A wind power grid-connected low-voltage ride-through control method

Abstract

The application discloses a wind power grid connection low-voltage ride-through control method, and a specific flowchart of the method is multi-dimensional feature acquisition, probability-based pre-judgment of drop level, presetting of Crowbar resistance grade, zero-delay hard triggering, verification and dynamic fine-tuning. The closed-loop control logic of the multi-dimensional feature acquisition-probability-based pre-judgment of drop level-presetting of Crowbar resistance grade-zero-delay hard triggering-dynamic fine-tuning can pre-judge voltage drop, and can realize pre-regulation of resistance grade according to drop conditions, breaks the limitation of traditional post-regulation, avoids the impact of overcurrent peak on a converter, and simultaneously adapts to different drop levels through the grade, so as to meet the requirement of a user.

Description

Technical Field

[0001] This invention relates to the field of wind power grid connection control technology, and in particular to a low voltage ride-through control method for wind power grid connection. Background Technology

[0002] In wind power grid-connected systems, grid voltage can drop momentarily due to faults. To address this issue, existing technologies generally employ low-voltage ride-through control strategies. One widely used protection method in engineering is the Crowbar circuit, which involves activating a set of fixed resistors when the rotor current or DC bus voltage exceeds a threshold. This provides a bypass path for the overcurrent, converting excess energy into heat to dissipate it, thereby protecting the converter from damage. Once the grid voltage recovers, the resistors are disconnected, allowing the wind turbine to return to normal operation.

[0003] However, the magnitude of actual grid voltage drops is uncertain. Existing Crowbar circuits use a single fixed-value resistor, which is usually designed for the most severe fault conditions. This makes it difficult to adjust flexibly according to actual conditions. More importantly, the adjustment logic of existing Crowbars is all reactive. It must wait until overcurrent or overvoltage has occurred before detecting, judging, and then triggering the switch. But in real faults, the rotor current rises very quickly. From the detection of the problem to the actual switch conduction, it can take hundreds of microseconds or even milliseconds. However, due to the error during this period, the first wave of overcurrent spike has already passed. By the time the resistor switches in, the impact it should have caused has already occurred. This lag reduces the protection effect and has limitations.

[0004] The resistance value is usually set to a small value to ensure the energy dissipation speed. However, when a slight voltage drop occurs, using a small resistance will cause the doubly-fed motor to enter a deep asynchronous operation state, absorbing a large amount of reactive power from the grid, which will have a negative impact on voltage recovery. Conversely, if a large resistance value is designed for the mild voltage drop condition, the current peak cannot be effectively limited due to insufficient energy dissipation during a severe voltage drop, resulting in overcurrent damage to the converter. This has limitations. Summary of the Invention

[0005] The purpose of this invention is to provide a low-voltage ride-through control method for wind power grid connection to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a wind power grid-connected low-voltage ride-through control method, comprising the following steps:

[0007] S1. Multidimensional feature acquisition: Acquire operational and external feature data before the occurrence of grid voltage dips;

[0008] S2. Based on probability prediction of voltage drop level, the collected feature data is input into the pre-built voltage drop prediction model to calculate the probability value of future voltage drop. When the probability value exceeds the preset start threshold, the impending fault is classified into multiple prediction levels, including at least mild voltage drop and severe voltage drop, according to the voltage drop depth prediction value output by the model.

[0009] S3. Preset Crowbar resistor level: Before the actual voltage drop waveform occurs, send a pre-adjustment command to the Crowbar circuit based on the predicted level.

[0010] S4. Zero-delay hard triggering: Real-time monitoring of the instantaneous rate of change or amplitude of grid voltage or rotor current. When a voltage drop threshold is detected, the pre-adjustment command is stopped, and the resistor controllable switch is directly triggered, so that the preset resistance level is immediately put into protection.

[0011] S5. Verify dynamics and make fine adjustments. During the Crowbar deployment period, periodically collect the actual drop depth and dynamically adjust the drop level according to the specific situation.

[0012] Preferably, the operational and external characteristic data collected in step S1 includes at least one of the following three categories:

[0013] Power grid operation data includes instantaneous three-phase voltage values ​​at the grid connection point, voltage fluctuation variance, subsynchronous oscillation components, frequency change rate, regional load factor, and estimated short-circuit ratio.

[0014] Meteorological and environmental data: including lightning warning level, wind speed fluctuation rate, temperature and humidity, and dust accumulation level;

[0015] Historical and dispatch data: including historical voltage drop statistics of wind farms, power grid maintenance plans, and the switching status of adjacent lines.

[0016] Preferably, in step S2, the specific method for calculating the probability of a future voltage drop is as follows: the multi-dimensional feature data collected in step one is used to form a feature vector, which is then input into a pre-trained lightweight recurrent neural network. This network is a long short-term memory network or a gated recurrent unit network. The model outputs a probability value and a predicted drop depth value. When the probability value is greater than or equal to a preset probability threshold, it is determined that a drop is about to occur.

[0017] Preferably, in step S2, the predicted level includes the following four levels:

[0018] Level 1, normal state, indicates that no voltage drop is predicted in the future, and the pre-regulation command is not triggered at this time;

[0019] The second level, mild voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 70% and 90% of the rated voltage.

[0020] Level 3, moderate voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 40% and 70% of the rated voltage.

[0021] Level 4, severe voltage drop warning, indicates that the lowest voltage value predicted after a future voltage drop is below 40% of the rated voltage.

[0022] Preferably, the Crowbar circuit in step S3 includes at least two resistor units with different resistance values, and each resistor unit is connected in series with an independently controllable semiconductor switch.

[0023] The pre-adjustment command is as follows: before the voltage drop actually occurs, control the semiconductor switch corresponding to the target resistance level to pre-conduct, so that the resistance circuit is in a hot standby state, but the Crowbar main bypass switch remains open, so as not to affect normal grid-connected operation.

[0024] Preferably, in step S3, the Crowbar circuit includes three resistor units with the following resistance values: the high resistance value is greater than or equal to twice the medium resistance value, the medium resistance value is greater than or equal to twice the low resistance value, and the resistor units are connected in parallel.

[0025] The pre-adjustment instructions in step three correspond to the following relationships:

[0026] Prediction level one, i.e., mild drop warning, corresponds to the pre-conduction of high resistance value resistor unit;

[0027] Prediction level two, namely moderate drop warning, corresponds to the pre-conduction resistance value resistor unit;

[0028] Level 3 prediction, namely severe drop warning, corresponds to the pre-conduction of low resistance resistor units.

[0029] Preferably, in step S4, the specific judgment condition for detecting the voltage drop threshold is as follows:

[0030] The instantaneous effective value or positive sequence component amplitude of the three-phase voltage is calculated in real time, and a voltage drop is determined to have occurred when any of the following conditions are met:

[0031] Condition 1: The instantaneous value of any phase voltage drops by more than 0.1 times the rated voltage within one millisecond;

[0032] Condition 2: The positive sequence voltage amplitude is lower than 0.9 times the rated voltage at two consecutive points;

[0033] Condition 3: The instantaneous value of the rotor current exceeds 1.2 times the rated value and the rate of change exceeds the preset slope threshold.

[0034] Preferably, in step S4, when a voltage drop threshold is detected, the controller immediately performs the following operations:

[0035] First, cancel the drive signal that maintains the pre-conduction state in S3;

[0036] Secondly, a conduction pulse is sent to the Crowbar main bypass switch;

[0037] Third, a pulse blocking signal is simultaneously sent to the rotor-side converter to stop the converter from modulating.

[0038] Preferably, the specific process for periodically collecting the actual drop depth and dynamically correcting it in step S5 is as follows:

[0039] After the Crowbar is put into operation, the actual grid voltage amplitude is continuously collected at a cycle of no more than 50 microseconds.

[0040] Calculate the deviation between the actual drop depth and the estimated value in S2. This deviation is equal to the absolute value of the difference between the actual drop depth and the predicted drop depth.

[0041] If the deviation is less than or equal to 0.1 times the rated voltage, the prediction is considered accurate, and the current resistance setting should be kept unchanged.

[0042] If the deviation is greater than 0.1 times the rated voltage and the actual drop is more severe than expected, then a forced switch to a lower resistance level will be initiated.

[0043] If the deviation is greater than 0.1 times the rated voltage and the actual drop is less severe than predicted, then the current gear level will remain unchanged.

[0044] The technical effects and advantages of this invention are as follows:

[0045] 1. This invention utilizes a closed-loop control logic of multi-dimensional feature acquisition, probability prediction of voltage drop levels, preset Crowbar resistance levels, zero-delay hard triggering, and dynamic fine-tuning. This logic enables early prediction of voltage drops and pre-adjustment of resistance levels in response to voltage drops, overcoming the limitations of traditional post-event adjustments and preventing the impact of overcurrent spikes on the converter. Furthermore, by adapting resistance levels to different voltage drop levels, this invention meets the needs of users.

[0046] 2. This invention integrates three types of data—power grid operation, meteorological environment, and historical dispatch—through multi-dimensional feature acquisition, and constructs a voltage drop prediction model using a lightweight recurrent neural network. This model can not only predict the probability and depth of voltage drops, but also distinguish between mild, moderate, and severe voltage drops in advance, thereby improving the targeting and accuracy of low-voltage ride-through control and reducing the risk of protection failure caused by prediction errors. Attached Figure Description

[0047] Figure 1 This is the main flowchart of the overall method of the present invention;

[0048] Figure 2This is a schematic diagram of multi-dimensional feature acquisition in step S1 of the present invention;

[0049] Figure 3 This is a schematic diagram of the process for predicting the drop level in step S2 of the present invention;

[0050] Figure 4 This is a diagram showing the preset resistance settings in step S3 of the present invention.

[0051] Figure 5 This is a schematic diagram of the Crowbar circuit structure of the present invention;

[0052] Figure 6 This is a flowchart of the zero-delay hard triggering process in step S4 of the present invention;

[0053] Figure 7 This is a flowchart of step S5, dynamic fine-tuning, of the present invention.

[0054] Figure 8 This is the complete timing flowchart of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] This invention provides, for example Figure 1 The wind power grid-connected low voltage ride-through control method shown includes the following steps:

[0057] S1. Multidimensional feature acquisition: Acquire operational and external feature data before the occurrence of grid voltage dips;

[0058] S2. Based on probability prediction of voltage drop level, the collected feature data is input into the pre-built voltage drop prediction model to calculate the probability value of future voltage drop. When the probability value exceeds the preset start threshold, the impending fault is classified into multiple prediction levels, including at least mild voltage drop and severe voltage drop, according to the voltage drop depth prediction value output by the model.

[0059] S3. Preset Crowbar resistor level: Before the actual voltage drop waveform occurs, send a pre-adjustment command to the Crowbar circuit based on the predicted level.

[0060] S4. Zero-delay hard triggering: Real-time monitoring of the instantaneous rate of change or amplitude of grid voltage or rotor current. When a voltage drop threshold is detected, the pre-adjustment command is stopped, and the resistor controllable switch is directly triggered, so that the preset resistance level is immediately put into protection.

[0061] S5. Verify dynamics and make fine adjustments. During the Crowbar deployment period, periodically collect the actual drop depth and dynamically adjust the drop level according to the specific situation.

[0062] Example 1: In this example, relevant characteristic data such as power grid operation, meteorological environment, and historical dispatch are first collected in advance. Then, the collected characteristic data is input into a pre-trained prediction model to predict the probability and depth of future voltage drops, and classify them into mild, moderate, and severe drop levels. Next, according to the predicted drop level, the corresponding Crowbar resistor is pre-conducted to put it in a hot standby state without affecting the normal grid-connected operation of the wind turbine. Then, the grid voltage and rotor current are monitored in real time. When the voltage drop threshold is detected, the preset resistor is immediately triggered for protection and the rotor-side converter is blocked. During the resistor protection period, the actual drop depth is periodically collected and compared with the predicted value. The resistor level is dynamically adjusted according to the deviation between the two to avoid secondary impact. Finally, when the grid voltage returns to normal, the Crowbar resistor is disconnected, the rotor-side converter is unblocked, and the wind turbine resumes normal grid-connected operation.

[0063] This invention provides, for example Figures 2 to 8 The wind power grid-connected low voltage ride-through control method shown includes the following steps:

[0064] S1. Multidimensional feature acquisition: Acquire operational and external feature data before the occurrence of grid voltage dips;

[0065] S2. Based on probability prediction of voltage drop level, the collected feature data is input into the pre-built voltage drop prediction model to calculate the probability value of future voltage drop. When the probability value exceeds the preset trigger threshold, the impending fault is classified into multiple prediction levels, including at least mild voltage drop and severe voltage drop, according to the voltage drop depth prediction value output by the model. The model is trained offline using historical wind farm operation data and is updated online incrementally based on the deviation between the actual voltage drop event and the prediction result during operation.

[0066] S3. Preset Crowbar resistor level: Before the actual voltage drop waveform occurs, send a pre-adjustment command to the Crowbar circuit based on the predicted level.

[0067] S4. Zero-delay hard triggering: Real-time monitoring of the instantaneous rate of change or amplitude of grid voltage or rotor current. When a voltage drop threshold is detected, the pre-adjustment command is stopped, and the resistor controllable switch is directly triggered, so that the preset resistance level is immediately put into protection.

[0068] S5. Verify dynamics and make fine adjustments. During the Crowbar deployment period, periodically collect the actual drop depth and dynamically adjust the drop level according to the specific situation.

[0069] The operational and external characteristic data collected in step S1 include at least one of the following three categories:

[0070] Power grid operation data includes instantaneous three-phase voltage values ​​at the grid connection point, voltage fluctuation variance, subsynchronous oscillation components, frequency change rate, regional load factor, and estimated short-circuit ratio.

[0071] Meteorological and environmental data: including lightning warning level, wind speed fluctuation rate, temperature and humidity, and dust accumulation level;

[0072] Historical and dispatch data includes historical voltage sag statistics for wind farms, grid maintenance plans, and the switching status of adjacent lines. The specific process for periodically collecting actual voltage sag depth and dynamically correcting it in step S5 is as follows:

[0073] After the Crowbar is put into operation, the actual grid voltage amplitude is continuously collected at a cycle of no more than 50 microseconds.

[0074] Calculate the deviation between the actual drop depth and the estimated value in S2. This deviation is equal to the absolute value of the difference between the actual drop depth and the predicted drop depth.

[0075] If the deviation is less than or equal to 0.1 times the rated voltage, the prediction is considered accurate, and the current resistance setting should be kept unchanged.

[0076] If the deviation is greater than 0.1 times the rated voltage and the actual drop is more severe than expected, then a forced switch to a lower resistance level will be initiated.

[0077] If the deviation is greater than 0.1 times the rated voltage and the actual drop is less severe than expected, the current gear level will be maintained to avoid secondary impact during the switching process.

[0078] Example 2: In this example, the power grid operation data is as follows: the instantaneous value of the three-phase voltage at the grid connection point is collected by the grid connection point voltage sensor, frequency sensor, and load monitoring device. The sampling frequency is not less than 10kHz, the calculation period of the voltage fluctuation variance is 100ms, the subsynchronous oscillation component is extracted by Fourier transform into the oscillation component in the range of 10-50Hz, the frequency change rate, the regional load factor, specifically the ratio of the current regional actual load to the rated load, and the short-circuit ratio estimate, which is estimated by the power grid impedance measurement device.

[0079] Meteorological and environmental data: Lightning warning level, wind speed fluctuation rate (the ratio of the difference between the maximum and minimum wind speed within 1 minute to the average wind speed), temperature and humidity, and dust accumulation level are collected through the wind farm meteorological monitoring station. These are obtained through the photovoltaic panel or wind turbine blade dust accumulation monitoring device and are divided into four levels: no dust accumulation, light dust accumulation, moderate dust accumulation, and heavy dust accumulation.

[0080] Historical and dispatch data: Retrieve voltage drop statistics for the past 3 years from the wind farm monitoring system, including drop time, drop depth, duration, and maintenance plans issued by the power grid dispatch center, including maintenance periods, maintenance lines, and the switching status of adjacent lines.

[0081] After the Crowbar resistance setting is activated, the actual grid voltage amplitude is continuously collected at a period of 50 microseconds. The deviation between the actual drop depth and the predicted drop depth in step S2 is calculated, i.e., deviation = actual drop depth - predicted drop depth. Dynamic corrections are made based on the magnitude of the deviation. The specific process is as follows:

[0082] 1. If the deviation is ≤0.1 times the rated voltage, it indicates that the prediction is accurate. Keep the current resistance setting unchanged and continue to monitor the changes in the mains voltage.

[0083] 2. If the deviation is greater than 0.1 times the rated voltage, and the actual drop depth is more severe than the predicted drop depth, that is, the actual minimum voltage value is lower than the predicted minimum voltage value, then the resistor will be forcibly switched to a lower level to ensure rapid energy discharge and limit the peak value of the rotor current.

[0084] 3. If the deviation is greater than 0.1 times the rated voltage, and the actual drop depth is slightly less than the predicted drop depth, that is, the actual minimum voltage value is higher than the predicted minimum voltage value, then the current gear position shall be maintained to avoid secondary current surges during gear switching, which would affect the stable operation of the wind turbine and the power grid.

[0085] When the grid voltage recovers to more than 90% of the rated voltage and lasts for more than 100ms, the controller issues a command to disconnect the Crowbar resistor setting, release the pulse blockade of the rotor-side converter, and restore the wind turbine to normal grid-connected operation.

[0086] This application solves the problems of delayed post-adjustment and poor adaptability of existing Crowbar circuits by predicting in advance, setting preset gears, triggering with zero delay, and making dynamic fine adjustments. While protecting the converter from damage, it also takes into account the needs of grid voltage recovery, thereby improving the low voltage ride-through capability and operational stability of the wind power grid-connected system.

[0087] This invention provides, for example Figure 1 The wind power grid-connected low voltage ride-through control method shown includes the following steps:

[0088] S1. Multidimensional feature acquisition: Acquire operational and external feature data before the occurrence of grid voltage dips;

[0089] S2. Based on probability prediction of voltage drop level, the collected feature data is input into the pre-built voltage drop prediction model to calculate the probability value of future voltage drop. When the probability value exceeds the preset start threshold, the impending fault is classified into multiple prediction levels, including at least mild voltage drop and severe voltage drop, according to the voltage drop depth prediction value output by the model.

[0090] S3. Preset Crowbar resistor level: Before the actual voltage drop waveform occurs, send a pre-adjustment command to the Crowbar circuit based on the predicted level.

[0091] S4. Zero-delay hard triggering: Real-time monitoring of the instantaneous rate of change or amplitude of grid voltage or rotor current. When a voltage drop threshold is detected, the pre-adjustment command is stopped, and the resistor controllable switch is directly triggered, so that the preset resistance level is immediately put into protection.

[0092] S5. Verify dynamics and make fine adjustments. During the Crowbar deployment period, periodically collect the actual drop depth and dynamically adjust the drop level according to the specific situation.

[0093] In step S2, the specific method for calculating the probability of a future voltage drop is as follows: the multi-dimensional feature data collected in step one is used to form a feature vector, which is then input into a pre-trained lightweight recurrent neural network. This network is a long short-term memory network or a gated recurrent unit network. The model outputs a probability value and a predicted drop depth value. When the probability value is greater than or equal to a preset probability threshold, it is determined that a drop is about to occur.

[0094] In step S2, the prediction level includes the following four levels:

[0095] Level 1, normal state, indicates that no voltage drop is predicted in the future, and the pre-regulation command is not triggered at this time;

[0096] The second level, mild voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 70% and 90% of the rated voltage.

[0097] Level 3, moderate voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 40% and 70% of the rated voltage.

[0098] Level 4, severe voltage drop warning, indicates that the lowest voltage value predicted after a future voltage drop is below 40% of the rated voltage.

[0099] The Crowbar circuit in step S3 includes at least two resistor units with different resistance values, and each resistor unit is connected in series with an independently controllable semiconductor switch.

[0100] The pre-regulation command is as follows: before the voltage drop actually occurs, control the semiconductor switch corresponding to the target resistance level to pre-conduct, so that the resistance circuit is in a hot standby state, but the Crowbar main bypass switch remains open, so as not to affect normal grid-connected operation.

[0101] In step S3, the Crowbar circuit includes three resistor units with the following resistance values: the high resistance value is greater than or equal to twice the medium resistance value, the medium resistance value is greater than or equal to twice the low resistance value, and the resistor units are connected in parallel.

[0102] The pre-adjustment instructions in step three correspond to the following relationships:

[0103] Prediction level one, i.e., mild drop warning, corresponds to the pre-conduction of high resistance value resistor unit;

[0104] Prediction level two, namely moderate drop warning, corresponds to the pre-conduction resistance value resistor unit;

[0105] Level 3 prediction, namely severe drop warning, corresponds to the pre-conduction of low resistance resistor units.

[0106] In step S4, the specific judgment condition for detecting the voltage drop threshold is as follows:

[0107] The instantaneous effective value or positive sequence component amplitude of the three-phase voltage is calculated in real time, and a voltage drop is determined to have occurred when any of the following conditions are met:

[0108] Condition 1: The instantaneous value of any phase voltage drops by more than 0.1 times the rated voltage within one millisecond;

[0109] Condition 2: The positive sequence voltage amplitude is lower than 0.9 times the rated voltage at two consecutive points;

[0110] Condition 3: The instantaneous value of the rotor current exceeds 1.2 times the rated value and the rate of change exceeds the preset slope threshold;

[0111] The judgment period for the above conditions shall not exceed one hundred microseconds to ensure that the triggering is completed within the first sampling window at the voltage drop initiation point.

[0112] In step S4, when a voltage drop threshold is detected, the controller immediately performs the following operations:

[0113] First, cancel the drive signal that maintains the pre-conduction state in S3;

[0114] Secondly, within the same control cycle, which does not exceed fifty microseconds, a conduction pulse is sent to the Crowbar main bypass switch;

[0115] Third, a pulse blocking signal is simultaneously sent to the rotor-side converter to stop the converter from modulating;

[0116] The time interval between the second and third steps is less than ten microseconds, and the grid-side converter continues to operate normally, continuing to provide reactive power support to the grid.

[0117] Example 3: In this example, the multi-dimensional feature data collected in step S1 is normalized to form a feature vector, which is then input into a pre-built voltage drop prediction model. The voltage drop prediction model is a lightweight long short-term memory network, which is trained using historical voltage drop case data from wind farms. During training, cross-validation is used to optimize the model parameters. When the model is running, it outputs the probability value and the estimated value of the voltage drop depth within the next 500ms. The preset probability threshold is 80%. When the probability value is ≥80%, it is determined that a voltage drop is about to occur, and the prediction level is divided according to the estimated value of the voltage drop depth. The specific level division is as follows:

[0118] Level 1, or normal state: No voltage drop is predicted in the future. The probability value is <80%, and the pre-regulation command is not triggered in this state.

[0119] The second level, namely the mild voltage drop warning, is when the lowest voltage value after the predicted voltage drop is between 70% and 90% of the rated voltage.

[0120] Level 3, also known as moderate voltage drop warning: The lowest voltage value after the predicted voltage drop is between 40% and 70% of the rated voltage.

[0121] Level 4, or severe voltage drop warning: The lowest voltage value predicted after a future voltage drop is less than 40% of the rated voltage.

[0122] The drop prediction model can also use a gated recurrent unit network. Its training method and prediction logic are the same as those of the lightweight long short-term memory network, with only slight differences in network structure. The appropriate network type can be selected according to the actual hardware computing power.

[0123] In this embodiment, the Crowbar circuit includes three resistor units in parallel topology. Each resistor unit is connected in series with an independent and controllable IGBT semiconductor switch. The three resistor units specifically include a high-resistance resistor (R1), a medium-resistance resistor (R2), and a low-resistance resistor (R3). Their resistance values ​​are related as follows: R1 ≥ 2 × R2, R2 ≥ 2 × R3. The specific values ​​can be adapted according to the rated power of the fan.

[0124] Before the actual voltage drop waveform occurs, a pre-adjustment command is issued to the Crowbar circuit based on the predicted level determined in step S2. The specific correspondence is as follows:

[0125] Mild drop warning: The IGBT switch of the pre-conducting high-resistance resistor unit is turned on, so that the R1 circuit is in hot standby state;

[0126] Moderate drop warning: The IGBT switch corresponding to the pre-conduction medium resistance unit makes the R2 circuit in hot standby state;

[0127] Severe drop warning: The IGBT switch of the pre-conducting low-resistance resistor unit is turned on, so that the R3 circuit is in hot standby state.

[0128] It should be noted that the pre-adjustment command only controls the IGBT switch of the corresponding resistor unit to be pre-conducted, while the Crowbar main bypass switch remains open, which will not affect the normal grid-connected operation of the wind turbine. At the same time, the hot standby state can ensure that the resistor circuit can be instantly turned on when triggered later, shortening the response time.

[0129] The Crowbar circuit can also be configured with two or four resistance units. The resistance value relationships and pre-adjustment command correspondences can be adjusted according to actual needs, as long as they can adapt to different voltage drop levels. The controller monitors the grid voltage and rotor current in real time, with a monitoring cycle not exceeding 100 microseconds. The specific criteria for determining the voltage drop threshold are any of the following:

[0130] Condition 1: The instantaneous value of any phase voltage drops by more than 0.1 times the rated voltage within one millisecond;

[0131] Condition 2: The positive sequence voltage amplitude at two consecutive sampling points is lower than 0.9 times the rated voltage, and the sampling interval is 100 microseconds;

[0132] Condition 3: The instantaneous value of the rotor current exceeds 1.2 times the rated value and the rate of change exceeds the preset slope threshold.

[0133] When any of the above conditions is detected, it is determined that a voltage drop has occurred, and the controller immediately performs the following operations to ensure zero-delay triggering:

[0134] First, cancel the drive signal that maintains the pre-conduction state of the resistor unit in step S3;

[0135] Secondly, within the same control cycle, a conduction pulse is sent to the Crowbar main bypass switch to immediately activate the preset resistance level for protection.

[0136] Third, a pulse blocking signal is simultaneously sent to the rotor-side converter to stop the converter from modulating, thus preventing the converter from being damaged due to overcurrent.

[0137] The time interval between the second and third steps is less than 10 microseconds, and the grid-side converter continues to operate normally, providing reactive power support to the grid and helping the grid voltage to recover.

[0138] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for low-voltage ride-through control of wind power grid connection, characterized in that, Includes the following steps: S1. Multidimensional feature acquisition: Acquire operational and external feature data before the occurrence of grid voltage dips; S2. Based on probability prediction of voltage drop level, the collected feature data is input into the pre-built voltage drop prediction model to calculate the probability value of future voltage drop. When the probability value exceeds the preset start threshold, the impending fault is classified into multiple prediction levels, including at least mild voltage drop and severe voltage drop, according to the voltage drop depth prediction value output by the model. S3. Preset Crowbar resistor level: Before the actual voltage drop waveform occurs, send a pre-adjustment command to the Crowbar circuit based on the predicted level. S4. Zero-delay hard triggering: Real-time monitoring of the instantaneous rate of change or amplitude of grid voltage or rotor current. When a voltage drop threshold is detected, the pre-adjustment command is stopped, and the resistor controllable switch is directly triggered, so that the preset resistance level is immediately put into protection. S5. Verify dynamics and make fine adjustments. During the Crowbar deployment period, periodically collect the actual drop depth and dynamically adjust the drop level according to the specific situation.

2. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, The operational and external characteristic data collected in step S1 include at least one of the following three categories: Power grid operation data includes instantaneous three-phase voltage values ​​at the grid connection point, voltage fluctuation variance, subsynchronous oscillation components, frequency change rate, regional load factor, and estimated short-circuit ratio. Meteorological and environmental data: including lightning warning level, wind speed fluctuation rate, temperature and humidity, and dust accumulation level; Historical and dispatch data: including historical voltage drop statistics of wind farms, power grid maintenance plans, and the switching status of adjacent lines.

3. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, In step S2, the specific method for calculating the probability of a future voltage drop is as follows: the multi-dimensional feature data collected in step one is used to form a feature vector, which is then input into a pre-trained lightweight recurrent neural network. This network is a long short-term memory network or a gated recurrent unit network. The model outputs a probability value and a predicted drop depth value. When the probability value is greater than or equal to a preset probability threshold, it is determined that a drop is about to occur.

4. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, In step S2, the prediction level includes the following four levels: Level 1, normal state, indicates that no voltage drop is predicted in the future, and the pre-regulation command is not triggered at this time; The second level, mild voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 70% and 90% of the rated voltage. Level 3, moderate voltage drop warning, indicates that the lowest voltage value after the predicted voltage drop is between 40% and 70% of the rated voltage. Level 4, severe voltage drop warning, indicates that the lowest voltage value predicted after a future voltage drop is below 40% of the rated voltage.

5. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, The Crowbar circuit in step S3 includes at least two resistor units with different resistance values, and each resistor unit is connected in series with an independently controllable semiconductor switch. The pre-adjustment command is as follows: before the voltage drop actually occurs, control the semiconductor switch corresponding to the target resistance level to pre-conduct, so that the resistance circuit is in a hot standby state, but the Crowbar main bypass switch remains open, so as not to affect normal grid-connected operation.

6. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, In step S3, the Crowbar circuit includes three resistor units with the following resistance values: the high resistance value is greater than or equal to twice the medium resistance value, the medium resistance value is greater than or equal to twice the low resistance value, and the resistor units are connected in parallel. The pre-adjustment instructions in step three correspond to the following relationships: Prediction level one, i.e., mild drop warning, corresponds to the pre-conduction of high resistance value resistor unit; Prediction level two, namely moderate drop warning, corresponds to the pre-conduction resistance value resistor unit; Level 3 prediction, namely severe drop warning, corresponds to the pre-conduction of low resistance resistor units.

7. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, In step S4, the specific judgment condition for detecting the voltage drop threshold is as follows: The instantaneous effective value or positive sequence component amplitude of the three-phase voltage is calculated in real time, and a voltage drop is determined to have occurred when any of the following conditions are met: Condition 1: The instantaneous value of any phase voltage drops by more than 0.1 times the rated voltage within one millisecond; Condition 2: The positive sequence voltage amplitude is lower than 0.9 times the rated voltage at two consecutive points; Condition 3: The instantaneous value of the rotor current exceeds 1.2 times the rated value and the rate of change exceeds the preset slope threshold.

8. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, In step S4, when a voltage drop threshold is detected, the controller immediately performs the following operations: First, cancel the drive signal that maintains the pre-conduction state in S3; Secondly, a conduction pulse is sent to the Crowbar main bypass switch; Third, a pulse blocking signal is simultaneously sent to the rotor-side converter to stop the converter from modulating.

9. The wind power grid-connected low-voltage ride-through control method according to claim 1, characterized in that, The specific process for periodically collecting and dynamically correcting the actual drop depth in step S5 is as follows: After the Crowbar is put into operation, the actual grid voltage amplitude is continuously collected at a cycle of no more than 50 microseconds. Calculate the deviation between the actual drop depth and the estimated value in S2. This deviation is equal to the absolute value of the difference between the actual drop depth and the predicted drop depth. If the deviation is less than or equal to 0.1 times the rated voltage, the prediction is considered accurate, and the current resistance setting should be kept unchanged. If the deviation is greater than 0.1 times the rated voltage and the actual drop is more severe than expected, then a forced switch to a lower resistance level will be initiated. If the deviation is greater than 0.1 times the rated voltage and the actual drop is less severe than predicted, then the current gear level will remain unchanged.

Images