Offshore wind power plant control method and system

Abstract

The invention discloses an offshore wind power plant control method and system, and relates to the technical field of wind power generation control, and the method comprises the steps: carrying out the collection through a sensor array disposed at a wind generating set, and forming a three-dimensional environment parameter vector; inputting the real-time environment parameters into a pre-trained classification model, and matching the real-time environment parameters to a plurality of preset environment operation states; constructing a Markov state transition probability matrix based on the historical operation data, and recording environmental parameter change rate characteristics during state transition; determining the adaptation degree of the preset control parameters of the wind generating set in each environment operation state to other environment operation states; fusing to generate a mixed prediction result; and triggering a control strategy according to the comprehensive judgment of the prediction result and the adaptation degree evaluation. The method has the advantages that dynamic Markov state transition prediction and cross-state adaptation degree evaluation are combined, so that the adaptive capacity of the offshore wind power plant to complex sea conditions is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation control, and specifically relates to a control method and system for an offshore wind farm. Background Art

[0002] As an important part of clean energy, offshore wind power generation has a complex and changeable operating environment, and often faces multiple challenges such as sudden changes in wind speed, wave impact, and salt spray corrosion. Traditional control methods mostly rely on single environmental parameters or static thresholds for decision-making, and it is difficult to dynamically adapt to comprehensive environmental changes, resulting in fluctuations in power generation efficiency, increased equipment losses, and even potential safety hazards under extreme working conditions. Especially under harsh sea conditions, the existing technologies lack the ability to synergistically analyze multi-dimensional environmental parameters and are difficult to predict the state migration trend in a timely manner, resulting in lagging control strategies and restricting the reliability and economy of offshore wind farms.

[0003] In current solutions, environmental state recognition is mostly based on fixed rules or single machine learning models, and the temporal correlation of historical operation data and the state transition law are insufficiently mined. At the same time, the adjustment of control parameters often only considers the current working conditions, without combining the future environmental evolution trend and the cross-state adaptability evaluation, which is prone to frequent start-stop or sub-optimal control. In addition, the parameter update efficiency of the traditional Markov prediction model is insufficient in a dynamic environment, and it is unable to effectively integrate real-time data and historical state transition characteristics, resulting in limited prediction accuracy. There is an urgent need for a closed-loop control system that integrates multi-source environmental perception, intelligent state classification, and dynamic prediction decision-making to improve the adaptive operation ability and risk prevention and control level under complex sea conditions. Summary of the Invention

[0004] To solve the above technical problems, a control method and system for an offshore wind farm are provided. This technical solution solves the problem that the existing technologies lack the ability to synergistically analyze multi-dimensional environmental parameters, are difficult to predict the state migration trend in a timely manner, resulting in lagging control strategies and restricting the reliability and economy of offshore wind farms.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A control method for an offshore wind farm, comprising:

[0007] Collecting real-time wind speed, significant wave height, and salt spray concentration data through a sensor array deployed on a wind turbine generator to form a three-dimensional environmental parameter vector;

[0008] Inputting the real-time environmental parameters into a pre-trained classification model to match to a plurality of preset environmental operating states, at least including a normal power generation mode, a storm warning mode, and an extreme shutdown mode;

[0009] Construct a Markov state transition probability matrix based on historical operation data, and record the characteristics of the environmental parameter change rate during each state transition;

[0010] Based on the preset control parameters of the wind turbine for each environmental operating state, determine the adaptability of the preset control parameters of the wind turbine in each environmental operating state to other environmental operating states;

[0011] Predict the future environmental parameter changes based on the characteristics of the environmental parameter change rate during each state transition, and fuse with the state transition probability matrix to generate a mixed prediction result;

[0012] According to the comprehensive judgment of the prediction result and the adaptability evaluation, trigger one of the control strategies of maintaining operation, parameter adjustment or emergency shutdown.

[0013] Preferably, the specific training process of the classification model is as follows:

[0014] Set the environmental parameter clustering center vector for each environmental operating state respectively;

[0015] Based on the historical control logs of the offshore wind farm, determine several environmental parameter vectors when each environmental operating state is adjusted, and record them as sample environmental parameter vectors;

[0016] Train based on multiple sample environmental parameter vectors and environmental parameter clustering center vectors to determine the environmental parameter boundary values for each environmental operating state;

[0017] Based on the environmental parameter clustering center vector and environmental parameter boundary values, determine the classification model of the environmental operating state.

[0018] Preferably, the construction of the Markov state transition probability matrix based on historical operation data and the recording of the characteristics of the environmental parameter change rate during each state transition specifically include:

[0019] Based on the control attributes of the wind turbine, determine an analysis duration;

[0020] Set a sample collection duration, and set the time period within the sample collection duration closest to the current moment as the sample collection period;

[0021] Based on the analysis duration, divide the sample collection period into several analysis periods;

[0022] Based on the classification model, determine the environmental state of each analysis period;

[0023] Based on the environmental state of each analysis period, construct a Markov state transition probability matrix, P = [p ij K×K ,p ij ​$P_{ij}$ is the Markov state transition probability from the $i$-th environmental operating state to the $j$-th environmental operating state, and $K$ is the total number of preset environmental operating states;

[0024] Among them, $N_i$ i is the total number of analysis periods in the $i$-th environmental state, and $N_{ij}$ i→j is the number of adjacent analysis periods in which the environmental state changes from the $i$-th environmental state to the $j$-th environmental operating state;

[0025] Respectively, count the average value of environmental parameters in each analysis period and the change rate of environmental parameters over time in each analysis period, and summarize to obtain the change rate characteristics of environmental parameters at the time of state transition.

[0026] Preferably, the specific process of determining the adaptability of the preset control parameters of the wind turbine in each environmental operating state to other environmental operating states includes:

[0027] Determine the performance indicators of the wind turbine under the preset control parameters of the wind turbine in each environmental operating state, and the performance indicators at least include yaw delay, power fluctuation rate, and peak blade stress;

[0028] Then the calculation formula of the adaptability is:

[0029]

[0030] In the formula, $\alpha_{kh}$ k→h is the adaptability of the preset control parameters of the wind turbine in the $k$-th environmental operating state to the $h$-th environmental operating state, $m$ is the total number of performance indicators, $x_{kn}$ is the $n$-th performance indicator corresponding to the preset control parameters of the wind turbine in the $k$-th environmental operating state, and $y_{hn}$ n is the $n$-th performance indicator corresponding to the preset control parameters of the wind turbine in the $h$-th environmental operating state, and $w_n$

[0031] Preferably, the specific process of predicting the change of environmental parameters in the future period based on the change rate characteristics of environmental parameters at the time of each state transition includes:

[0032] Based on the real-time parameter value of the environmental parameter collected in real time and the real-time change rate of the environmental parameter, and the average value of the environmental parameter in each analysis period and the change rate of the environmental parameter over time, solve the vector distance to obtain the environmental similarity distance index between the real-time environment and each analysis period;

[0033] Count the environmental operating state in the next analysis period of each analysis duration;

[0034] Calculate the transition fitting probability of each environmental operation state within the future analysis duration based on the environmental similarity distance index in the real-time environment and all analysis durations, as well as the environmental operation state in the next analysis duration of each analysis duration.

[0035] Specifically, the calculation formula is:

[0036]

[0037] Among them, G i is the transition fitting probability of the i-th environmental operation state within the future analysis duration, F i is the set of analysis durations of the i-th environmental operation state in the next analysis duration, c is the c-th element in F i in, E is the set of all analysis durations, o is the o-th element in E, L max is the maximum value in the environmental similarity distance index between the real-time environment and all analysis durations, L min is the minimum value in the environmental similarity distance index between the real-time environment and all analysis durations, L c is the environmental similarity distance index between the real-time environment and c.

[0038] Preferably, the specific steps for fusing with the state transition probability matrix to generate a mixed prediction result are as follows:

[0039] Determine the environmental operation state at the current moment. Based on the environmental operation state at the current moment, determine the Markov state transition probability of transitioning to each environmental operation state next from the Markov state transition probability matrix.

[0040] Respectively set the mixing weights of the transition fitting probability and the Markov state transition probability, and perform weighted summation of the transition fitting probability and the Markov state transition probability to obtain a mixed prediction result.

[0041] Preferably, the specific steps for triggering one of the control strategies of maintaining operation, parameter adjustment, or emergency shutdown according to the comprehensive judgment of the prediction result and the fitness evaluation are as follows:

[0042] If the prediction probability of the extreme shutdown mode in the mixed prediction result exceeds 0.5, trigger an emergency shutdown;

[0043] Otherwise, use the prediction probability of each environmental operation state in the mixed prediction result as the weight, and perform weighted summation on the fitness between the preset control parameters of the wind turbine generator under each environmental operation state and the environmental operation state to obtain the predicted fitness between the preset control parameters of the wind turbine generator under each environmental operation state and the environment;

[0044] If the environmental operation state corresponding to the maximum value of the predicted fitness is the same as the current environmental operation state, trigger maintaining operation;

[0045] If the environmental operating state corresponding to the maximum predicted adaptation degree is different from the current environmental operating state, parameter adjustment is triggered, and the control parameters of the offshore wind farm are adjusted to the preset control parameters of the wind turbine generator under the environmental operating state corresponding to the maximum predicted adaptation degree.

[0046] Furthermore, an offshore wind farm control system is proposed for implementing the offshore wind farm control method as described above, including:

[0047] A sensor array, deployed on the wind turbine generator, for collecting real-time data of wind speed, significant wave height, and salt fog concentration;

[0048] A data processing module for integrating the collected real-time data into a three-dimensional environmental parameter vector;

[0049] A state classification module with a pre-trained classification model, which matches the three-dimensional environmental parameter vector to a plurality of preset environmental operating states, including normal power generation mode, storm warning mode, and extreme shutdown mode;

[0050] A Markov analysis module for constructing a state transition probability matrix based on historical operation data and storing the environmental parameter change rate characteristics during each state transition;

[0051] An adaptation degree evaluation module for calculating the control adaptation degree of each environmental operating state to other states according to the preset control parameters;

[0052] An environmental prediction module for generating a mixed prediction result by combining the real-time environmental parameter change rate and the state transition probability matrix;

[0053] A control strategy execution module for triggering one of the control instructions of maintaining operation, parameter adjustment, or emergency shutdown according to the mixed prediction result and the adaptation degree evaluation result.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] By integrating real-time sensing of multi-source environmental parameters and intelligent state classification, combining dynamic Markov state transition prediction and cross-state adaptability evaluation, the present invention significantly improves the adaptive ability of offshore wind farms to complex sea conditions. Its multi-dimensional environmental parameter collaborative analysis and hybrid prediction mechanism effectively improve the early warning accuracy of extreme working conditions, avoiding misjudgment or response lag caused by traditional static thresholds; by predicting the environmental evolution trend and the adaptability of cross-state control parameters, dynamically optimizing the operation strategy of the generator set, reducing unnecessary shutdown times and frequent parameter adjustments, while ensuring equipment safety, improving power generation efficiency. In addition, the fusion prediction model based on historical data and real-time change rate enhances the robustness of the system to sudden environmental disturbances, realizing reduced power generation losses and lower equipment maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flowchart of a control method for an offshore wind farm proposed by this solution;

[0057] Figure 2 It is a flowchart of the specific training method of the classification model proposed by this solution;

[0058] Figure 3 It is a flowchart of the method for constructing the Markov state transition probability matrix and recording the environmental parameter change rate characteristics when each state transitions proposed by this solution;

[0059] Figure 4 It is a flowchart of the method for predicting the change of environmental parameters in the future period proposed by this solution;

[0060] Figure 5 It is a flowchart of the method for fusing and generating a hybrid prediction result proposed by this solution;

[0061] Figure 6 It is an architecture diagram of the electronic device in this solution;

[0062] Figure 7 It is a schematic diagram of the structure of the computer-readable storage medium in this solution.

[0063] The labels in the figure are:

[0064] 500 - Electronic device; 501 - Bus; 502 - CPU; 503 - ROM; 504 - RAM; 505 - Communication port; 506 - Input / output component; 507 - Hard disk; 508 - User interface; 600 - Computer-readable storage medium. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0066] Referring to Figure 1 as shown, a control method for an offshore wind farm includes:

[0067] Collecting real-time wind speed, significant wave height, and salt fog concentration data through a sensor array deployed on a wind turbine generator to form a three-dimensional environmental parameter vector;

[0068] Through the real-time collaborative monitoring of multi-dimensional environmental parameters, breaking through the limitations of traditional single-parameter perception, comprehensively integrating core environmental factors such as wind speed, waves, and salt fog, accurately characterizing the dynamic characteristics of complex sea conditions, providing multi-dimensional data support for subsequent state classification and risk assessment, and significantly reducing the risk of abnormal equipment loss caused by environmental factor coupling;

[0069] Inputting the real-time environmental parameters into a pre-trained classification model to match them to a plurality of preset environmental operating states, at least including a normal power generation mode, a storm warning mode, and an extreme shutdown mode;

[0070] Based on the multi-state dynamic matching mechanism of the intelligent classification model, realizing the rapid and accurate discrimination of environmental states, and avoiding the adaptability defects of manually setting static thresholds. By coupling multi-parameter boundary dynamic association rules, enhancing the recognition ability of extreme weather precursors, providing a key decision window for control strategy switching, and reducing passive responses under sudden working conditions;

[0071] Constructing a Markov state transition probability matrix based on historical operation data and recording the characteristics of environmental parameter change rates during each state transition;

[0072] By quantifying the historical state transition law and environmental parameter change characteristics, constructing a dynamic probability prediction framework, and strengthening the pre-judgment ability for the progressive environmental deterioration process. Combining the parameter change trend and state transition correlation analysis to support a more accurate maintenance cycle planning and spare part scheduling strategy, and optimizing the full life cycle management of equipment;

[0073] Based on the preset control parameters of the wind turbine generator for each environmental operating state, determining the adaptability of the preset control parameters of the wind turbine generator in each environmental operating state to other environmental operating states;

[0074] Through cross-state control parameter adaptability analysis, evaluating the compatibility of the current control strategy with future potential working conditions, and reducing mechanical shocks and power fluctuations during strategy switching. On the premise of ensuring core safety indicators, preferentially selecting an adaptation plan that takes into account both power generation efficiency and equipment stability to improve the overall reliability of the system;

[0075] Predicting the change of environmental parameters in the future period based on the characteristics of environmental parameter change rates during each state transition and fusing with the state transition probability matrix to generate a mixed prediction result;

[0076] Combining the dynamic change characteristics of the real-time environment with the long-term statistical laws, a hybrid prediction model is constructed to effectively overcome the blind spots of traditional prediction methods for sudden disturbances. By enhancing the time coverage and prediction robustness, it provides a forward-looking decision-making basis for active control such as unit yaw adjustment and power curve optimization, thereby improving the stability of the power generation process;

[0077] Based on the comprehensive judgment of the prediction results and the fitness evaluation, a control strategy of maintaining operation, parameter adjustment or emergency shutdown is triggered.

[0078] Through the multi-objective optimization decision-making mechanism, the safety and economic indicators are dynamically balanced to reduce unnecessary downtime caused by environmental misjudgment. The protection strategy is triggered first in the warning of extreme working conditions, and smooth parameter adjustment is achieved based on the fitness evaluation under normal working conditions, ultimately achieving the dual goals of power generation efficiency optimization and equipment risk prevention and control.

[0079] Reference Figure 2 As shown in Figure 2, the specific training process of the classification model is:

[0080] Set the environment parameter clustering center vector for each environment operation state respectively;

[0081] Based on the historical control log of the offshore wind farm, several environmental parameter vectors are determined when each environmental operation state is adjusted, and recorded as sample environmental parameter vectors;

[0082] Based on multiple sample environmental parameter vectors and environmental parameter cluster center vectors, training is performed to determine the environmental parameter boundary values ​​for each environmental operation state;

[0083] Based on the environmental parameter clustering center vector and the environmental parameter boundary value, a classification model of the environmental operation status is determined.

[0084] Through the collaborative training mechanism of cluster center vectors and historical sample data, an adaptive environmental state classification model is constructed to overcome the empirical dependence problem of traditional manually set classification boundaries. Based on the dynamic correction of parameter boundaries of cluster centers, the recognition accuracy of transition areas between multiple states such as normal, warning, and shutdown is improved, and the risk of misjudgment of complex coupling factors such as salt spray concentration mutations and wave resonance is effectively avoided. At the same time, through boundary learning driven by historical samples, the classification model's ability to characterize the nonlinear characteristics of the marine environment is enhanced, providing a high-confidence state discrimination basis for preventive maintenance and dynamic control strategy generation.

[0085] Reference Figure 3 As shown in the figure, a Markov state transition probability matrix is ​​constructed based on historical operation data, and the environmental parameter change rate characteristics during each state transition are recorded, including:

[0086] Based on the control attributes of the wind turbine generator set, an analysis duration is determined. The analysis duration is determined by the control delay of the wind turbine generator set. The maximum value of the duration from the issuance of the control command to the completion of the parameter adjustment by the wind turbine generator set is taken as the analysis duration;

[0087] Set a sample collection duration, and set the time period within the sample collection duration closest to the current moment as the sample collection period;

[0088] Based on the analysis duration, divide the sample collection period into several analysis periods;

[0089] Determine the environmental state of each analysis period based on the classification model;

[0090] Based on the environmental state of each analysis period, construct a Markov state transition probability matrix, P = [p ij K×K where p ij is the Markov state transition probability from the i-th environmental operating state to the j-th environmental operating state, and K is the total number of preset environmental operating state types;

[0091] where, N i is the total number of analysis periods in the i-th environmental state, and N i→j is the number of adjacent analysis periods in which the environmental state changes from the i-th environmental state to the j-th environmental operating state;

[0092] Statistically calculate the average value of the environmental parameters within each analysis period and the change rate of the environmental parameters over time within each analysis period, and summarize to obtain the environmental parameter change rate characteristics at the time of state transition.

[0093] By dynamically setting the analysis duration matching the control delay of the unit, ensure that the calculation of the state transition probability is strictly synchronized with the actual response characteristics of the unit, and improve the timing consistency between the prediction model and the execution action. Based on the division of the real-time sample collection period and the construction of the Markov state transition probability matrix, quantify the correlation characteristics between the environmental state migration law and the parameter change rate, and significantly enhance the prediction credibility for short-term sudden disturbances (such as sudden changes in wind speed). At the same time, through the statistical modeling of the environmental parameter change rate, accurately capture the dynamic evolution trend during the state transition process, provide key prior information for the pre-adjustment of the unit control parameters, effectively reduce the risk of equipment damage caused by state switching delays, and improve the global stability of the system under complex sea conditions.

[0094] Among them, determining the adaptability of the preset control parameters of the wind turbine generator set in each environmental operating state to other environmental operating states specifically includes:

[0095] ​Determine the performance indicators of the wind turbine under the preset control parameters of the wind turbine in each environmental operating state. The performance indicators at least include yaw delay, power fluctuation rate, and peak blade stress;

[0096] Then the calculation formula for the fitness is:

[0097]

[0098] In the formula, α k→h is the fitness of the preset control parameters of the wind turbine in the kth environmental operating state to the hth environmental operating state. m is the total number of performance indicators. is the nth performance indicator corresponding to the preset control parameters of the wind turbine in the kth environmental operating state. is the nth performance indicator corresponding to the preset control parameters of the wind turbine in the hth environmental operating state. w n is the emphasis weight of the nth performance indicator.

[0099] Through the cross-state fitness quantification evaluation of multi-dimensional performance indicators, construct a control parameter compatibility analysis framework to break through the limitations of traditional single-state parameter fixed configuration. Based on the weighted calculation model of performance differences, dynamically identify the shortcomings of the current control strategy under potential future working conditions. Such as the risk of stress exceeding the standard of extreme shutdown parameters in the warning state, and preferentially select an adaptation plan with the least mechanical impact and controllable power fluctuation. Flexibly adjust the priority of safety and economic indicators through the weight coefficient. For example, stress suppression is emphasized in the storm mode to realize the intelligent optimization of the control strategy at different environmental evolution stages and improve the robustness of the unit's full-condition operation.

[0100] Refer to Figure 4 As shown, predicting the future environmental parameter changes based on the environmental parameter change rate characteristics at each state transition specifically includes:

[0101] Based on the real-time parameter values of the environmental parameters collected in real time and the real-time change rate of the environmental parameters, and the average value of the environmental parameters and the change rate of the environmental parameters over time in each analysis period, solve the vector distance to obtain the real-time environment and the environmental similarity distance index in each analysis period;

[0102] Count the environmental operating states in the next analysis period of each analysis duration;

[0103] Based on the real-time environment and the environmental similarity distance index in all analysis periods and the environmental operating states in the next analysis period of each analysis period, calculate the transfer fitting probability of each environmental operating state in the future analysis duration;

[0104] Specifically, the calculation formula is:

[0105]

[0106] wherein, G i is the transition fitting probability of the i-th environmental operation state within the future analysis duration, F i is the set of analysis durations of the i-th environmental operation state within the next analysis period, c is the c-th element in F i E is the set of all analysis durations, o is the o-th element in E, L max is the maximum value among the environmental similarity distance metrics between the real-time environment and all analysis durations, L min is the minimum value among the environmental similarity distance metrics between the real-time environment and all analysis durations, L c is the environmental similarity distance metric between the real-time environment and c.

[0107] Through the dynamic similarity matching mechanism of real-time environmental parameters and historical periods, a data-driven state transition fitting probability calculation model is constructed, effectively solving the problem of insufficient sensitivity of traditional static prediction methods to real-time changes. The similarity between the real-time environment and historical scenarios is quantified using vector distance, and combined with the statistical law of the next state, the influence trend of sudden disturbances on state migration is accurately captured. By normalizing the similarity distance weight allocation, highly correlated historical samples are dynamically screened, enhancing the adaptability of the prediction results to non-linear environmental evolution, providing a highly reliable basis for the unit to switch control strategies in advance, and significantly reducing the risks of power oscillation or protection lag caused by prediction deviation.

[0108] Referring to Figure 5 as shown, the specific steps for generating a hybrid prediction result by fusing with the state transition probability matrix are as follows:

[0109] Determine the environmental operation state at the current moment. Based on the environmental operation state at the current moment, determine the Markov state transition probability of transitioning to each environmental operation state next from the Markov state transition probability matrix;

[0110] Respectively set the hybrid weights of the transition fitting probability and the Markov state transition probability, and perform weighted summation on the transition fitting probability and the Markov state transition probability to obtain the hybrid prediction result.

[0111] Specifically, in some embodiments, the transition fitting probability and the Markov state transition probability are respectively taken as 0.7 and 0.3, then the hybrid prediction result is 0.7×transition fitting probability + 0.3×Markov state transition probability;

[0112] By integrating the long-term statistical laws of the Markov model and the short-term dynamic characteristics of real-time fitting probability, a hybrid prediction mechanism is constructed to effectively balance the dual impacts of historical experience and real-time environmental evolution. Based on the dynamic weight allocation strategy, the decision-making weight of real-time fitting probability is strengthened to capture rapid changes in the scenario of sudden disturbances, and the long-term trend prediction of Markov probability is emphasized under steady-state working conditions, enhancing the adaptability of prediction results to complex environmental patterns. Through the complementary optimization of the dual models, the risks of missed reports in extreme working conditions and false alarms in steady state are significantly reduced, providing a prediction basis that takes into account both timeliness and stability for the unit control strategy, and realizing the upgrade of the decision-making mode from passive response to active defense.

[0113] Specifically, according to the comprehensive judgment of the prediction results and fitness evaluation, triggering one of the control strategies of maintaining operation, parameter adjustment, or emergency shutdown specifically includes:

[0114] If the prediction probability of the extreme shutdown mode in the hybrid prediction results exceeds 0.5, an emergency shutdown is triggered;

[0115] Otherwise, taking the prediction probability of each environmental operating state in the hybrid prediction results as the weight, the fitness of the preset control parameters of the wind turbine generator under each environmental operating state to the environmental operating state is weighted and summed to obtain the predicted fitness of the preset control parameters of the wind turbine generator under each environmental operating state to the environment;

[0116] If the environmental operating state corresponding to the maximum predicted fitness is the same as the current environmental operating state, maintaining operation is triggered;

[0117] If the environmental operating state corresponding to the maximum predicted fitness is different from the current environmental operating state, parameter adjustment is triggered, and the control parameters of the offshore wind farm are adjusted to the preset control parameters of the wind turbine generator under the environmental operating state corresponding to the maximum predicted fitness.

[0118] Specifically, the calculation formula of the predicted fitness is:

[0119]

[0120] where S i is the predicted fitness of the preset control parameters of the wind turbine generator under the i-th environmental operating state to the environment, Y k is the prediction probability of the k-th environmental operating state in the hybrid prediction results, α i→k is the fitness of the preset control parameters of the wind turbine generator under the i-th environmental operating state to the k-th environmental operating state, and K is the total number of preset environmental operating states.

[0121] By integrating prediction probability and fitness evaluation through a multi-level decision-making mechanism, a control strategy generation framework for dynamic balance between safety and economy is constructed. Based on the probability threshold of extreme working conditions, emergency shutdown is triggered to ensure the rigid constraints of the core safety indicators of the equipment; in normal scenarios, through the weighted fusion calculation of predicted fitness, comprehensively considering the future multi-state evolution trend and the compatibility of control parameters, the globally optimal solution is preferentially selected. This mechanism can not only avoid the over-conservative or risky decisions caused by a single prediction model, but also avoid the lag of static fitness evaluation in responding to the dynamic environment, and achieve the coordinated improvement of power generation efficiency optimization and risk prevention and control. At the same time, through the probabilistic mapping of fitness weights, the robustness of the control strategy to uncertain environmental evolution is enhanced, and the energy loss and mechanical wear caused by unnecessary mode switching are reduced.

[0122] Furthermore, based on the same inventive concept as the above method, this solution also proposes an offshore wind farm control system, including:

[0123] A sensor array, deployed on the wind turbine generator, for real-time collection of wind speed, significant wave height, and salt fog concentration data;

[0124] A data processing module, for integrating the collected real-time data into a three-dimensional environmental parameter vector;

[0125] A state classification module, with a pre-trained classification model built-in, for matching the three-dimensional environmental parameter vector to a plurality of preset environmental operating states, including normal power generation mode, storm warning mode, and extreme shutdown mode;

[0126] A Markov analysis module, for constructing a state transition probability matrix based on historical operation data and storing the environmental parameter change rate characteristics during each state transition;

[0127] A fitness evaluation module, for calculating the control fitness of each environmental operating state to other states according to the preset control parameters;

[0128] An environmental prediction module, for generating a hybrid prediction result by combining the real-time environmental parameter change rate and the state transition probability matrix;

[0129] A control strategy execution module, for triggering one of the control instructions of maintaining operation, parameter adjustment, or emergency shutdown according to the hybrid prediction result and the fitness evaluation result.

[0130] Furthermore, the method according to the embodiment of the present application can also be implemented by means of Figure 6 the architecture of the electronic device shown. As Figure 6As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the offshore wind farm control method provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 6 the architecture shown is only exemplary. When implementing different devices, one or more components in the electronic device shown may be omitted according to actual needs. Figure 6

[0131] Figure 7 It is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of this application. As Figure 7 shown, it is a computer-readable storage medium 600 according to an embodiment of this application. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are run by a processor, the offshore wind farm control method according to the embodiment of this application described with reference to the above drawings can be executed. The storage medium 600 includes but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0132] In summary, the advantages of the present invention are as follows: By integrating real-time perception of multi-source environmental parameters and intelligent state classification, combining dynamic Markov state transition prediction and cross-state adaptability evaluation, the present invention significantly improves the adaptability of offshore wind farms to complex sea conditions. Its multi-dimensional environmental parameter collaborative analysis and hybrid prediction mechanism effectively improves the early warning accuracy of extreme working conditions, avoiding misjudgment or response lag caused by traditional static thresholds; by predicting the environmental evolution trend and the adaptability of cross-state control parameters, dynamically optimizing the operation strategy of the generator set, reducing unnecessary shutdown times and frequent parameter adjustments, while ensuring equipment safety, improving power generation efficiency. In addition, the fusion prediction model based on historical data and real-time change rate enhances the robustness of the system to sudden environmental disturbances, realizing reduced power generation losses and decreased equipment maintenance costs.

[0133] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.​

Claims

1. A control method for an offshore wind farm, characterized in that Including: Collect real-time wind speed, significant wave height, and salt fog concentration data through a sensor array deployed on a wind turbine generator to form a three-dimensional environmental parameter vector; Input the real-time environmental parameters into a pre-trained classification model to match to a plurality of preset environmental operating states, at least including a normal power generation mode, a storm warning mode, and an extreme shutdown mode; Construct a Markov state transition probability matrix based on historical operation data and record the environmental parameter change rate characteristics when each state transitions; Based on the preset control parameters of the wind turbine generator for each environmental operating state, determine the adaptability of the preset control parameters of the wind turbine generator in each environmental operating state to other environmental operating states; Predict the environmental parameter changes in the future period based on the environmental parameter change rate characteristics when each state transitions and fuse with the state transition probability matrix to generate a mixed prediction result; According to the comprehensive judgment of the prediction result and the adaptability evaluation, trigger one of the control strategies of maintaining operation, parameter adjustment, or emergency shutdown.

2. The control method for an offshore wind farm according to claim 1, characterized in that, The specific training process of the classification model is as follows: Set an environmental parameter clustering center vector for each environmental operating state respectively; Based on the historical control logs of the offshore wind farm, determine several environmental parameter vectors when each environmental operating state is adjusted, denoted as sample environmental parameter vectors; Train based on multiple sample environmental parameter vectors and environmental parameter clustering center vectors to determine the environmental parameter boundary values for each environmental operating state; Based on the environmental parameter clustering center vector and environmental parameter boundary values, determine the classification model of the environmental operating state.

3. A method for controlling an offshore wind farm according to claim 2, characterized in that, The construction of the Markov state transition probability matrix based on historical operation data and recording the environmental parameter change rate characteristics when each state transitions specifically includes: Determine an analysis duration based on the control attributes of the wind turbine generator; Set a sample collection duration, and set the time period within the sample collection duration closest to the current moment as the sample collection period; Based on the analysis duration, divide the sample collection period into several analysis periods; Determine the environmental state of each analysis period based on the classification model; Construct a Markov state transition probability matrix based on the environmental state of each analysis period. p ij is the Markov state transition probability from the i-th environmental operating state to the j-th environmental operating state, and K is the total number of preset environmental operating states. Among them, N i is the total number of analysis periods in the i-th environmental state, and N i→j is the number of adjacent analysis periods when the environmental state changes from the i-th environmental state to the j-th environmental operating state; Statistically calculate the average value of the environmental parameters within each analysis period and the change rate of the environmental parameters over time within each analysis period, and summarize to obtain the environmental parameter change rate characteristics when the state transitions.

4. A method for controlling an offshore wind farm according to claim 3, characterized in that The determination of the adaptability of the preset control parameters of the wind turbine generator in each environmental operating state to other environmental operating states specifically includes: Determine the performance indicators of the wind turbine generator under the preset control parameters of the wind turbine generator in each environmental operating state, and the performance indicators at least include yaw delay, power fluctuation rate, and peak blade stress; Then the calculation formula for the adaptability is: where α k→h is the adaptability of the preset control parameters of the wind turbine in the k-th environmental operating state to the h-th environmental operating state, m is the total number of performance indicators, is the n-th performance indicator corresponding to the preset control parameters of the wind turbine in the k-th environmental operating state, is the n-th performance indicator corresponding to the preset control parameters of the wind turbine in the h-th environmental operating state, w n is the emphasis weight of the n-th performance indicator.

5. A method for controlling an offshore wind farm according to claim 4, characterized in that, The prediction of the environmental parameter changes in the future period based on the environmental parameter change rate characteristics when each state transitions specifically includes: Based on the real-time parameter values of the environmental parameters collected in real-time and the real-time change rate of the environmental parameters, solve the vector distance with the average value of the environmental parameters within each analysis period and the change rate of the environmental parameters over time to obtain the environmental similarity distance index between the real-time environment and each analysis period; Statistically calculate the environmental operating state in the next analysis period for each analysis duration; Calculate the transfer fitting probability of each environmental operating state within the future analysis duration based on the environmental similarity distance metric in the real-time environment and all analysis durations, as well as the environmental operating state in the next analysis duration of each analysis duration. Specifically, the calculation formula is: Among them, G i is the transfer fitting probability of the i-th environmental operation state within the future analysis duration, F i is the analysis duration set of the i-th environmental operation state within the next analysis period, c is the c-th element in F i , E is the set of all analysis durations, o is the o-th element in E, L max is the maximum value among the environmental similarity distance metrics between the real-time environment and the environments within all analysis durations, L min is the minimum value among the environmental similarity distance metrics between the real-time environment and the environments within all analysis durations, L c is the environmental similarity distance metric between the real-time environment and c.

6. The control method for an offshore wind farm according to claim 5, characterized in that, The specific steps for fusing with the state transition probability matrix to generate a hybrid prediction result are as follows: Determine the environmental operating state at the current moment. Based on the environmental operating state at the current moment, determine the Markov state transition probability of transitioning to each environmental operating state next from the Markov state transition probability matrix. Set the hybrid weights of the transfer fitting probability and the Markov state transition probability respectively, and perform weighted summation of the transfer fitting probability and the Markov state transition probability to obtain a hybrid prediction result.

7. A method for controlling an offshore wind farm according to claim 6, characterized in that, The specific content of triggering one of the control strategies of maintaining operation, parameter adjustment, or emergency shutdown according to the comprehensive judgment of the prediction result and the fitness evaluation includes: If the prediction probability of the extreme shutdown mode in the hybrid prediction result exceeds 0.5, trigger an emergency shutdown. Otherwise, use the prediction probability of each environmental operating state in the hybrid prediction result as the weight to perform weighted summation of the fitness of the preset control parameters of the wind turbine generator under each environmental operating state with the environmental operating state, to obtain the predicted fitness of the preset control parameters of the wind turbine generator under each environmental operating state with the environment. If the environmental operating state corresponding to the maximum predicted fitness is the same as the current environmental operating state, trigger maintaining operation. If the environmental operating state corresponding to the maximum predicted fitness is different from the current environmental operating state, trigger parameter adjustment, and adjust the control parameters of the offshore wind farm to the preset control parameters of the wind turbine generator under the environmental operating state corresponding to the maximum predicted fitness.

8. An offshore wind farm control system for implementing the offshore wind farm control method described in any one of claims 1-7, comprising: A sensor array, a sensor array deployed on the wind turbine generator for real-time collection of wind speed, significant wave height, and salt fog concentration data. A data processing module for integrating the collected real-time data into a three-dimensional environmental parameter vector. A state classification module with a pre-trained classification model for matching the three-dimensional environmental parameter vector to a plurality of preset environmental operating states, including a normal power generation mode, a storm warning mode, and an extreme shutdown mode. A Markov analysis module for constructing a state transition probability matrix based on historical operation data and storing the environmental parameter change rate characteristics during each state transition. A fitness evaluation module for calculating the control fitness of each environmental operating state with respect to other states according to the preset control parameters under each environmental operating state. An environmental prediction module for generating a hybrid prediction result by combining the real-time environmental parameter change rate and the state transition probability matrix. A control strategy execution module for triggering one of the control instructions of maintaining operation, parameter adjustment, or emergency shutdown according to the hybrid prediction result and the fitness evaluation result.

9. An electronic device, characterized in that, Comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the offshore wind farm control method according to any one of claims 1-8.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by a processor, the offshore wind farm control method according to any one of claims 1-8 is implemented.

Images