wind farm with no more than 4 turbines is obtained.

[0104] Step S105, simulation verification:

[0105] A detailed model of a 24-machine DFIG wind farm was built in MATLAB / Simulink, and based on the proposed improved LightGBM classification model, the feasibility of the protection equivalent method proposed in this paper was verified by simulation.

[0106] The detailed topology of the wind farm is as follows: Figure 2 As shown in the figure, the short-circuit resistance is set from 1Ω to 3Ω in steps of 0.1Ω, 21 groups of faults with different severity are simulated; 10 groups of different wind speed conditions are set to obtain 21*10*24=5040 groups of data, and they are labeled according to the proposed operating status criteria; the model is trained with the training set, and the model training effect is judged with the test set, and the confusion matrix is ​​drawn as shown in the figure. Figure 3 As shown in the figure, the overall classification accuracy of the model reaches 99.58%, which can significantly distinguish fans in different operating states. At the same time, based on the MPFM matrix proposed in the present invention, YAYD achieves a classification accuracy of 100% with the smallest sample size, proving that the improved LightGBM proposed in the present invention can greatly reduce the misjudgment rate and improve the classification accuracy.

[0107] A set of scenarios were selected to verify the accuracy of the equivalent model: a three-phase short circuit fault occurred at the wind farm grid connection point at 10s, and the grid connection point voltage dropped below 0.2. The pre-fault power, pre-fault steady-state voltage, minimum voltage during the fault process, steady-state voltage during the fault, and wind speed data of 24 units were collected and input into the improved LightGBM to obtain the penalty factor m. 20 =3, m21 is 2.45, and the grouping results are: 1, 2, 3, 9, 10, 11, 17, 18 are YAYD, 4, 5, 6, 12, 13, 14, 19, 20, 21 are YAND, 7, 8, 15, 16, 22, 23, 24 are NAND. The classification results are completely consistent with the actual action situation, and a three-machine equivalent model is obtained.

[0108] Figure 4 The comparison between the active power, reactive power, voltage and current of the equivalent model and the actual wind farm at the grid connection point is shown. The closer the curves are, the better the effect of the equivalent model.

[0109] The method proposed in the present invention is compared with the single-machine equivalence method and the popular two-machine equivalence method of whether Crowbar is in action or not. Through the comparison, it can be found that under the operating conditions where some wind turbines in a wind farm are disconnected from the grid, the existing methods cannot accurately identify the offline units, and thus will produce large errors; the method proposed in the present invention is based on the improved classification function of LightGBM and the proposed wind turbine operating status judgment criterion, which accurately classifies the units in different protection action conditions into a group, thereby achieving the highest accuracy.

[0110] In order to verify the universality of the method proposed in the present invention, an operation situation in which no unit is disconnected from the grid is selected for verification: a three-phase short circuit fault occurs at the wind farm grid connection point at 10s, the grid connection point voltage drops to around 0.25, and no wind turbine is disconnected from the wind farm; the pre-fault power, pre-fault steady-state voltage, minimum voltage during the fault process, steady-state voltage during the fault, and wind speed data of 24 units are collected and input into the improved LightGBM model, and the grouping results are obtained as follows: 1, 2, 3, 4, 5, 9, 10, 11, 12, 13, 17, 18 are YAND, 6, 7, 8, 14, 15, 16, 19, 20, 21, 22, 23, 24 are NAND, and the classification results are completely consistent with the actual operation situation; the NAND group is further divided into two groups according to the wind speed range to obtain a three-machine equivalent model, and the comparison is as follows Figure 5 As shown, the method proposed in the present invention still has the highest accuracy.

[0111] Example 2

[0112] like Figure 6 As shown in the figure, the DFIG wind farm dynamic equivalent system based on protection action state identification includes:

[0113] Data acquisition module: It is used to build a wind farm system with multiple DFIGs. Each wind turbine in the wind farm system is equipped with low voltage ride-through LVRT control, Crowbar protection, and low voltage grid-tripping protection. Different fault levels and wind speeds are set to simulate the protection actions inside the wind turbines under different operating scenarios. The power, voltage, and wind speed of each wind turbine are collected to construct a sample data set. The sample data set is normalized and processed and then divided into a training set and a test set.

[0114] Operation status judgment module: used to set the fan operation status judgment criteria and label the fan operation status of the fan unit based on the fan operation status judgment criteria;

[0115] Model training module: it is used to use the data in the training set constructed in the data acquisition step as model input, and the fan operation status label in the operation status judgment step as output, to build a fan protection action recognition classification model based on the improved light gradient boosting machine, and to improve the model based on the misjudgment penalty factor matrix to obtain an improved fan protection status classification model, train the improved fan protection status classification model, and evaluate the classification effect with the data in the test set obtained in the data acquisition step;

[0116] Wind farm equivalent module: It is used to obtain the power, voltage, and wind speed of each wind turbine in the actual wind farm and input them into the improved wind turbine protection status classification model, and group wind turbines with the same action status into one group; a two-stage grouping method is proposed to further divide the Crowbar non-action unit into two units according to the wind speed range; the power and capacity of the equivalent unit are calculated based on the capacity weighted method, the line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed, so as to obtain an equivalent model of a wind farm with no more than 4 units;

[0117] Simulation verification module: It is used to build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification.

[0118] DFIG wind farm data acquisition module 21, wind turbine operation status judgment module 22, model training module 23, DFIG wind farm equivalent module 24, simulation verification module 25.

[0119] The DFIG wind farm data acquisition module 21 is used to acquire wind turbine operation data, which is used to:

[0120] Simulate the protection actions inside the wind turbine under different operating scenarios, collect the power, voltage, and wind speed of each wind turbine to construct a sample data set input;

[0121] The fan operation state judgment module 22 is used to judge the state of the fan in different operation states, and is used to:

[0122] Set the fan operation status criterion and label the fan units based on the fan operation status criterion. Divide them into two groups according to Crowbar action and inaction, and label the Crowbar inaction units NAND as 0; Crowbar action units include offline units and non-offline units, label the non-offline units YAND as 1, and label the offline units YAYD as 2;

[0123] The model training module 23 is used to train the fan operation status recognition model, and is used to:

[0124] Taking the power, voltage and wind speed of the wind turbine as the model input and the different action conditions of the wind turbine as the output, a wind turbine protection action recognition classifier based on the improved LightGBM is constructed, and the proposed MPFM model is improved to improve the classification accuracy.

[0125] The DFIG wind farm equivalent module 24 is used to perform equivalent on the DFIG wind farm, and is used to:

[0126] After obtaining the power, voltage and wind speed of each wind turbine in the actual wind farm, the wind turbines with the same action status are divided into a group based on the improved LightGBM classification model; the equivalent machine parameters are calculated to obtain the equivalent model of the wind farm with no more than 4 machines.

[0127] The simulation verification module 25 is used for simulation verification and is used to:

[0128] Build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification;

[0129] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0130] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, if these modifications and variations of the embodiments of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0131] Other parts not described in detail are all prior art.

Claims

1. A dynamic equivalent method for DFIG wind farm based on protection action state identification, It is characterized in that The method comprises the following steps: Data acquisition steps: Build a wind farm system with multiple DFIGs. Each wind turbine in the wind farm system is equipped with low voltage ride-through LVRT control, Crowbar protection, and low voltage grid-tripping protection. Set different fault levels and wind speeds to simulate the protection actions inside the wind turbines under different operating scenarios. Collect the power, voltage, and wind speed of each wind turbine to construct a sample data set. After normalizing and processing the sample data set, divide it into a training set and a test set. Operation status judgment step: setting the fan operation status judgment criteria and marking the fan operation status label on the fan unit based on the fan operation status judgment criteria; Model training step: using the data in the training set constructed in the data acquisition step as model input, using the fan operation status label in the operation status judgment step as output, constructing a fan protection action recognition classification model based on the improved light gradient boosting machine, and improving the model based on the misjudgment penalty factor matrix to obtain an improved fan protection status classification model, training the improved fan protection status classification model, and using the data in the test set obtained in the data acquisition step to evaluate the classification effect; Wind farm equivalent steps: obtain the power, voltage, and wind speed of each wind turbine in the actual wind farm and input them into the improved wind turbine protection status classification model, and group the wind turbines with the same action status into one group; propose a two-stage grouping method to further divide the Crowbar non-action unit into two units according to the wind speed range; calculate the power and capacity of the equivalent unit based on the capacity weighted method, calculate the line impedance based on the equal voltage loss method, and calculate the equivalent wind speed based on the relationship curve between power and wind speed, and obtain the equivalent model of a wind farm with no more than 4 units; Simulation verification steps: Build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification.

2. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 1, It is characterized in that The data acquisition steps are as follows: By setting different short-circuit resistance values ​​and simulating different fault degrees, the voltage at the wind farm grid-connected point drops to 0.1pu to 0.5pu. By setting different wind speeds, the protection action inside the wind turbine under different operating scenarios is simulated. Five characteristic quantities, namely steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed, are collected for each wind turbine to construct a sample data set, which is then divided into training set and test set after normalization and processing.

3. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 1, It is characterized in that The operation status determination steps are as follows: Fan operating status criteria: In the formula, i r represents the rotor current, U min Represents the lowest voltage, NAND represents the unit where Crowbar does not operate; YAND represents the unit where Crowbar operates but is not disconnected from the grid; YAYD represents the unit where Crowbar operates and is disconnected from the grid; The Crowbar units are divided into two groups according to their Crowbar action and inaction. The Crowbar inaction units are labeled NAND as 0. The Crowbar action units include the offline units and the non-offline units. The non-offline units are labeled YAND as 1, and the offline units are labeled YAYD as 2.

4. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 2, It is characterized in that The model training steps are as follows: Based on the LightGBM machine learning framework, the histogram algorithm and the Leaf-wise decision tree growth strategy with depth limit are introduced to build a wind turbine protection action recognition and classification model based on the improved light gradient boosting machine; The misjudgment penalty factor matrix is ​​introduced to improve the model and obtain the improved wind turbine protection status classification model; The five characteristic vectors of steady-state power, steady-state voltage, fault lowest point voltage, fault steady-state voltage and wind speed are used as the input of the improved wind turbine protection status classification model, and the labels of wind turbine units according to the wind turbine operation status criterion are used as the output corresponding to each group of inputs; The improved wind turbine protection status classification model is trained and the classification effect is evaluated using the test set.

5. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 4, It is characterized in that The misjudgment penalty factor matrix is ​​as follows: Among them, m 01 Represents the MPF when the actual value is 0 but predicted to be 1, m 10 Represents the MPF when the actual value is 1 but predicted to be 0, m 02 Represents the MPF when the actual value is 0 but predicted to be 2, m 20 Represents the MPF when the actual value is 2 but the prediction is 0, m 12 Represents the MPF when the actual value is 1 but predicted to be 2, m 21 Represents the MPF when the actual value is 2 but the prediction is 1, Set m 01 , m 02 , m 10 , m 12 is 1.5; set m 20 The range is 2.5-3, m 21 The range is 2-2.5; The optimal parameters are selected through grid search to minimize the loss function and the lowest misjudgment rate.

6. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 5, It is characterized in that The misjudgment penalty factor matrix is ​​introduced to improve the model, and the improved wind turbine protection status classification model is obtained, including: The loss function of LightGBM is: In the formula, N represents the number of decision trees, F m-1 (x i ; A m-1 ) is the first m-1 trees for x i The predicted value of A m-1 is the parameter, y i is the actual value, F m-1 (x i ; A m-1 ) is the predicted value, L(y i , F m-1 (x i ; A m-1 )) is the error function between the true value and the predicted value, and the logarithmic function is selected here; The loss function expression after adding MPFM is as follows: Where, L modified Represents the improved loss function.

7. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 1, It is characterized in that The wind farm equivalent steps are specifically as follows: The pre-fault power, pre-fault steady-state voltage, minimum voltage during the fault, steady-state voltage during the fault, and wind speed data of each wind turbine in the actual wind farm are obtained and input into the improved wind turbine protection status classification model; The improved wind turbine protection status classification model divides wind turbines with the same operating status into a group, and further divides the Crowbar non-operating units into two units according to the wind speed range; Calculate the power and capacity of equivalent units based on the capacity weighting method: In the formula, S equ , P equ Represents the capacity and active power of the equivalent unit, S i , P i represents the capacity and active power of each unit, and M represents the number of action groups belonging to this type; At the same time, the equivalent line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed; Based on the above steps, an equivalent model of a wind farm with no more than 4 turbines is obtained.

8. The DFIG wind farm dynamic equivalent method based on protection action state identification according to claim 1, It is characterized in that The specific steps of the simulation verification based on the MATLAB / Simulink simulation platform are as follows: A detailed wind farm is established based on MATLAB / Simulink, and a large amount of training data is obtained through a large number of operating scenario settings to train and improve the LightGBM model; some scenarios are selected to perform dynamic protection of the wind farm and verify the rationality of the equivalent results.

9. A DFIG wind farm dynamic equivalent system based on protection action state recognition, It is characterized in that include: Data acquisition module: It is used to build a wind farm system with multiple DFIGs. Each wind turbine in the wind farm system is equipped with low voltage ride-through LVRT control, Crowbar protection, and low voltage grid-tripping protection. Different fault levels and wind speeds are set to simulate the protection actions inside the wind turbines under different operating scenarios. The power, voltage, and wind speed of each wind turbine are collected to construct a sample data set. The sample data set is normalized and processed and then divided into a training set and a test set. Operation status judgment module: used to set the fan operation status judgment criteria and label the fan operation status of the fan unit based on the fan operation status judgment criteria; Model training module: it is used to use the data in the training set constructed in the data acquisition step as model input, and the fan operation status label in the operation status judgment step as output, to build a fan protection action recognition classification model based on the improved light gradient boosting machine, and to improve the model based on the misjudgment penalty factor matrix to obtain an improved fan protection status classification model, train the improved fan protection status classification model, and evaluate the classification effect with the data in the test set obtained in the data acquisition step; Wind farm equivalent module: It is used to obtain the power, voltage, and wind speed of each wind turbine in the actual wind farm and input them into the improved wind turbine protection status classification model, and group wind turbines with the same action status into one group; a two-stage grouping method is proposed to further divide the Crowbar non-action unit into two units according to the wind speed range; the power and capacity of the equivalent unit are calculated based on the capacity weighted method, the line impedance is calculated based on the equal voltage loss method, and the equivalent wind speed is calculated based on the relationship curve between power and wind speed, so as to obtain an equivalent model of a wind farm with no more than 4 units; Simulation verification module: It is used to build a detailed model of a 24-machine DFIG wind farm in MATLAB / Simulink for simulation verification; The DFIG wind farm dynamic equivalent system based on protection action state identification is used to execute the steps in the DFIG wind farm dynamic equivalent method based on protection action state identification described in any one of claims 1-8.

10. A computer-readable storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, the steps of the DFIG wind farm dynamic equivalent method based on protection action state identification as described in any one of claims 1 to 8 are implemented.

Citation Information

Patent Citations

  • Doubly-fed wind farm group aggregating method and doubly-fed wind farm group aggregating system based on dynamic characteristics of rotor current

    CN106451527A

  • Wind power plant aggregation equivalent modeling method and system considering fault ride-through characteristic and medium

    CN117955163A

  • New energy power system equivalent inertia partition evaluation method and system, and storage medium

    CN118074165A

  • System and method for controlling a dual-fed induction generator in response to high-voltage grid events

    US20140062424A1

  • Power transformer winding fault positioning method based on deep convolutional neural network integrated with visual identification

    US20210048487A1