Offshore wind power plant active time sequence optimization control method and system

Abstract

The invention provides an active time sequence optimization control method and system for an offshore wind power plant, and relates to the technical field of wind power plant control. By judging whether the turbulence intensity, the growth rate, the wind speed change and the power fluctuation are too large at the same time, the intense turbulence surge state is accurately identified. If the wind power station is in the state, wind speed ultra-short-term prediction is carried out based on an ARIMA model, an active power track is predicted through a wind power station power transfer function, and whether the power change rate or absolute value over-limit risk exists or not is pre-judged. Once the risk exists, the unit is divided into a high-risk unit and an adjustable unit according to the degree of sensitivity to turbulence, and the adjustment potential of the adjustable unit is evaluated. And finally, aiming at smoothing the power trajectory and avoiding over-limit, solving an optimal adjustment trajectory by adopting a genetic algorithm, thereby realizing power adjustment of the wind power plant unit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind farm control, in particular to a kind of offshore wind farm active time sequence optimization control method and system. BACKGROUND

[0002] As an important part of clean energy, offshore wind power is developing towards large scale and clusterization. However, the offshore environment is complex and changeable, and the wind conditions at the outlet of the wind farm, especially large-scale offshore wind farm, often show significant turbulent characteristics. This turbulence is caused by sea-air interaction, complex terrain and wind turbine wake superposition effect, resulting in rapid fluctuations in wind speed and direction in a very short time. For grid-connected wind farms, this strong turbulence condition can cause severe oscillation and superposition of power of each unit in the field, resulting in rapid and large fluctuations in total active power of the wind farm. The existing wind farm power control strategy is mostly based on average wind speed for steady-state optimization, or only for independent adjustment of single machine, lacking of collaborative awareness and forward inhibition ability to the time sequence fluctuation characteristics of the total power of the field. This leads to the wind farm easily triggering the power change rate limit and absolute upper limit specified by the power grid under strong turbulence conditions, not only facing examination fines, but also posing a serious threat to the frequency stability and power quality of the power grid.

[0003] Currently, wind farm-level power control relies on tracking of automatic generation control instructions, which is a passive response control mode. The problem is that the response lags behind the fluctuation and cannot cope with the rapid increase of turbulence in seconds or even sub-seconds. Especially in offshore scenarios, the interaction of wake effect and turbulence makes the power fluctuation problem particularly prominent, and there is an urgent need for a method that can assess the turbulence state in real time, predict the risk of power overrun, and implement collaborative optimization control based on the differential adjustment potential of units in the field, to realize the time sequence smoothing of wind farm active power output and ensure the safe and stable operation of the power grid.

[0004] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The present application aims to provide a kind of offshore wind farm active time sequence optimization control method and system to solve the problems raised in the above background.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions: A kind of offshore wind farm active time sequence optimization control method, the specific steps include: Step 1: Based on the wind speed and wind direction data at the outlet of the offshore wind farm in the current sampling period, and the wind speed and output power of each wind turbine nacelle, the total active power of the wind farm is calculated in real time, and the average turbulence intensity of the wind farm is evaluated; and the average turbulence intensity, wind speed variation and total active power fluctuation of the wind farm in a plurality of consecutive sampling periods are analyzed to determine whether the wind farm is in a strong turbulence surge state; Step 2: If the wind farm is in a strong turbulence surge state, based on the ultra-short-term prediction of the wind speed and the preset wind farm power transfer function, the predicted trajectory of the total active power of the wind farm in the next sampling period is generated; and it is determined whether there is an over-limit risk in the predicted trajectory; Step 3: If there is an over-limit risk, according to the sensitivity of the wind turbine to turbulence response, all grid-connected wind turbines in the wind farm are divided into high-risk wind turbines and adjustable wind turbines, and the adjustment potential of the adjustable wind turbines under the current wind speed is obtained; Step 4: Based on the genetic algorithm, the optimal adjustment trajectory is calculated to smooth the predicted trajectory of the total active power of the wind farm, and the optimal control instruction is generated to avoid the over-limit risk; the optimal control instruction includes preventive power limiting instruction applied to the high-risk wind turbines and compensatory adjustment instruction applied to the adjustable wind turbines; Step 5: The optimal control instruction is sent to the corresponding wind turbine controller, and the above process is repeated in the next sampling period until it is determined that the wind farm is not in a strong turbulence surge state.

[0007] Further, the outlet of the offshore wind farm is the outlet position of the power collection line and the offshore booster station, and determining whether the wind farm is in the strong turbulence surge state specifically includes: A wind measurement tower is installed at the outlet of the offshore wind farm to collect wind speed and wind direction data at the outlet of the offshore wind farm at a first preset sampling frequency, the first preset sampling frequency is not less than 1Hz; through the wind farm monitoring system, real-time data messages uploaded by all wind turbine controllers in the wind farm are received at a second preset sampling frequency; the real-time data messages at least include the nacelle wind speed measured by the ultrasonic anemometer installed at the top of the wind turbine nacelle, and the output power measured by the converter sensor of the wind turbine, and the second preset sampling frequency is not less than 0.5Hz; The average value of the output power of each grid-connected wind turbine in the wind farm at all sampling times in the current sampling period is calculated, and the sum of the average values of the output power of all wind turbines is taken as the total active power of the wind farm; The average turbulence intensity of the wind farm is evaluated as follows: According to the wind speed data, the average turbulence intensity of the current sampling period is calculated, and the formula is as follows: ; wherein, is the average turbulence intensity of the current sampling period, is the standard deviation of the wind speed of all sampling time points in the current sampling period, is the average value of the wind speed of all sampling time points in the current sampling period, k represents the current sampling period; When the following four conditions are met simultaneously, it is determined that the wind farm is in a strong turbulence surge state, specifically: a. The average turbulence intensity of the current sampling period is greater than the first preset threshold value; b. The growth rate of the average turbulence intensity of the current sampling period relative to the average value of the previous M sampling periods is greater than the second preset threshold value; c. The absolute value of the average wind speed change rate monitored in the current sampling period is greater than the third preset threshold value; d. The standard deviation of the total active power of the wind farm in the previous M sampling periods is greater than the fourth preset threshold value.

[0008] Further, the calculation formula of the growth rate is as follows: ; Wherein, represents the growth rate, is the average turbulence intensity of the previous M sampling periods; The calculation formula of the average wind speed change rate is as follows: ; Wherein, is the average wind speed change rate, represents the average value of the wind speed of all sampling time points in the previous sampling period, is the time length of the sampling period.

[0009] Further, determining whether the over-limit risk exists specifically includes: The ultra-short-term prediction refers to the prediction of the wind speed and wind direction of each sampling time point in the next sampling period; based on the wind speed and wind direction data monitored at the outlet of the wind farm, the wind speed sequence and the wind direction sequence in the current sampling period and the previous M sampling periods are arranged to form a sample set, an ARIMA prediction model is established using time series analysis method, the wind speed sequence and the wind direction sequence of the previous N sampling periods in the sample set are taken as input, and the wind speed sequence and the wind direction sequence of the next sampling period are taken as output, the model is trained; and the wind speed sequence and the wind direction sequence of the current and the previous N-1 sampling periods are input into the trained model to obtain the predicted wind speed sequence and wind direction sequence of the next sampling period; the wind speed sequence refers to the sequence formed by the wind speed data of all sampling time points in the sampling period, and the wind direction sequence refers to the sequence formed by the wind direction data of all sampling time points in the sampling period; The current grid-connected unit data is acquired from a wind farm monitoring system, and a predicted wind speed sequence is substituted into a preset wind farm power transfer function to calculate the total active power at each sampling time in the next sampling period to form a predicted trajectory, the wind farm power transfer function being as follows: ; wherein, Ptotal(t) is the total active power of the wind farm at the tth sampling time in the next sampling period, is the current air density, determined according to the temperature and air pressure data of the location of the wind farm, is the swept area of the wind wheel, is the wind energy utilization coefficient, is the predicted wind speed at the tth sampling time in the next sampling period, and t is the index of the sampling time in the sampling period, represents the number of grid-connected units in the wind farm, is the average efficiency factor of the wind farm, is the array effect reduction coefficient; The over-limit risk refers to the risk of exceeding the limit value of the power grid, including two judgment conditions. Condition one is to judge whether there is a power change rate over-limit risk, specifically: if the absolute value of the total active power growth rate of adjacent sampling times in the next sampling period is greater than a preset threshold one, it is determined that there is a power change rate over-limit risk; the preset threshold one is the maximum power change rate limit value per minute specified by the power grid; the total active power change rate of adjacent sampling times is obtained by calculating the total active power difference between the next sampling time and the previous sampling time in adjacent sampling times, and calculating the ratio of the difference to the time interval of adjacent sampling times as the total active power change rate of adjacent sampling times; Condition two is to judge whether there is a power absolute value over-limit risk, specifically: if the absolute value of the total active power of a certain sampling time in the next sampling period is greater than a preset threshold two, it is determined that there is a power absolute value over-limit risk; the preset threshold two is the upper limit value of the active power specified by the power grid; If any of the conditions is met, it is determined that there is an over-limit risk.

[0010] Further, the adjustment potential is obtained by: if there is an over-limit risk, starting the adjustment mechanism of the total active power of the wind farm; According to the sensitivity of the unit to the turbulence response, all grid-connected units in the wind farm are divided into high-risk units and adjustable units, and the sensitivity is represented by the standard deviation of the output power change rate of the unit. The specific division process is as follows: calculate the standard deviation of the output power change rate of each unit in a historical time window containing several historical sampling periods, and divide the units with a standard deviation higher than the preset sensitivity threshold into high-risk units, and divide the remaining units into adjustable units; the output power change rate of the unit in any sampling period is obtained as follows: calculate the output power difference between the unit in the sampling period and the previous sampling period, and calculate the ratio of the difference to the adjacent sampling time interval as the output power change rate of the unit in the sampling period; The adjustment potential includes a maximum up-regulation power potential and a maximum down-regulation power capacity; the maximum up-regulation power potential is calculated as follows: ; Wherein, represents the maximum up-regulation power potential of the ith adjustable unit at the tth sampling time in the next sampling period, represents the theoretical maximum available power of the ith adjustable unit at the tth sampling time in the next sampling period, represents the maximum power change rate allowed by the ith adjustable unit, is the time length of the sampling period; the theoretical maximum available power is determined as follows: if the predicted wind speed at the tth sampling time in the next sampling period is not greater than the rated wind speed of the ith adjustable unit, the theoretical maximum available power of the ith adjustable unit at the tth sampling time in the next sampling period is calculated as follows: , if the predicted wind speed is greater than the rated wind speed, the theoretical maximum available power is the rated power of the unit, and i is the index of the adjustable unit; The maximum down-regulation power capacity is calculated as follows: ; Wherein, is the maximum down-regulation power capacity of the ith adjustable unit at the current time, represents the actual output power of the ith adjustable unit at the current time, represents the minimum output power allowed by the ith adjustable unit; the maximum down-regulation power capacity of all sampling times in the next sampling period is the value; If it is judged that there is no over-limit risk, instructions are sent to all units to maintain normal operation.

[0011] Further, the optimal control instruction is obtained as follows: based on the predicted trajectory of the total active power of the wind farm in the next sampling period, the predicted trajectory being composed of the total active power predicted values corresponding to all sampling times in the next sampling period; For each sampling time in the predicted trajectory, the upper limit of the total active power adjustment at this time is determined according to the sum of the maximum up-regulation power potential of all adjustable units, and the lower limit of the total active power adjustment at this time is determined according to the sum of the maximum down-regulation power capacity of all adjustable units, so as to form the power allowable value range at each sampling time; The genetic algorithm is adopted to generate a smooth and non-overlimit risk-adjusted total active power trajectory as the target for optimization and solving, and the specific process is as follows: The initial population is randomly generated, each individual represents a candidate adjusted total active power trajectory, which is composed of the same number of sampling times as the predicted trajectory, and the total active power value at each sampling time is taken as its gene value, which is randomly generated within the corresponding power allowable value range; The smoothness of the adjusted trajectory represented by the individual is taken as its fitness value, and the smoothness is quantified by calculating the reciprocal of the sum of the absolute values of the power change rates of adjacent sampling times in the trajectory; The constraint condition setting step: before calculating the fitness, first judge whether the adjusted trajectory represented by the individual has over-limit risk; if there is over-limit risk, directly set the fitness of the individual to a predetermined penalty value, so that it is eliminated in the subsequent selection; Genetic operation step: selection, crossover and mutation operations are performed on the population to generate a new generation of population; selection is performed based on the fitness value using the tournament selection method; multi-point crossover is performed between the gene positions of the individual with a certain probability; the value of any gene position in the individual is randomly changed with a certain probability, and the value after mutation is within the power allowable value range of the sampling time; Iteration termination step: repeat the fitness value calculation step to the genetic operation step until the maximum number of iterations is reached; Output result step: the adjusted total active power trajectory represented by the individual with the highest fitness in the final population is determined as the optimal adjustment trajectory; According to the difference between the optimal adjustment trajectory and the original predicted trajectory at each sampling time, the adjustment amount of the total active power of the wind farm is calculated; The total adjustment amount is distributed to each adjustable unit according to the adjustment potential of each adjustable unit to generate a compensatory adjustment instruction for each adjustable unit; if the adjustment amount of the total active power of the wind farm is positive, it indicates that the output power of the wind farm is to be increased, and the weight is determined according to the ratio of the maximum up-regulation power potential of each adjustable unit to the adjustment amount of the total active power; if the adjustment amount of the total active power of the wind farm is negative, it indicates that the output power of the wind farm is to be reduced, and the weight is determined according to the ratio of the maximum down-regulation power capacity of each adjustable unit to the adjustment amount of the total active power; at the same time, for high-risk units, a preventive power limiting instruction is generated, and the value of the preventive power limiting instruction is the current output power of the unit minus a power buffer margin preset according to the rated power of the unit.

[0012] Further, the compensatory adjustment instruction is used to instruct each adjustable unit to change the output power according to the power adjustment amount allocated to the adjustable unit; and the preventive power limiting instruction is used to instruct each high-risk unit to limit the output power to a level lower than the current value.

[0013] The application further provides a wind farm active timing optimization control system, which is used to implement the wind farm active timing optimization control method. A real-time monitoring and turbulence identification module is configured to calculate the total active power of the wind farm and evaluate the average turbulence intensity of the wind farm in real time according to the wind speed and wind direction data at the outlet of the wind farm and the wind speed and output power of each wind turbine in a current sampling period, and analyze the average turbulence intensity, wind speed change and total active power fluctuation of the wind farm in a plurality of continuous sampling periods to determine whether the wind farm is in a strong turbulence surge state. A power overrun risk prediction module is configured to generate a predicted trajectory of the total active power of the wind farm in a next sampling period based on the ultra-short-term prediction of the wind speed and a preset power transfer function of the wind farm if the wind farm is in the strong turbulence surge state, and determine whether there is an overrun risk in the predicted trajectory. A unit grouping and potential evaluation module is configured to divide all grid-connected units in the wind farm into high-risk units and adjustable units according to the sensitivity of the units to turbulence response and obtain the adjustment potential of the adjustable units under the current wind speed if there is an overrun risk. A power optimization control module is configured to calculate an optimal adjustment trajectory based on a genetic algorithm with the smooth predicted trajectory of the total active power of the wind farm as a target and the avoidance of the overrun risk as a constraint, and generate optimal control instructions based on the optimal adjustment trajectory; the optimal control instructions include a preventive power limiting instruction applied to the high-risk units and a compensatory adjustment instruction applied to the adjustable units. An instruction execution and feedback module is configured to issue the optimal control instruction to the corresponding wind turbine controller and repeat the above process in the next sampling period until it is determined that the wind farm is not in the strong turbulent burst state.

[0014] In the above technical solution, the present application has the following technical effects and advantages: The present application can accurately identify the strong turbulent burst state caused by complex sea conditions by calculating the average turbulent intensity of the wind farm in real time and comprehensively analyzing multi-period data, and early perception of power fluctuation risk is achieved. Furthermore, in combination with the ultra-short-term wind condition prediction based on time series analysis and the accurate wind farm power transfer function, the trajectory of the total active power can be predicted in advance and the over-limit risk can be judged, and the control mode can be changed from passive response to active prevention. Finally, by dynamically grouping the wind turbines according to their sensitivity to turbulence response and determining the optimal adjustment trajectory based on the genetic algorithm, the preventive power limitation of high-risk wind turbines and the compensatory adjustment of adjustable wind turbines are coordinated, and the collaborative optimization and suppression of power time sequence fluctuations in the entire wind farm are achieved. This method effectively solves the problem of power fluctuation and easy grid limit value violation of offshore wind farms under strong turbulent conditions, significantly smooths the grid-connected power curve, reduces the risk of examination, and enhances the friendliness and support capacity of large-scale offshore wind farms to the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The figure is a schematic diagram of the overall method of the present application. Figure 2 The figure is a schematic diagram of the system structure of the present application. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical scheme and advantages of the present application clearer, further detailed description will be made in combination with specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the usual meaning understood by those skilled in the art to which the present application belongs. The terms "first", "second" and similar words used in the present application do not represent any order, quantity or importance, but are only used to distinguish different components. The terms "include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. The terms "connect" or "connect" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "up", "down", "left", "right" and the like only represent relative positional relationships, which may change accordingly when the absolute position of the described object changes.

[0018] Embodiment: Referring to Figure 1 The present application provides a technical solution: A method for active time sequence optimization control of an offshore wind farm, comprising the following specific steps: Step 1: According to the wind speed and wind direction data at the outlet of the offshore wind farm in the current sampling period, as well as the wind speed and output power of each wind turbine nacelle, the total active power of the wind farm is calculated in real time, and the average turbulence intensity of the wind farm is evaluated; and the average turbulence intensity, wind speed variation and total active power fluctuation of the wind farm in a plurality of continuous sampling periods are analyzed to determine whether the wind farm is in a strong turbulence surge state; In this embodiment, the outlet of the offshore wind farm is the outlet position of the connection between the power collection line of the wind farm and the offshore booster station, and at least one wind measurement tower is installed. The wind measurement tower should be equipped with a high-precision ultrasonic anemometer or a sensor with equivalent or higher precision for measuring three-dimensional wind speed and direction. Determining whether the wind farm is in the strong turbulence surge state specifically includes: A wind measurement tower is installed at the outlet of the offshore wind farm to collect wind speed and wind direction data at the outlet of the offshore wind farm at a first preset sampling frequency, and the first preset sampling frequency is not less than 1 Hz. Preferably, the sampling frequency is 1 Hz to 4 Hz. The reason is that turbulence is a physical phenomenon with high frequency variation, and a too low sampling frequency (e.g. once per minute) will lose key turbulence spectrum information and cannot accurately calculate the turbulence intensity. Sampling above 1 Hz can effectively capture the wind speed fluctuation in seconds, providing a reliable data basis for subsequent analysis. The collected raw data should at least include instantaneous wind speed and instantaneous wind direction.

[0019] Through the central monitoring system (SCADA) of the wind farm, real-time data messages actively uploaded by each wind turbine controller in the field are received in a communication protocol (such as IRC 61400-25). Real-time data messages uploaded by all wind turbine controllers in the wind farm are received at a second preset sampling frequency. The second preset sampling frequency is not less than 0.5 Hz. Preferably, the frequency is 0.5 Hz to 2 Hz, and is synchronized with the data acquisition time stamp of the wind measurement tower. The reason is that the power response of the unit to wind speed change has certain inertia, and a slightly lower wind speed sampling frequency is acceptable, but it still needs to be high enough to track the power fluctuation trend. A frequency lower than 0.5 Hz makes it difficult to effectively evaluate the power change in seconds to minutes. The real-time data message at least contains the nacelle wind speed measured by the ultrasonic anemometer installed on the top of the wind turbine nacelle, and the output power measured by the power sensor of the unit converter, such as the current transformer and voltage transformer; the nacelle wind speed is measured by the ultrasonic anemometer installed on the top of the wind turbine nacelle, reflecting the true inflow wind speed at the hub height of the unit. The output power is measured by the power sensor on the side of the unit converter, such as the current transformer and voltage transformer, reflecting the real-time active power output of the unit.

[0020] The received wind speed and wind direction data at the outlet of the offshore wind farm and the wind speed and output power data of each wind turbine nacelle are preprocessed, including abnormal value identification and processing, missing value filling, and filtering, to ensure that the data quality meets the requirements of subsequent model training and analysis.

[0021] The abnormal value identification and processing of wind speed data specifically includes: Although wind speed data may change rapidly due to turbulence, there is a physical upper limit to its instantaneous acceleration. For wind speed data (including wind speed at the outlet of the offshore wind farm and wind speed at the nacelle of each wind turbine), the effective range of wind speed is set to [0 m / s, 40 m / s]. Any wind speed value less than 0 m / s or greater than 40 m / s is determined to be abnormal. The lower limit of 0 m / s is based on physical common sense, and the upper limit of 40 m / s is based on the reference extreme cut-out wind speed with a margin, which can effectively filter out non-physical values caused by sensor failure or communication errors. For wind speed at the outlet of the offshore wind farm and wind speed at the nacelle of each wind turbine, if the wind speed of a sampling point exceeds the effective range of wind speed, the wind speed data of that sampling point is marked as an abnormal value and is removed.

[0022] The wind speed change rate threshold is set to For wind speed at the outlet of the offshore wind farm and wind speed at the nacelle of each wind turbine, if the wind speed change rate of a sampling point relative to the wind speed of the previous sampling point exceeds this threshold, the wind speed data of that sampling point is marked as an abnormal value and is removed.

[0023] This change rate threshold is based on an engineering estimate of the physical upper limit of wind speed instantaneous change under extreme turbulence conditions, aiming to capture jump points caused by communication instantaneous interruption or sensor instantaneous interference. The wind speed change rate threshold set in this embodiment is based on the engineering consensus of aerodynamic characteristics under extreme turbulence conditions. This value is far beyond the reasonable range of wind speed change in the normal atmospheric boundary layer. Using this wind speed change rate threshold can effectively capture abnormal spikes caused by sensor instantaneous power failure and restart, signal transmission jump, and other non-physical factors. If these abnormal points are not removed, they will seriously distort the calculation results of turbulence intensity.

[0024] Further, for wind speed data at the outlet of the offshore wind farm and wind speed at the nacelle of each wind turbine, which are determined to be abnormal values after the above abnormal value identification and removal, there may be missing data. In addition, due to the limitations of sensor signal transmission, there may also be missing data at some sampling points during the sampling process of wind speed data. For missing data of wind speed at the outlet of the offshore wind farm and wind speed at the nacelle of each wind turbine, the missing value filling step is performed, which specifically includes: If there is only one isolated missing value in the wind speed sequence, i.e. the data at the previous and next time points are valid: the arithmetic mean of the wind speed data of the previous and next valid sampling points is used for filling, this method is simple and efficient, and can better maintain the local trend of the data, and is suitable for single data point missing caused by instantaneous interference under high frequency sampling, if the data at the previous and next two end points in the wind speed data sequence are missing, since there are no two neighbor sampling points, in this case, the data of the neighbor sampling points is directly copied for filling.

[0025] If there are at least two consecutive missing values in the sequence, the consecutive missing values are regarded as a missing value segment, the wind speed data of the sampling point at the previous time point and the wind speed data of the sampling point at the next time point of the missing value segment are selected, the gradient change rate is calculated according to the difference between the wind speed data at the previous and next time points of the missing value segment, and combined with the number of missing values, and the gradient is adjusted based on the wind speed data of the sampling point at the previous time point according to the gradient change rate.

[0026] For example, the wind speed data sequence segment of a certain segment is: …, 12 m / s, V1, V2, V3, 18 m / s, …, wherein V1, V2, V3 are three consecutive missing values, which constitute a missing value segment, the wind speed data of the sampling point at the previous time point of the missing value segment is 12 m / s, and the wind speed data of the sampling point at the next time point is 18 m / s, then the difference between the wind speed data at the previous and next time points of the missing value segment is 6 m / s, the missing value is 3, and there are 4 gradients, then the gradient is 1.5 m / s, then V1 is 13.5 m / s, V2 is 15 m / s, and V3 is 16.5 m / s.

[0027] Such filling of missing values strictly maintains the overall trend determined by the first and last two valid points of the missing data segment. In this example, the overall upward trend from 12 m / s to 18 m / s is completely retained and evenly distributed to each time step (i.e. V1, V2, V3). Such filling sequence is strictly linear, avoiding sharp jumps in the filled values, thereby ensuring that the first derivative (change rate) of the data is constant, which can prevent the introduction of new false fluctuations due to improper filling.

[0028] For abnormal value identification and processing of wind direction data, specifically including: The wind direction data is a physical quantity that circulates in a fixed range, and its normal change pattern is continuous rotation rather than step jump. For wind direction data, abnormal values are determined based on a preset physical threshold, and the effective range of wind direction is set as [0°, 360°]. Values outside this range are determined to be abnormal. Under the premise that the sampling frequency is high enough (not less than 1 Hz), the real wind direction change is gradual between adjacent sampling points, even when a weather system with rapid wind direction switching passes. Therefore, any wind direction value outside the effective range is almost certainly caused by sensor failure or data decoding error, and no additional change rate threshold needs to be set. The invalid data can be reliably identified based on the physical range alone.

[0029] If there is only one isolated missing value in the wind direction data, and the data at the previous and next time points are valid: the arithmetic mean of the wind direction data of the previous and next valid sampling points is used to fill in the missing value.

[0030] If there are at least two consecutive missing values in the sequence, the consecutive missing values are treated as a missing value segment, the wind direction data of the sampling point at the previous time point and the sampling point at the next time point of the missing value segment are selected, the gradient change rate is calculated based on the difference between the wind direction data at the previous and next time points of the missing value segment and the number of missing values, and the gradient is adjusted based on the wind direction data of the sampling point at the previous time point according to the gradient change rate.

[0031] For example, a wind direction data sequence segment of a certain segment is: …, 20°, W1, W2, W3, 60°, …, where W1, W2, and W3 are three consecutive missing values, forming a missing value segment, the wind speed data of the sampling point at the previous time point of the missing value segment is 20°, and the wind speed data of the sampling point at the next time point is 60°. The difference between the wind speed data at the previous and next time points of the missing value segment is 40°, the missing value is 3, and there are 4 gradients, so the gradient is 10 m / s. Therefore, W1 is 30°, W2 is 40°, and W3 is 50°.

[0032] For abnormal value identification and processing of wind turbine output power data, specifically including: As a large rotating machine, the power output of a wind turbine is limited by the response speed of the variable pitch system and the generator torque, and there is a clear maximum power change rate. For output power data, the effective range of power is set to 0.95 times the rated power of the unit to 1.05 times the rated power of the unit. A slight negative value is allowed to consider measurement error or temporary reactive power, and the upper limit is 1.10 times the rated power to consider possible short-term overload or positive sensor drift.

[0033] The absolute value of the power change rate at adjacent sampling time is calculated, and the change rate threshold is set to 1.2 times the maximum power change rate (unit: MW / s) specified in the technical specification of the unit. The 1.2 times coefficient is to increase the safety margin on the basis of the hardware limit to filter the power mutation caused by controller abnormality or data packet loss. The change rate threshold set here respects the physical limit of the equipment and provides a buffer for measurement noise and transient overshoot. The change rate threshold can accurately identify the power value jump caused by controller logic error, data communication packet loss or sensor transient failure.

[0034] Similarly, if there is only one isolated missing value in the output power sequence, that is, the data before and after the time is valid: the arithmetic mean of the output power of the two valid sampling points before and after is used for filling, this method is simple and efficient, and can better maintain the local trend of the data, if the data of the two end points in the output power sequence is missing, since there are no two neighbor sampling points, in this case, the data of the neighbor sampling point is directly copied for filling.

[0035] If there are at least two consecutive missing values in the sequence, the consecutive missing values are regarded as a missing value segment, the output power of the sampling point at the previous time of the missing value segment and the output power of the sampling point at the next time are selected, the gradient change rate is calculated according to the difference between the output power at the previous time and the next time of the missing value segment, and combined with the number of missing values, the gradient is adjusted based on the output power of the sampling point at the previous time.

[0036] For example, assuming that the rated power is 10MW, the output power sequence segment of a segment is: …, 9.6, P1, P2, P3, 10, …, wherein P1, P2, P3 are three consecutive missing values, which constitute a missing value segment, the output power of the sampling point at the previous time of the missing value segment is 9.6MW, and the wind speed data of the sampling point at the next time is 10MW, then the difference between the output power at the previous time and the next time of the missing value segment is 0.4MW, the missing value is 3, and there are 4 gradients, then the gradient is 0.1MW, so P1 is 9.7MW, P2 is 9.8MW, and P3 is 9.9MW.

[0037] The data filtering preprocessing specifically includes: The data sequence after the treatment of outliers and missing values is smoothed by a first-order low-pass filter. The time constant of the filter is set to 0.5 seconds (corresponding to a discrete filter coefficient of about 0.6-0.7). The purpose of filtering is to suppress high-frequency noise and small disturbances that have not been completely removed from the sensor, which are mainly from the background noise of electronic devices and environmental electromagnetic interference. Through filtering, the main frequency signal of wind speed fluctuation used to calculate the turbulence intensity can be extracted more clearly, and the small glitches in the power data that are not mechanism can be smoothed, thereby improving the calculation accuracy of the average turbulence intensity and the total active power, and providing a more reliable data basis for subsequent state judgment.

[0038] Through the above systematic preprocessing process, the invalid and interference information is maximally eliminated while the physical meaning and time sequence characteristics of the data are retained, thereby laying a solid and reliable data foundation for subsequent accurate identification of turbulence surge state, ultra-short-term power prediction, and optimization control.

[0039] For the current sampling period, the output power data of each unit at all sampling times in the period is extracted. First, for a single unit, the average value of its output power at all sampling times in the current sampling period is calculated. This is to smooth out the instantaneous small fluctuations of the unit itself and better reflect the average output level in the period. Then, the average power values of all units in the wind farm that are in operation (excluding units in shutdown or maintenance state) are summed up to obtain the total active power of the wind farm.

[0040] The average turbulence intensity of the wind farm is evaluated as follows: The wind speed sequence collected by the aforementioned wind measurement tower is used as the source data for calculation. The wind measurement tower data is not affected by the wind turbine wake effect and the control strategy of the unit, and can better represent the real natural wind conditions at the entrance of the wind farm. First, the average value of the wind speed at all sampling times in the current sampling period is calculated, and the standard deviation of the wind speed at all sampling times in the same period is calculated. The average turbulence intensity of the current sampling period is calculated as follows: ; wherein, is the average turbulence intensity of the current sampling period, is the standard deviation of the wind speed at all sampling times in the current sampling period, is the average value of the wind speed at all sampling times in the current sampling period, and k represents the current sampling period; usually expressed in percentage. This index quantifies the fluctuation degree of the wind speed around the average level.

[0041] The wind farm is determined to be in a strong turbulence surge state when the following four conditions are met simultaneously. These conditions comprehensively consider the absolute intensity, relative change, wind speed dynamics, and power consequences. The judgment conditions are as follows: a. The average turbulence intensity of the current sampling period is greater than a preset first threshold. The value of the first threshold is determined based on historical wind resource data of the location where the wind farm is located, the category of the wind farm and the requirements of the power grid. Generally, the background value of the turbulence intensity of an offshore wind farm is low. The high percentile (such as the 95% or 98% percentile) of the normal turbulence intensity distribution of the wind farm can be selected as the first threshold. For example, by analyzing the wind measurement data of the site for more than one year, it is found that the annual average value of the turbulence intensity is about 8%, and the 98% percentile value is about 15%. The first threshold can be set to about 0.15 (i.e. 15%). The first threshold is set in this way because the threshold can screen out those working conditions that are in extremely high turbulence intensity, and these working conditions are the potential high-risk period of power fluctuation.

[0042] b. The growth rate of the average turbulence intensity of the current sampling period relative to the average value of the previous M sampling periods is greater than a preset second threshold. The second threshold is an enhancement threshold, and the threshold is based on the standard deviation of the normal fluctuation of the turbulence intensity in the historical data. The standard deviation of the difference value of the turbulence intensity of adjacent sampling periods in the historical data is calculated. The second threshold can be set to twice the standard deviation or three times the standard deviation. For example, the standard deviation of the average turbulence intensity of the historical sampling periods is calculated to be 2%, and the second threshold can be set to 4% or 6% here. This judgment condition is used to capture the mutation of turbulence, that is, even if the absolute turbulence intensity is not the highest, but a rapid and abnormal growth often indicates that the new wind condition will be more unstable and the risk is increasing.

[0043] c. The absolute value of the average wind speed change rate monitored in the current sampling period is greater than a preset third threshold; the value of the third threshold is based on the rated wind speed change rate that the wind turbine can withstand and historical statistical values. A smaller proportion (such as 50%) can be referred to the extreme wind speed change rate value in the design standard of the unit. For example, the instantaneous wind speed change rate allowed by the unit design is 3 m / s², and the third threshold can be set to 0.5-1.0 m / s². A statistical method similar to condition b can also be used to take the high percentile (such as the 95% percentile) of the absolute value of the historical wind speed change rate. The sharp change (sudden rise or sudden drop) of the wind speed is the direct cause of the large fluctuation of the power. This condition is used as a supplement to the turbulence condition to directly capture the input dynamics of the wind energy.

[0044] d. The standard deviation of the total active power of the wind farm over the previous M sampling periods is greater than a fourth threshold value. The fourth threshold value is directly related to the power fluctuation limit specified by the grid. If the grid specifies that the power change per minute should not exceed 10% of the rated power of the wind farm, and the sampling period is 30 seconds (i.e. 2 periods per minute), the fourth threshold value can be set to half of 10% of the rated power of the wind farm or a more conservative value. For example, for a 400 MW wind farm, the limit per minute is 40 MW, and the fluctuation threshold for 30 seconds can be set to 15-20 MW. This condition is the final factual judgment. It indicates that the turbulence and wind speed changes have actually translated into fluctuations in power output, and the fluctuation level has approached the edge of what the grid allows, thus confirming the necessity of intervention.

[0045] The value of the parameter M is also related to the length of each sampling period. For example, if the sampling period is 30 seconds, M can be set to 10; if the sampling period is 10 seconds, M can be set to 60. Therefore, a preferred value range of M is that the corresponding total length of time is between 5 minutes and 30 minutes. If the value of M is too small, such as M = 2, the window is extremely short, and the calculated historical average value is very close to the value of the previous time. This makes the growth rate used for judgment similar to the difference between adjacent periods, which is too sensitive to instantaneous noise and prone to false positives, such as judging normal small fluctuations as a surge. If the value of M is too large, the historical average value becomes very smooth, representing the average background turbulence level over a long period of time. This dilutes recent changes and makes the system insensitive and slow to respond to truly threatening, rapidly developing turbulence surge events, which can miss the best control window. Selecting a window of 5-30 minutes perfectly balances the two extremes. It is long enough to smooth out minor fluctuations on a second and minute scale and establish a stable recent background reference, and short enough to capture significant trend changes on a scale of tens of minutes.

[0046] In summary, only when all four conditions are met is the wind farm determined to be in a strong turbulence surge state. This multi-condition and logic ensures the accuracy and reliability of the judgment and avoids false actions due to abnormal single indicators.

[0047] In this embodiment, the calculation formula of the growth rate is as follows: ; wherein, represents the growth rate, is the average turbulence intensity mean value of the previous M sampling periods; The calculation formula of the average wind speed change rate is as follows: ; wherein, ​is the average wind speed change rate, is the average value of the wind speed at all sampling time points in the last sampling period, is the time length of the sampling period.

[0048] Step 2: If the wind farm is in a strong turbulent burst state, a predicted trajectory of the total active power of the wind farm in the next sampling period is generated based on the ultra-short-term prediction of the wind speed and a preset wind farm power transfer function; it is judged whether there is an over-limit risk in the predicted trajectory; In this embodiment, judging whether there is the over-limit risk specifically includes: The ultra-short-term prediction refers to high-frequency prediction of the wind speed and wind direction at each sampling time point in the next sampling period. In this embodiment, a time series analysis method is preferably used, specifically an autoregressive integrated moving average model. Based on the wind speed and wind direction data monitored at the outlet of the wind farm, the wind speed sequence and wind direction sequence in the current sampling period and the previous M sampling periods are sorted to form a sample set. The value of M determines the length of the historical window, which can refer to the value of M described above, and the preferred range is 10 to 30 periods. The time series analysis method is used to establish an ARIMA prediction model, the wind speed sequence and wind direction sequence of the first N sampling periods in the sample set are used as input, and the wind speed sequence and wind direction sequence of the next sampling period are used as output. The parameters of the ARIMA model (such as the autoregressive order, the difference order, and the moving average order) are trained and determined. The training of the model can be performed online, or a basic model can be trained using historical big data, and then adjusted online. The wind speed sequence and wind direction sequence of the current and previous N-1 sampling periods are input into the trained model to obtain the predicted wind speed sequence and wind direction sequence of the next sampling period; the wind speed sequence refers to a sequence formed by the wind speed data at all sampling time points in the sampling period, and the wind direction sequence refers to a sequence formed by the wind direction data at all sampling time points in the sampling period; N is a natural number, and the value range is [5, 8], ensuring that N is not greater than M.

[0049] The ultra-short-term prediction model is not limited to the ARIMA model, and other machine learning models such as long short-term memory networks and gated recurrent units can also be used, as long as the same high-frequency and short-time wind speed prediction function can be achieved.

[0050] After obtaining the predicted wind speed sequence, it needs to be converted into a predicted trajectory of the total active power of the wind farm. The current on-grid unit data is obtained from the wind farm monitoring system, and the predicted wind speed sequence is substituted into the preset wind farm power transfer function to calculate the total active power at each sampling time point in the next sampling period to form a predicted trajectory, and the wind farm power transfer function is as follows: ; wherein, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period, Ptotal(t) is the total active power of the wind farm at the tth sampling moment in the next sampling period,

[0051] The over-limit risk refers to the risk of exceeding the limit value of the power grid, including two judgment conditions. Condition one is to judge whether there is a power change rate over-limit risk, specifically: if the absolute value of the total active power growth rate of adjacent sampling moments in the next sampling period is greater than a preset threshold one, it is judged that there is a power change rate over-limit risk; the preset threshold one is the maximum power change rate limit value per minute specified by the power grid; the total active power change rate of adjacent sampling moments is obtained by calculating the difference between the total active power of the next sampling moment and the previous sampling moment, and calculating the ratio of the difference to the time interval of adjacent sampling moments as the total active power change rate of adjacent sampling moments; Condition two is to judge whether there is a power absolute value over-limit risk, specifically: if the absolute value of the total active power of a sampling moment in the next sampling period is greater than a preset threshold two, it is judged that there is a power absolute value over-limit risk; the preset threshold two is the upper limit value of the active power specified by the power grid; If any of the conditions is met, it is determined that there is an over-limit risk.

[0052] The step is used for early prediction of over-limit risk, and solves the problem of lagging behind in after-remedy. The wind turbine is not an ideal device for instantaneous response. From receiving an instruction (such as a power reduction instruction) to the action of the pitch system and then to the actual power reduction, there is a delay time of mechanical and electrical response. If the power is controlled after the power has exceeded the limit (in the process) or even after the limit (after the event), it is too late, and the power grid has already been impacted. The step is used for forward-looking judgment, which reserves valuable action time for all subsequent steps. The whole system intervenes one sampling period in advance, ensures that the control instruction takes effect before the risk occurs, and thus realizes truly preventive control.

[0053] Step 3: If there is an over-limit risk, all grid-connected operating units in the wind farm are divided into high-risk units and adjustable units according to the sensitivity of the units to the turbulence response, and the adjustment potential of the adjustable units at the current wind speed is obtained. In the embodiment, the adjustment potential specifically includes: If it is judged in the previous step that there is an over-limit risk, the active adjustment mechanism of the wind farm is immediately started. The core of the step is to identify sensitive units and quantify the adjustment capacity, so as to provide accurate input for subsequent optimization control. The specific implementation process is as follows: All grid-connected operating units in the wind farm are divided into high-risk units and adjustable units according to the sensitivity of the units to the turbulence response, and the sensitivity is represented by the standard deviation of the output power change rate. The specific division process is as follows: the standard deviation of the output power change rate of each unit in a historical time window containing a plurality of historical sampling periods is calculated, and the units with a standard deviation higher than a preset sensitivity threshold are divided into high-risk units, and the remaining units are divided into adjustable units. The output power change rate of the unit in any sampling period is obtained by calculating the output power difference between the unit in the sampling period and the previous sampling period, and calculating the ratio of the difference to the adjacent sampling time interval, so as to obtain the output power change rate of the unit in the sampling period. In the embodiment, a simple method is used, and the volatility of the output power change rate is preferably used for quantification. In addition, historical turbulence data of the location where the unit is located can be combined to determine whether the region is often in a strong turbulence surge state, and all units in the region that are often in the state are divided into high-risk units.

[0054] Turbulence is a meteorological phenomenon with strong randomness and rapid change. Due to factors such as wake effect and terrain, the response speed and amplitude of units at different positions to turbulence are significantly different. The output power of a sensitive unit will fluctuate more sharply and frequently, which is the main source of the risk of over-limit power. The present application innovatively uses the volatility of the historical output power change rate to quantify the sensitivity, which is a scientific and simple calculation method and can effectively distinguish the characteristics of the units.

[0055] The method for determining the sensitivity threshold should be adjusted and optimized according to the specific conditions of the wind farm. Long-term operating data of the wind farm under typical strong turbulent weather conditions (but not severe enough to necessitate power throttling) should be extracted. The output power variation rate of all units in the entire site should be calculated over multiple time periods, and its distribution should be plotted. The sensitivity threshold can be set as the average value of all units plus a standard deviation, or as the 75th percentile of the entire distribution. The aim is to identify particularly sensitive units whose performance is significantly different from the majority. This approach ensures that the threshold is adapted to the actual layout and unit characteristics of the wind farm, dynamically and relatively identifying the most sensitive units, rather than using an absolute value that may not be applicable to all sites.

[0056] The adjustment potential includes the maximum upward power adjustment potential and the maximum downward power adjustment capacity; the formula for calculating the maximum upward power adjustment potential is as follows: ; in, This represents the maximum up-adjustment power potential of the i-th adjustable unit at the t-th sampling time in the next sampling period. This represents the theoretical maximum available power of the i-th adjustable unit at the t-th sampling time in the next sampling period. This represents the maximum allowable power change rate for the i-th adjustable unit. This value is provided by the unit manufacturer and is a fixed parameter. Let be the time length of the sampling period; the theoretical maximum available power is determined as follows: if the predicted wind speed at the t-th sampling moment in the next sampling period is not greater than the rated wind speed of the i-th adjustable unit, then the formula for calculating the theoretical maximum available power of the i-th adjustable unit at the t-th sampling moment in the next sampling period is: If the predicted wind speed is greater than the rated wind speed, it indicates that the unit may have reached or exceeded the rated power in the next sampling period. At this time, the theoretical maximum available power is the rated power of the unit, and i is the index of the adjustable unit.

[0057] The formula for calculating the maximum upsizing potential reflects a delicate balance between physical constraints and engineering practice. Its rationality stems primarily from a profound understanding of the nature of wind energy capture: the theoretical upper limit of power that a unit can theoretically increase at any given time fundamentally depends on current and near-future wind conditions. The physical limit represented by the term. This calculation intelligently switches the calculation model by judging whether the predicted wind speed exceeds the rated wind speed, ensuring the accuracy of the theoretical value. That is, below the rated wind speed, the wind energy cube law is followed, and above it, it is limited by the mechanical design rating of the unit, which is completely consistent with the actual operating characteristics of the wind turbine.

[0058] However, theoretically available power alone is insufficient; power regulation in any physical device is subject to inertia and cannot be completed instantaneously. (Formula introduction) Item, which is the quantification of the key engineering constraint of the dynamic response capability of the unit. It acknowledges the rigid limitation of the maximum response speed of the hardware such as the pitch system, the converter, etc. on the power ramping process. Finally, by taking the minimum of the two, the formula cleverly identifies the short board that constrains the ramping-up capability at a specific moment. The ramping-up potential is ultimately determined by the bottleneck of the two limiting conditions. For example, even if the unit has a strong ramping capability ( ), the theoretical power it can capture ( ) is limited if the wind speed is small; conversely, even if the wind is sufficient ( ), the hardware ramping capability of the unit limits its instantaneous power increase. Whether it is a case of having wind but the unit cannot climb up or the unit can climb but the wind is not enough, the formula can capture the real upper limit of the power increment that can be safely executed, thus ensuring the feasibility and safety of subsequent optimization instructions.

[0059] The formula for calculating the maximum down-regulation power capacity is as follows: ; where, is the maximum down-regulation power capacity of the i-th adjustable unit at the current moment, represents the actual output power of the i-th adjustable unit at the current moment, represents the minimum output power allowed by the i-th adjustable unit; the maximum down-regulation power capacity at all sampling times in the next sampling period is this value; this is because the down-regulation operation is an emergency intervention based on the current state (current power), and it is assumed that within a very short prediction period, the down-regulation capacity of the unit is relatively stable and available. Taking the minimum value is also to find the bottleneck of the down-regulation capability. It is ensured that the down-regulation instruction neither lets the unit fall below the minimum technical output nor exceeds the power reduction speed that the hardware can withstand.

[0060] Secondly, similar to the ramping-up potential, the down-regulation is also subject to the constraints of the physical properties of the equipment. The term explicitly states the maximum power reduction amplitude that the unit can achieve within a unit control period, which is determined by the actuation rate of the pitch mechanism, the energy dissipation capacity, and other hardware indicators. The formula adopts the minimum value principle, which is also to accurately locate the constraints of the current down-regulation operation. The calculation formula of the maximum down-regulation power capacity reflects the deep understanding of the emergency and conservative nature of the grid's down-regulation demand. Its rationality is based on two solid foundations. First is the operational safety boundary, i.e. The item defines the absolute spatial lower limit of the down-regulation operation, ensures that any power reduction instruction will not endanger the stable operation of the unit itself, avoids triggering protective off-grid or damaging auxiliary systems due to too low power, and is the premise for maintaining the grid-connected state of the wind farm. It is worth noting that the down-regulation capacity is calculated as a constant value determined at the current time and remaining unchanged in the next entire prediction period. This processing method has high engineering practicability, which is based on a reasonable assumption that in the short emergency control window triggered by strong turbulence, the fast down-regulation decision based on the current state is more reliable and stable than frequently re-predicting a small down-regulation capacity, which is conducive to the fast response and execution of the control system.

[0061] If it is judged that there is no risk of exceeding the limit, an instruction is sent to all units to maintain normal operation.

[0062] Step 4: Based on the genetic algorithm, the optimal adjustment trajectory is calculated with the smooth predicted trajectory of the total active power of the wind farm as the target and the risk of exceeding the limit as the constraint, and the optimal control instruction is generated based on the optimal adjustment trajectory; the optimal control instruction includes a preventive power limiting instruction applied to the high-risk unit and a compensatory adjustment instruction applied to the adjustable unit; In this embodiment, obtaining the optimal control instruction specifically includes: based on the predicted trajectory of the total active power of the wind farm in the next sampling period, the predicted trajectory being composed of total active power prediction values corresponding to all sampling time points in the next sampling period; For each sampling time point in the predicted trajectory, the upper limit of the total active power adjustment at the time point is determined according to the sum of the maximum up-regulation power potentials of all adjustable units, and the lower limit of the total active power adjustment at the time point is determined according to the sum of the maximum down-regulation power capacities of all adjustable units, to form a power allowable value range for each sampling time point; The genetic algorithm is used to optimize and solve the target of generating a smooth and non-exceeding-limit adjusted total active power trajectory, and the specific process is as follows: Initialization step: randomly generate multiple individuals to form an initial population, and the population size is usually related to the dimension of the decision variable (i.e. the number of sampling time points). Preferably, the value range is 50 to 200 individuals, i.e. the population size is five to ten times the number of sampling time points. If the population is too large, the calculation efficiency is low, which affects the real-time performance of the control; if the population is too small, the diversity is insufficient, which easily falls into a local optimal solution and cannot find a globally optimal smooth trajectory. The above value range is an engineering compromise between solution accuracy and calculation real-time performance. Each individual represents a candidate adjusted total active power trajectory, which is composed of the same number of sampling time points as the predicted trajectory, and the total active power value at each sampling time point is taken as its gene value, which is randomly generated within the power allowable value range of the sampling time point.

[0063] Step of calculating fitness: the fitness value of an individual is the inverse of the sum of the absolute values of the power rate of change between adjacent sampling instants of the trajectory represented by the individual; the greater the fitness value, the smoother the trajectory.

[0064] Step of setting constraint: before calculating fitness, first determine whether the trajectory represented by the individual has the risk of exceeding the limit; if there is a risk of exceeding the limit, directly set the fitness of the individual to a pre-set penalty value, so that it is eliminated in the subsequent selection; the penalty value is set to a very large negative number (such as ) or zero as the fitness value of the individual that violates the constraint. This is a death penalty mechanism. Since the constraint (especially the grid limit) is hard and must be met, any solution that violates the constraint is infeasible. By setting its fitness to a very poor value, it can be ensured that these infeasible solutions are completely eliminated in the selection process, thereby guiding the population to evolve towards the feasible region.

[0065] Step of genetic operation: select, cross and mutate the population to generate a new generation; select based on the fitness value using the tournament selection method, which is the most commonly used selection method; the higher the fitness of an individual, the greater the chance of being selected to the next generation, which meets the natural law of survival of the fittest and can ensure the optimization direction. Cross operation is a multi-point crossover between individual gene positions with a certain probability; mutation operation randomly changes the value of any gene position in the individual with a certain probability, and the mutated value is within the power allowed value range at that sampling instant. Here, the crossover probability is usually high, ranging from 0.7 to 0.9. Crossover is the main method to generate new individuals, and a high crossover probability helps to efficiently explore the solution space, promotes the exchange and spread of good genes, and speeds up convergence. The mutation probability is usually low, ranging from 0.01 to 0.1. Mutation operation introduces population diversity, preventing the algorithm from falling into local optima too early. However, too high a mutation probability will make the algorithm degenerate into random search. Therefore, a lower mutation probability is sufficient to introduce new exploration while maintaining good patterns.

[0066] Step of iteration termination: repeat the steps of calculating fitness value to genetic operation until the maximum number of iterations is reached; the maximum number of iterations is preferably in the range of 100 to 500. The number of iterations is the main termination condition of the algorithm. A sufficient number of iterations is a prerequisite for convergence. Since wind farm control has high real-time requirements, it must be completed before the next sampling period arrives, so the number of iterations must be set in offline simulation based on historical data to ensure that a satisfactory solution can be converged within a limited computation time.

[0067] Output result step: the total active power trajectory represented by the individual with the highest fitness in the final population is determined as the optimal adjustment trajectory after adjustment; According to the difference between the optimal adjustment trajectory and the original predicted trajectory at each sampling time, the adjustment amount of the total active power of the wind farm is calculated; According to the adjustment potential of each adjustable unit, the total adjustment amount is distributed by weight to generate compensatory adjustment instructions for each adjustable unit; if the adjustment amount of the total active power of the wind farm is positive, it indicates that the output power of the wind farm is to be increased, and the weight is determined according to the ratio of the maximum up-regulation power potential of each adjustable unit to the adjustment amount of the total active power; distributing according to the proportion of the maximum up-regulation potential is a fair and efficient way. Units with greater potential bear more of the increased load, avoiding the rapid reaching of the limit of some units and losing adjustment capacity, optimizing the use of adjustment resources. If the adjustment amount of the total active power of the wind farm is negative, it indicates that the output power of the wind farm is to be reduced, and the weight is determined according to the ratio of the maximum down-regulation power capacity of each adjustable unit to the adjustment amount of the total active power; distributing according to the maximum down-regulation capacity. This ensures that the load reduction instructions will not cause any unit to be forced to reduce below its minimum technical output, thereby ensuring the safe and stable operation of the unit, while also effectively utilizing the down-regulation capacity of all units. At the same time, for high-risk units, a preventive power limit instruction is generated, whose value is the current output power of the unit minus a power buffer margin preset according to its rated power. The power buffer margin is usually set to a percentage of the rated power of the unit, preferably in the range of [5%, 15%]. High-risk units are sensitive to turbulence, and their power itself has a large natural fluctuation. Reserving a buffer margin based on the rated power is equivalent to providing a "buffer zone" or "safety cushion" for its natural fluctuations. This can effectively limit the fluctuation of high-risk units to a smaller absolute range, thereby reducing their contribution to the total power fluctuation of the entire field from the source, enhancing the robustness and preventability of the control The technical effects of implementing the optimization control step are significant and multi-level. First, the most direct effect is to greatly enhance the grid-friendliness of offshore wind farms under extreme weather conditions such as strong turbulence. Through forward-looking optimization based on ultra-short-term prediction and genetic algorithm, this method can generate a smooth power output trajectory that strictly meets the requirements of grid regulations, actively avoiding the risk of triggering grid protection actions due to excessive power change rate or absolute power exceeding limits. This not only ensures the continuous operation of the wind farm itself and reduces the loss of power generation due to disconnection, but more importantly, it maintains the stability of the grid frequency and enhances the confidence and ability of the grid to accommodate a high proportion of wind power, providing a solid guarantee for the safe and stable operation of the grid.

[0068] Further, the embodiment is not a simple power reduction of all units, but a fine management based on the characteristics of the units. By dividing the units into high-risk units and adjustable units, and applying preventive power reduction instructions and compensatory adjustment instructions to the two, the method ensures smoothness while also taking into account power generation benefits and equipment safety. For adjustable units, the real-time adjustment potential is used to allocate tasks by weight, fully and reasonably tapping the adjustment capacity of each unit, avoiding the situation where some units reach the limit too early while others are idle, thereby maximizing the utilization efficiency of the droop space under the premise of meeting grid constraints, and improving the overall power generation efficiency of the entire field. Thus forming a closed-loop, self-adaptive real-time control loop. The system can continuously monitor, predict, optimize and execute the wind farm state, forming a dynamic adjustment virtuous cycle. The wind farm can adapt to the complex and changeable marine weather environment for a long time and actively, turning one-time control actions into a stable function for continuous operation, thereby fundamentally improving the reliability, economy and dispatchability of offshore wind farm operation, and providing effective technical support for the intelligent operation of large-scale offshore wind power.

[0069] Step 5: After the optimal adjustment trajectory obtained in the previous step, the central controller (such as the wind farm energy management system) will generate two types of specific and executable control instructions, and send them to the corresponding wind turbine controller (such as PLC or main controller) through the internal communication network of the wind farm (such as optical fiber Ethernet, industrial wireless network, etc.). The compensatory adjustment instruction is used to instruct each adjustable unit to change the output power according to the allocated power adjustment amount. The instruction is a specific power set value or power increment instruction. In combination with the previous step, it can be seen that the power increment instruction (i.e. adjustment amount) is distributed here. The instruction format should include the unit number, instruction type (such as power increment), and adjustment amount. After receiving the instruction, the adjustable unit controller will immediately switch the target of its inner loop control circuit (such as variable pitch control and torque control) from maximum power point tracking (MPPT) mode to power control mode. The controller will smoothly adjust its output power to the target value required by the instruction without exceeding the limit of its maximum power change rate. This process is achieved by adjusting the pitch angle and generator torque, which belongs to a mature unit-level control technology. The preventive power limiting instruction is used to instruct each high-risk unit to limit its output power to a lower level than the current value. The instruction is also a power set value, but its value is lower than the actual output power of the unit. After receiving the instruction, the high-risk unit controller will also switch to power control mode and limit its output power to a lower level specified by the instruction. The purpose of this is not to let it participate in frequent adjustment, but to create a buffer space for it in advance in case of sudden power surge caused by turbulence. For example, if a unit may instantaneously surge by 2MW due to turbulence, by pre-reducing its power by 0.5MW (setting a buffer margin), the impact of its instantaneous surge on the total power of the wind farm can be reduced from 2MW to 1.5MW, thereby greatly reducing the risk of the total output power of the wind farm exceeding the grid limit.

[0070] The present application is a dynamic and continuous closed-loop control process. After the control instructions are issued, the system does not stop working. Instead, it automatically and repeatedly performs the whole process of steps 1 to 5 until it is determined that the wind farm is not in a state of strong turbulence surge. At this time, the control system will send instructions to all units to remove their power limits and restore them to normal maximum power point tracking operation mode, thereby maximizing the power generation effect of the wind farm under the premise of ensuring the safety of the power grid. This automatic exit mechanism based on clear criteria ensures that the present application is only activated when necessary, avoiding unnecessary loss of power generation.

[0071] Please refer to Figure 2 , the present application further provides a kind of offshore wind farm active time sequence optimization control system, the kind of offshore wind farm active time sequence optimization control system for realizing the offshore wind farm active time sequence optimization control method described above, comprising: a real-time monitoring and turbulence identification module, configured to calculate total active power of the offshore wind farm in real time according to wind speed and wind direction data at the outlet of the offshore wind farm in a current sampling period, wind speed in each wind turbine nacelle and output power, and evaluate average turbulence intensity of the offshore wind farm; and analyze average turbulence intensity, wind speed variation and total active power fluctuation of the offshore wind farm in a plurality of continuous sampling periods to determine whether the offshore wind farm is in a strong turbulence surge state; a power overrun risk prediction module, configured to, if the offshore wind farm is in the strong turbulence surge state, generate a predicted trajectory of total active power of the offshore wind farm in a next sampling period based on ultra-short-term prediction of wind speed and a preset power transfer function of the offshore wind farm, and determine whether there is an overrun risk in the predicted trajectory; a turbine grouping and potential evaluation module, configured to, if there is an overrun risk, divide all grid-connected operating turbines in the offshore wind farm into high-risk turbines and adjustable turbines according to sensitivity of the turbines to turbulence response, and obtain adjustment potential of the adjustable turbines under current wind speed; a power optimization control module, configured to, taking the predicted trajectory of total active power of the offshore wind farm as a target and avoiding the overrun risk as a constraint, calculate an optimal adjustment trajectory based on a genetic algorithm, and generate optimal control instructions based on the optimal adjustment trajectory; the optimal control instructions include preventive power limiting instructions applied to the high-risk turbines and compensatory adjustment instructions applied to the adjustable turbines; an instruction execution and feedback module, configured to issue the optimal control instructions to corresponding wind turbine controllers, and repeat the above process in the next sampling period until it is determined that the offshore wind farm is not in the strong turbulence surge state.

[0072] The above formulas are all dimensionless numerical calculations, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and preset parameters in the formula are set by a person skilled in the art according to actual conditions.

[0073] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially. Those skilled in the art can realize that units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software methods depends on specific application and design constraints of the technical solutions.

[0074] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, and may be located in one place, or distributed on multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment of the present application according to actual needs.

[0075] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method of active timing optimization control for an offshore wind farm, characterized in that, The specific steps include: Step 1: According to the wind speed and wind direction data at the outlet of the offshore wind farm in the current sampling period, and the wind speed and output power of each wind turbine nacelle, the total active power of the wind farm is calculated in real time, and the average turbulence intensity of the wind farm is evaluated; and the average turbulence intensity, wind speed variation and total active power fluctuation of the wind farm in a plurality of continuous sampling periods are analyzed to determine whether the wind farm is in a strong turbulence surge state; Step 2: If the wind farm is in a strong turbulence surge state, based on the ultra-short-term prediction of wind speed and the preset wind farm power transfer function, the predicted trajectory of the total active power of the wind farm in the next sampling period is generated; whether there is an over-limit risk in the predicted trajectory is determined; Step 3: If there is an over-limit risk, according to the sensitivity of the unit to turbulence response, all grid-connected units in the wind farm are divided into high-risk units and adjustable units, and the adjustment potential of the adjustable units under the current wind speed is obtained; Step 4: Taking the predicted trajectory of the smooth total active power of the wind farm as the target, and avoiding over-limit risk as the constraint, the optimal adjustment trajectory is calculated based on the genetic algorithm, and the optimal control instruction is generated accordingly; the optimal control instruction includes preventive power limiting instruction applied to high-risk units and compensatory adjustment instruction applied to adjustable units; Step 5: The optimal control instruction is sent to the corresponding wind turbine controller, and the above process is repeated in the next sampling period until it is determined that the wind farm is not in a strong turbulence surge state.

2. A method of active time sequence optimization control for an offshore wind farm according to claim 1, characterized in that, The outlet of the offshore wind farm is the outlet position of the wind farm power collection line and the offshore booster station, and whether the wind farm is in the strong turbulence surge state specifically includes: A wind measurement tower is installed at the outlet of the offshore wind farm to collect wind speed and wind direction data at the outlet of the offshore wind farm at a first preset sampling frequency, and the first preset sampling frequency is not less than 1Hz; through the wind farm monitoring system, real-time data messages uploaded by all wind turbine controllers in the wind farm are received at a second preset sampling frequency; the real-time data messages at least include the nacelle wind speed measured by the ultrasonic anemometer installed at the top of the wind turbine nacelle, and the output power measured by the unit converter sensor, and the second preset sampling frequency is not less than 0.5Hz; The average value of the output power of each grid-connected wind turbine in the wind farm at all sampling times in the current sampling period is calculated, and the sum of the average values of the output power of all wind turbines is taken as the total active power of the wind farm; The average turbulence intensity of the wind farm is evaluated as follows: According to the wind speed data, the average turbulence intensity of the current sampling period is calculated, and the formula is as follows: ； wherein, is the average turbulence intensity of the current sampling period, is the standard deviation of the wind speed at all sampling instants in the current sampling period, is the average value of the wind speed at all sampling instants in the current sampling period, k denotes the current sampling period; When the following four conditions are met simultaneously, it is determined whether the wind farm is in a strong turbulence surge state, specifically: a. The average turbulence intensity of the current sampling period is greater than the preset first threshold value; b. The growth rate of the average turbulence intensity of the current sampling period relative to the average value of the previous M sampling periods is greater than the preset second threshold value; c. The absolute value of the average wind speed variation rate monitored in the current sampling period is greater than the preset third threshold value; d. The standard deviation of the total active power of the wind farm in the previous M sampling periods is greater than the preset fourth threshold value.

3. A method of active time sequence optimization control for an offshore wind farm according to claim 2, characterized in that: The calculation formula of the growth rate is as follows: ； wherein, represents the growth rate, is the average turbulence intensity mean value for the previous M sampling periods; The calculation formula of the average wind speed change rate is as follows: ； wherein is the average wind speed variation rate, denotes the average value of the wind speed at all sampling instants in the previous sampling period, is the time length of the sampling period.

4. The method of claim 1, wherein, The determination of whether the over-limit risk exists specifically includes: The ultra-short-term prediction refers to the prediction of the wind speed and the wind direction at each sampling time in the next sampling period; based on the monitored wind speed and wind direction data at the outlet of the wind farm, the wind speed sequence and the wind direction sequence in the current sampling period and the previous M sampling periods are arranged to form a sample set, an ARIMA prediction model is established by using time series analysis, the wind speed sequence and the wind direction sequence in the previous N sampling periods in the sample set are taken as inputs, and the wind speed sequence and the wind direction sequence in the next sampling period are taken as outputs, the model is trained; and the wind speed sequence and the wind direction sequence in the current sampling period and the previous N-1 sampling periods are input into the trained model to obtain the predicted wind speed sequence and wind direction sequence in the next sampling period; the wind speed sequence refers to a sequence formed by the wind speed data at all sampling times in a sampling period, and the wind direction sequence refers to a sequence formed by the wind direction data at all sampling times in a sampling period; The unit data of the current grid-connected operation is obtained from the wind farm monitoring system, and the predicted wind speed sequence is substituted into the preset wind farm power transfer function to calculate the total active power at each sampling time in the next sampling period to form a predicted trajectory, and the wind farm power transfer function is as follows: ； wherein, is the total active power of the wind farm at the t-th sampling moment in the next sampling period, is the current air density, determined according to the temperature and air pressure data of the location where the wind farm is located, is the wind wheel swept area, is the wind energy utilization coefficient, is the predicted wind speed at the t-th sampling moment in the next sampling period, t is the index of the sampling moment in the sampling period, represents the number of grid-connected units in the wind farm, is the average efficiency factor of the wind farm, is the array effect reduction coefficient; The over-limit risk refers to the risk of exceeding the specified limit value of the power grid, including two judgment conditions, condition one is to determine whether there is a power change rate over-limit risk, specifically: if the absolute value of the total active power growth rate of adjacent sampling times in the next sampling period is greater than a preset threshold one, it is determined that there is a power change rate over-limit risk; the preset threshold one is the maximum power change rate limit value per minute specified by the power grid; the total active power change rate of adjacent sampling times is obtained by calculating the total active power difference between the next sampling time and the previous sampling time in adjacent sampling times, and calculating the ratio of the difference value to the time interval of adjacent sampling times as the total active power change rate of adjacent sampling times; Condition two is to determine whether there is a power absolute value over-limit risk, specifically: if the absolute value of the total active power of a certain sampling time in the next sampling period is greater than a preset threshold two, it is determined that there is a power absolute value over-limit risk; the preset threshold two is the upper limit value of the active power specified by the power grid; If any condition is met, it is determined that there is an over-limit risk.

5. A method of active time sequence optimization control for an offshore wind farm according to claim 4, characterized in that, The adjustment potential is obtained specifically by: If there is an over-limit risk, the adjustment mechanism of the total active power of the wind farm is started; According to the sensitivity of the unit to the turbulent flow, all grid-connected units in the wind farm are divided into high-risk units and adjustable units, and the sensitivity is represented by the standard deviation of the output power change rate of the unit. The specific division process is as follows: calculate the standard deviation of the output power change rate of each unit in a historical time window containing several historical sampling periods, and divide the units with the standard deviation higher than the preset sensitivity threshold into high-risk units, and divide the remaining units into adjustable units; the output power change rate of the unit in any sampling period is obtained as follows: calculate the output power difference between the unit in the sampling period and the previous sampling period, and calculate the ratio of the difference to the adjacent sampling time interval as the output power change rate of the unit in the sampling period; The adjustment potential includes a maximum up-regulation power potential and a maximum down-regulation power capacity; the formula for calculating the maximum up-regulation power potential is as follows: ； wherein, represents the maximum up-regulation power potential of the ith adjustable unit at the tth sampling moment in the next sampling period, represents the theoretical maximum available power of the ith adjustable unit at the tth sampling moment in the next sampling period, represents the maximum power change rate allowed by the ith adjustable unit, is the time length of the sampling period; the theoretical maximum available power is determined as follows: if the predicted wind speed at the tth sampling moment in the next sampling period is not greater than the rated wind speed of the ith adjustable unit, the theoretical maximum available power of the ith adjustable unit at the tth sampling moment in the next sampling period is calculated according to the following formula: if the predicted wind speed is greater than the rated wind speed, the theoretical maximum available power is the rated power of the unit, and i is the index of the adjustable unit; The formula for calculating the maximum down-regulation power capacity is as follows: ； wherein, is the maximum down-regulation power capacity of the i-th adjustable unit at the current time, is the actual output power of the i-th adjustable unit at the current time, is the minimum output power allowed for the i-th adjustable unit; the maximum down-regulation power capacity at all sampling times in the following sampling period is this value; If it is judged that there is no over-limit risk, instructions are sent to all units to maintain normal operation.

6. A method of active time sequence optimization control for an offshore wind farm according to claim 5, characterized in that, The optimal control instruction is obtained, and specifically includes: Based on the predicted trajectory of the total active power of the wind farm in the next sampling period, the predicted trajectory is composed of the total active power predicted value corresponding to all sampling time points in the next sampling period; For each sampling time point in the predicted trajectory, the upper limit of the total active power adjustment at the time point is determined according to the sum of the maximum up-regulation power potential of all adjustable units, and the lower limit of the total active power adjustment at the time point is determined according to the sum of the maximum down-regulation power capacity of all adjustable units, to form the power allowable value range of each sampling time point; A genetic algorithm is used to generate a smooth and over-limit risk-free adjusted total active power trajectory as the target for optimization and solving, and the specific process is as follows: Initialization step: randomly generate multiple individuals to form an initial population, each individual represents a candidate adjusted total active power trajectory, the trajectory is composed of the same number of sampling time points as the predicted trajectory, and the total active power value at each sampling time point is taken as its gene value, which is randomly generated within the corresponding power allowable value range; Adaptability calculation step: the smoothness of the adjusted trajectory represented by the individual is taken as its adaptability value, and the smoothness is quantified by calculating the reciprocal of the sum of the absolute values of the power change rates of adjacent sampling time points in the trajectory; Constraint condition setting step: before calculating the adaptability, first judge whether the adjusted trajectory represented by the individual has over-limit risk; if there is over-limit risk, directly set the adaptability of the individual to a preset penalty value, so that it is eliminated in the subsequent selection; Genetic operation step: selection, crossover and mutation operations are performed on the population to generate a new generation of population; selection is performed based on the adaptability value using the tournament selection method; crossover operation is performed between the gene positions of the individuals with a certain probability; the mutation operation randomly changes the value of any gene position in the individual with a certain probability, and the mutated value is within the power allowable value range of the sampling time point; Iteration termination step: repeat the adaptability value calculation step to the genetic operation step until the maximum iteration number is reached; An output result step: determining the adjusted total active power trajectory represented by the individual with the highest fitness in the final population as the optimal adjusted trajectory; According to the difference between the optimal adjusted trajectory and the original predicted trajectory at each sampling time, the adjustment amount of the total active power of the wind farm is calculated; According to the adjustment potential of each adjustable unit, the total adjustment amount is distributed to each adjustable unit to generate a compensatory adjustment instruction; if the adjustment amount of the total active power of the wind farm is positive, it indicates that the output power of the wind farm is increased, and the weight is determined according to the ratio of the maximum up-regulation power potential of each adjustable unit to the adjustment amount of the total active power; if the adjustment amount of the total active power of the wind farm is negative, it indicates that the output power of the wind farm is reduced, and the weight is determined according to the ratio of the maximum down-regulation power capacity of each adjustable unit to the adjustment amount of the total active power; at the same time, for high-risk units, a preventive power limiting instruction is generated, and the value is the current output power of the unit minus a power buffer margin preset according to its rated power.

7. A method of active time sequence optimization control for an offshore wind farm according to claim 6, characterized in that, The compensatory adjustment instruction is used to instruct each adjustable unit to change the output power according to the power adjustment amount allocated to it; the preventive power limiting instruction is used to instruct each high-risk unit to limit its output power to a level lower than the current value.

8. An offshore wind farm active timing optimization control system, characterized in that, The offshore wind farm active timing optimization control system is used to implement the offshore wind farm active timing optimization control method of any one of claims 1-7, comprising: A real-time monitoring and turbulence identification module is configured to calculate the total active power of the wind farm and evaluate the average turbulence intensity of the wind farm in real time according to the wind speed and direction data at the outlet of the offshore wind farm and the wind speed and output power of each wind turbine generator cabin in the current sampling period; and analyze the average turbulence intensity, wind speed variation and total active power fluctuation of the wind farm in a plurality of consecutive sampling periods to determine whether the wind farm is in a strong turbulence surge state; A power overrun risk prediction module is configured to generate a predicted trajectory of the total active power of the wind farm in the next sampling period based on the ultra-short-term prediction of the wind speed and the preset power transfer function of the wind farm if the wind farm is in a strong turbulence surge state; and determine whether there is an overrun risk in the predicted trajectory; A unit grouping and potential evaluation module is configured to divide all grid-connected units in the wind farm into high-risk units and adjustable units according to the sensitivity of the units to turbulence response if there is an overrun risk, and obtain the adjustment potential of the adjustable units under the current wind speed; A power optimization control module is configured to calculate an optimal adjustment trajectory based on a genetic algorithm with the smooth predicted trajectory of the total active power of the wind farm as the target and the avoidance of the overrun risk as the constraint, and generate optimal control instructions based on the optimal adjustment trajectory; the optimal control instructions include preventive power limiting instructions for high-risk units and compensatory adjustment instructions for adjustable units; An instruction execution and feedback module is configured to issue the optimal control instructions to the corresponding wind turbine generator controller, and repeat the above process in the next sampling period until it is determined that the wind farm is not in a strong turbulence surge state.

Images