Wind turbine generator intelligent start-stop decision-making method based on security domain perception

Abstract

The invention discloses a wind turbine generator intelligent start-stop decision-making method based on security domain perception. The method comprises the following steps: constructing a multi-state operation portrait table based on multi-source heterogeneous data of a wind turbine generator, dynamically calculating a security domain in each operation state by adopting a kernel density estimation algorithm, and constructing a security domain-instruction three-dimensional contrast matrix; generating a first-layer triggering judgment signal driven by wind-light fluctuation based on short-term wind-light power prediction and a security domain, generating a second-layer triggering judgment signal driven by power grid demand based on a power grid dispatching instruction, and performing priority arbitration by taking a security domain-instruction three-dimensional contrast matrix as a constraint when the two layers of signals conflict; and finally, a technical means of generating a start-stop strategy by fusing double-layer trigger signals is adopted, so that the requirements of safe operation of the wind turbine generator and power grid dispatching response are considered, the start-stop decision precision and safety are improved, decision one-sidedness caused by single data and misjudgment and missing judgment caused by fixed safety domain are avoided, and the safety of the wind turbine generator is improved. And equipment damage or power grid instability caused by signal conflicts can be avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wind power generation control, and particularly relates to a wind turbine intelligent start-stop decision method based on safety domain perception. BACKGROUND

[0002] In the field of new energy power generation, wind power generation, as an important clean energy utilization form, has a continuously expanding installed capacity and grid-connected scale. The start-stop decision of a wind turbine, as a core link of the operation control of a wind power generation system, directly affects the power generation efficiency, equipment life and power grid stability. However, the current wind turbine start-stop decision method has many deficiencies.

[0003] Traditional methods mostly formulate start-stop strategies based on a single operating state, and do not fully consider the complex working conditions faced by wind turbines in actual operation, resulting in a large deviation between the operation image table and the actual working condition, which easily causes misoperation. At the same time, in the existing decision process, the safety domain calculation mostly adopts a static threshold setting method, which cannot be dynamically adjusted according to the historical operation data of the wind turbine, real-time meteorological information and equipment health state, and in the extreme weather conditions or equipment aging scenarios, the safety domain is easily broken through, causing the risk of equipment damage.

[0004] Taking patent CN104037817A as an example, the patent proposes a wind turbine automatic start-stop control method of a wind farm with the smallest hour number deviation, which, according to the principle of fair scheduling, does not use the start-stop method when the continuously adjusted grid-connected wind turbine active power can meet the control requirements. However, this method has obvious deficiencies. Firstly, in terms of operating state consideration, it only makes start-stop decisions according to the utilization hours, a single dimension, without fully covering various complex working conditions in the actual operation of the wind turbine, so that the capture of the real-time operating state of the wind turbine in the decision process is not accurate enough to cope with the changing actual scenarios. Secondly, the safety domain setting has limitations. This method does not involve a mechanism for dynamically adjusting the safety domain according to real-time meteorological information and equipment health state, but uses a relatively fixed decision logic. In actual operation, when encountering sudden extreme weather conditions, this static safety domain setting method is easy to make the wind turbine break through the safety operation range, causing equipment failure or even safety accidents. SUMMARY

[0005] In view of the deficiencies of the prior art, the application provides a wind turbine intelligent start-stop decision method based on safety domain perception, which comprises the following steps: constructing a multi-state operation portrait table based on multi-source heterogeneous data of a wind turbine, dynamically calculating a safety domain in each operation state by using a kernel density estimation algorithm, constructing a safety domain-instruction three-dimensional comparison matrix, generating a first layer trigger judgment signal driven by wind and light fluctuations and a second layer trigger judgment signal driven by a power grid instruction, arbitrating and fusing the signals by the safety domain-instruction three-dimensional comparison matrix when there is a conflict, and generating a start-stop strategy.

[0006] The technical solution of the application is as follows:

[0007] According to an aspect of the application, a wind turbine intelligent start-stop decision method based on safety domain perception is provided, which comprises the following steps:

[0008] Based on multi-source heterogeneous data of a wind turbine, a multi-state operation portrait table is constructed, a safety domain in each state of the multi-state operation portrait table is dynamically calculated by using a kernel density estimation algorithm, and a safety domain-instruction three-dimensional comparison matrix is formed by comprehensively considering the safety domain and a power grid dispatching instruction.

[0009] Based on the safety domain-instruction three-dimensional comparison matrix, a double-layer trigger judgment signal of wind and light fluctuations and power grid demand is generated, priority arbitration is performed by using the safety domain-instruction three-dimensional comparison matrix as a constraint when the first layer trigger judgment signal and the second layer trigger judgment signal conflict, and a start-stop strategy is generated by fusing the double-layer trigger judgment signal.

[0010] The first layer trigger judgment signal is based on a short-term wind and light power prediction result and a safety domain, and the second layer trigger judgment signal is based on a response to a power grid dispatching instruction.

[0011] As a further selection of the method of the application, the multi-source heterogeneous data comprises historical operation data, real-time meteorological information and equipment health state data of the wind turbine, wherein the historical operation data covers SCADA data of power output, wind speed, generator speed and gear box temperature, the real-time meteorological information covers wind speed, wind direction, environmental temperature and air pressure of a wind measurement tower or meteorological forecast data, and the equipment health state data covers state monitoring data of gear box vibration, bearing temperature and hydraulic system pressure.

[0012] As a further selection of the method of the application, the multi-state operation portrait table divides the operation state of the wind turbine into a rated operation state, a sub-rated operation state, a standby state and a shutdown state.

[0013] The rated operating state is defined as the wind speed being between the rated wind speed and the cut-out wind speed and the power output fluctuation range being within ±5% of the rated power.

[0014] The sub-rated operating state is defined as a wind speed lower than the rated wind speed and a power output lower than 95% of the rated power but higher than the minimum power generation power corresponding to the cut-in wind speed.

[0015] The standby state is defined as the wind speed being lower than the cut-in wind speed, the power grid dispatching command requiring the suspension of operation, or the equipment being in maintenance mode.

[0016] The shutdown state is defined as when the equipment health monitoring data exceeds the safety threshold, the wind speed exceeds the safety range, or the power grid commands a forced shutdown.

[0017] As a further option of the method in this application, the multi-state operation profile table includes:

[0018] The key coupling variables for rated operating conditions include wind speed, power, generator speed, and gearbox vibration acceleration;

[0019] The key coupled variables for sub-rated operation include wind speed, pitch angle, power, and blade root bending moment.

[0020] The key set of coupled variables in standby mode includes grid frequency, voltage, device standby time, and ambient temperature.

[0021] The key set of coupled variables in the shutdown state includes extreme wind speed, peak vibration, fault codes, and safety chain status.

[0022] As a further option of the method in this application, the step of dynamically calculating the security domain using a kernel density estimation algorithm includes: applying a multivariate kernel density estimation formula to each operating state:

[0023]

[0024] in, This represents the state distribution probability density estimate of the input variable x, where x represents the input variable. i Let represent the variable value of the i-th data point, K(·) represent the kernel function, h is the bandwidth parameter used to control the smoothness of the estimation, and n represents the number of variables.

[0025] As a further option of the method in this application, the security domain is determined by setting a probability threshold α:

[0026]

[0027] in, It is the probability density of the state distribution obtained by kernel density estimation, and α is a preset probability threshold;

[0028] like Then x is considered a value within the security region; if Then x is considered to be outside the security domain.

[0029] As a further option of the method in this application, the mathematical expression of the security domain-command three-dimensional comparison matrix is ​​a three-dimensional tensor structure. Its operating state dimension includes security domain parameters for rated operating state, sub-rated operating state, standby state, and shutdown state. The security domain parameter dimension includes thresholds for wind speed, power output, generator speed, and gearbox temperature. The power grid dispatch command parameter dimension includes target power value, start / stop control signal, emergency shutdown signal, and response time constraint. Furthermore, the matrix maps the feasibility boundary of state switching through the security domain constraint dimension. The mathematical expression of the security domain-command three-dimensional comparison matrix is ​​as follows:

[0030]

[0031] Where: M: Security domain-instruction three-dimensional comparison matrix; under rated operating conditions, v rated Rated wind speed; P rated Rated power; ω rated Rated generator speed; T gearbox : Gearbox temperature threshold; under sub-rated operating conditions, v sub Sub-rated wind speed; P sub Subrated power; ω sub Sub-rated generator speed; in standby mode, v cut-in Cut-in wind speed, P standby Standby power, ω standby Standby speed; in the off state, v cut-out Cut off the wind speed, P off Shutdown power, ω off Stop speed;

[0032] Grid dispatch command dimension, P target Target power value, S∈{0,1}; Start-stop control signal, S=1: Grid dispatch requires unit to start; S=0: Grid dispatch requires unit to stop; E∈{0,1}: Emergency stop signal, E=1: Grid fault or extreme weather conditions require emergency stop of unit; E=0: Unit is operating normally, T response Response time constraints;

[0033] The safety domain constraint dimension is as follows: safety domain A is the safety domain of the unit under rated operating conditions; safety domain B is the safety domain of the unit under sub-rated operating conditions; safety domain C is the safety domain of the unit under standby conditions; and safety domain D is the safety constraint of the unit under shutdown conditions.

[0034] As a further option of the method in this application, the step of determining the first-layer triggering signal based on the short-term wind and solar power prediction results and the safety domain includes:

[0035] The short-term wind and solar power is predicted using a predictive model, expressed as follows:

[0036]

[0037] in, To predict power, W history For historical wind speed series, M weather For meteorological data, F turbulence G represents the intensity characteristics of turbulence. solar For the contribution of photovoltaic power, α is the model weighting coefficient (taken as 0.3-0.7, f LSTM f is the LSTM model function. RF This is the function for the random forest model;

[0038] Short-term forecast results Real-time comparison is performed with the corresponding security domain in the security domain-instruction three-dimensional comparison matrix M; the comparison process is as follows:

[0039] Calculate the matching degree between the power change rate within the future time window and the allowable fluctuation range of the safety domain; if the power change rate exceeds the threshold, trigger the start / stop judgment.

[0040] The relationship between predicted power and safety domain parameters is obtained from the power curve of the wind turbine, and the values ​​of safety domain parameters are obtained.

[0041] Determine whether the security domain parameter values ​​conform to the security domain and generate a trigger signal for switching the first-level operating state;

[0042] The process for determining the second-level trigger signal based on the response to power grid dispatch instructions is as follows:

[0043] Receive and identify parameters such as target power value, start / stop signal, emergency stop signal and response time constraint in the scheduling command;

[0044] The execution priority is set according to the command type. The command priority is divided into the following categories according to the power grid security level from high to low: emergency shutdown > state switching command > normal power adjustment.

[0045] Based on the priority order, the scheduling instruction parameters are matched with the security domains in the security domain-instruction three-dimensional comparison matrix to determine whether the instruction meets the safe operation conditions of the unit.

[0046] Based on the matching result between the instruction and the security domain, a trigger signal for switching the second-level running state is generated.

[0047] As a further option of the method in this application, when the first-level trigger judgment signal and the second-level trigger judgment signal conflict, the priority arbitration is performed based on the security domain-instruction three-dimensional comparison matrix: if the second-level trigger judgment signal contains an emergency stop signal, then the second-level trigger judgment signal is forcibly executed first; if the first-level trigger judgment signal indicates that the critical coupling variable exceeds the security domain, then the first-level trigger judgment signal is executed first; for regular instruction conflicts, the target state with the lowest risk and the smallest fluctuation with the current state is selected based on the overlapping area of ​​the security domain, and the arbitrated target state code is output, including the target state, expected duration, power setting value and security domain verification flag.

[0048] As a further option of the method in this application, the generation of the start-stop strategy includes the following steps:

[0049] The results of the first and second layer trigger judgment signals are combined. If there is a conflict, arbitration is performed based on the security domain-instruction three-dimensional comparison matrix.

[0050] The start / stop policy is output in the form of a status code, including the target status, expected duration, power setting value, and safety domain verification flag;

[0051] The start-stop strategy achieves coordinated optimization of wind and solar power fluctuations and grid demand by dynamically adjusting the safety domain constraint parameters.

[0052] The beneficial effects of this application are as follows:

[0053] This application constructs a multi-state operation profile table by integrating historical SCADA data of wind turbine operation, real-time meteorological information, and equipment health status data. The operation status is refined into four categories: rated, sub-rated, standby, and shutdown, and the switching rules are clearly defined. This effectively solves the problems of single data dimension and one-sided decision-making basis in existing technologies, and enables start-up and shutdown decisions to comprehensively cover three dimensions of information: unit operation, environmental changes, and equipment health, greatly improving the completeness and accuracy of decision-making basis.

[0054] This application uses a kernel density estimation algorithm to dynamically calculate the safety domain under various operating conditions. It determines the safety range by modeling the joint probability density of multiple variables and setting a probability threshold, replacing the traditional fixed threshold mode. This can accurately adapt to the parameter distribution characteristics of different operating conditions and avoid the problems of "false shutdown" or "missed judgment risk" caused by the inability of fixed thresholds to match state differences. It significantly improves the adaptability and accuracy of unit safety judgment and provides a more reliable safety guarantee for the stable operation of the unit.

[0055] The dual-layer triggering judgment mechanism and conflict arbitration rules for "wind and solar power fluctuations - grid commands" designed in this application identify wind and solar power fluctuation risks through short-term wind and solar power prediction, and respond to grid demands through grid command parsing. In the event of a signal conflict, a three-dimensional comparison matrix of safety domain and command is used as a constraint, and arbitration is based on the principle of safety priority and the highest priority of emergency commands. This takes into account both unit safety and grid dispatching needs, avoids the risk of equipment damage due to prioritizing grid commands and exceeding the safety domain, and also prevents the problem of delaying grid response by only focusing on unit safety. At the same time, it reduces power fluctuations during state transitions, taking into account both safety and grid compatibility. Attached Figure Description

[0056] Figure 1 A schematic diagram of the overall process of the intelligent start-stop decision-making method for wind turbines based on safety domain awareness;

[0057] Figure 2 A detailed flowchart of the S100 steps of the intelligent start-stop decision-making method for wind turbines based on safety domain awareness;

[0058] Figure 3 The detailed flowchart of the S200 steps of the intelligent start-stop decision-making method for wind turbines based on safety domain awareness is shown below. Detailed Implementation

[0059] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0060] Example 1

[0061] In the field of new energy power generation, the installed capacity and grid-connected scale of wind turbines continue to expand, and their start-up and shutdown decisions are crucial to power generation efficiency, equipment lifespan, and grid stability. However, existing methods are mostly based on single-state strategy formulation, with statically set safety domains, making it difficult to adapt to complex operating conditions. Furthermore, they fail to effectively coordinate wind and solar power fluctuations with grid demand, resulting in a singular optimization objective that struggles to balance safety, accuracy, and economy, thus requiring urgent improvement. Please refer to [link / reference]. Figure 1 This application illustrates an embodiment of a smart start-stop decision-making method for wind turbines based on safety domain awareness, the method comprising:

[0062] S100: Based on multi-source heterogeneous data of wind turbine units, construct a multi-state operation profile table; dynamically calculate the safety domain of each state in the multi-state operation profile table through kernel density estimation algorithm, and form a three-dimensional comparison matrix of safety domain and power grid dispatch instructions by integrating the safety domain and power grid dispatch instructions.

[0063] S200: Based on the safety domain-command three-dimensional comparison matrix, a two-layer trigger judgment signal for wind and solar fluctuations and grid demand is generated. When the first-layer trigger judgment signal and the second-layer trigger judgment signal conflict, priority arbitration is performed with the safety domain-command three-dimensional comparison matrix as a constraint. Start-up and shutdown strategies are generated by integrating the two-layer trigger judgment signals.

[0064] Among them, the first-level trigger judgment signal is based on the short-term wind and solar power prediction results and the safety domain, and the second-level trigger judgment signal is based on the response to the grid dispatch command.

[0065] Based on multi-source heterogeneous data from wind turbine generators, S100 constructs an operation profile table containing four states: rated, sub-rated, standby, and shutdown, after preprocessing. Kernel density estimation is used to calculate the safety domain of each state, and then combined with grid dispatch instructions to form a three-dimensional comparison matrix of safety domain and instruction.

[0066] Please refer to Figure 2 The diagram illustrates a flowchart of an example of a safety domain-aware intelligent start-stop decision method S100 for wind turbines, which includes:

[0067] S110: Construct a multi-state operation profile table based on multi-source heterogeneous data of wind turbine units.

[0068] The construction of a multi-state operational profile table relies on multi-source heterogeneous data.

[0069] In one possible implementation, the multi-source heterogeneous data encompasses historical operating data of the wind turbine, real-time meteorological information, and equipment health status.

[0070] Specifically, historical operating data includes SCADA data such as power output, wind speed, generator speed, and gearbox temperature; real-time meteorological information includes wind speed, wind direction, ambient temperature, and air pressure data from wind towers or weather forecasts; and equipment health status data includes status monitoring data such as gearbox vibration, bearing temperature, and hydraulic system pressure.

[0071] Data quality can be improved by preprocessing multi-source heterogeneous data, such as data cleaning, missing value imputation, outlier detection, and normalization.

[0072] When constructing the multi-state operation profile table, the collected historical operation data, real-time meteorological information and equipment health status data are mapped to different operation states, and the operation mode switching rules are integrated to form the multi-state operation profile table.

[0073] In one possible implementation, the operating states of wind turbine units are divided into four categories: rated operating state, sub-rated operating state, standby state, and shutdown state. The rules for classifying the states are as follows:

[0074] Rated operating conditions: When the wind speed is at the rated wind speed vrated With cut-out wind speed v cut-out During this period, the unit operates at rated power, and the generator speed is stable. This state is determined by the matching relationship between wind speed and power output, i.e., the power output is close to the rated power P. rated And the fluctuation range is within ±5%.

[0075] Sub-rated operating condition: When the wind speed is lower than the rated wind speed v < v rated At this time, the unit employs a maximum power point tracking strategy, and the power output fluctuates with wind speed. This state is defined as power output below 95% of rated power and above the cut-in wind speed v. cut-out The corresponding minimum power generation capacity.

[0076] Standby mode: When the wind speed is lower than the cut-in wind speed v < v cut-in When the power grid dispatch command requires the unit to suspend operation or the equipment to be under maintenance, the unit stops generating electricity but maintains its ability to start up quickly. The triggering conditions for this state include wind speed falling below a set threshold, activation of the power grid command signal, or the equipment being in normal health but not meeting the conditions for power generation.

[0077] Shutdown Status: When equipment health status is abnormal, extreme weather conditions occur, or the power grid fails, the unit will automatically shut down to ensure safety. This status is defined as equipment health monitoring data exceeding safety thresholds, wind speed exceeding the safety range, or a power grid command forcing a shutdown.

[0078] In one possible implementation, the switching rules for the operating modes of rated operating state, sub-rated operating state, standby state, and shutdown state are as follows:

[0079] Rated → Sub-rated: The wind speed decreases to below the rated wind speed but not below the cut-in wind speed, i.e.: v < v rated And v≥v cut-in ;

[0080] Sub-rated → Standby: The wind speed drops below the cut-in wind speed, i.e.: v < v cut-in Or, the power grid may instruct the system to shut down.

[0081] Standby → Rated / Sub-rated: The wind speed recovers to a level not lower than the cut-in wind speed, i.e.: v ≥ v cut-in And the equipment is in normal health condition;

[0082] Any state → shutdown: The equipment is in an abnormal health state, or the wind speed exceeds the cut-in wind speed, i.e., v > v cut-out .

[0083] In one possible implementation, the multi-state operation profile table is shown in Table 1 below:

[0084] Table 1 Multi-state operation profile

[0085] Running state Key coupling variable set Rated operation Wind speed, power, generator speed, gearbox vibration acceleration Sub-rated operation Wind speed, pitch angle, power, blade root bending moment Standby state Grid frequency, voltage, device standby duration, ambient temperature Shutdown state Extreme wind speed, vibration peak, fault code, safety chain status

[0086] S120: Calculate the security domain of each state in the multi-state operation profile table based on the kernel density estimation algorithm.

[0087] The calculation of the safety domain is crucial to ensuring the stable operation of the unit under different operating conditions. Based on the multi-state operating profile, the safety domain of each operating condition is dynamically calculated using the kernel density estimation algorithm.

[0088] Kernel density estimation (KDE) is a probability density function used to estimate data. In this step, kernel density estimation is used to model the state distribution of wind turbines under different operating conditions and to calculate the safety region accordingly.

[0089] In the context of multi-state operation profiling, kernel density estimation is performed for each operation state to obtain the state distribution function under different states. In one possible implementation, the formula for calculating the state distribution function is:

[0090]

[0091] in, This represents the state distribution probability density estimate of the input variable x, where x represents the input variable. i Let represent the variable value of the i-th data point, K(·) represent the kernel function, h is the bandwidth parameter used to control the smoothness of the estimation, and n represents the number of variables.

[0092] By using multivariate kernel density estimation, the joint probability density function of wind turbines under the same operating conditions is obtained, and the correlation between different variables is analyzed accordingly.

[0093] When calculating the safety domain, the probability density function obtained by kernel density estimation is combined to determine the safety threshold of the unit under different operating conditions.

[0094] In one possible implementation, the security domain is determined by setting a probability threshold α:

[0095]

[0096] in, It is the state distribution probability density obtained by kernel density estimation, and α is the preset probability threshold.

[0097] In specific implementation, if Then x is considered a value within the security region; if Then x is considered to be outside the security domain.

[0098] S130: Integrate the security domain and power grid dispatch instructions to form a three-dimensional comparison matrix of security domain and instructions.

[0099] The safety domain-command three-dimensional mapping matrix not only reflects the safety domain under different operating conditions, but also considers the impact of grid dispatch commands on unit operation, thus providing a comprehensive decision-making basis for subsequent intelligent start-up and shutdown decisions. The safety domain-command three-dimensional mapping matrix maps the safety domain of wind turbine units to grid dispatch commands to form a multi-dimensional safety domain decision space.

[0100] In one possible implementation, the safety domain-command three-dimensional mapping matrix involves multiple dimensions, including operating status, safety domain parameters, and grid dispatch command parameters. The operating status dimension includes rated operating status, sub-rated operating status, standby status, and shutdown status. The safety domain parameter dimension includes the safety domains of key variables such as wind speed, power output, generator speed, and gearbox temperature. The grid dispatch command parameter dimension includes target power values, start / stop control signals, emergency shutdown signals, and response time constraints.

[0101] In one possible implementation, the mathematical expression of the security domain-instruction stereo mapping matrix is ​​as follows:

[0102]

[0103] Where: M: Security domain-instruction three-dimensional comparison matrix; under rated operating conditions, v rated Rated wind speed; P rated Rated power; ω rated Rated generator speed; T gearbox : Gearbox temperature threshold. Under sub-rated operating conditions, v sub Sub-rated wind speed; P sub Subrated power; ω sub Sub-rated generator speed. In standby mode, v cut-in Cut-in wind speed, P standby Standby power, ω standby Standby speed. In the off state, v cut-out Cut off the wind speed, P off Shutdown power, ω off Stop speed:

[0104] Grid dispatch command dimension, P target Target power value, S∈{0,1}: Start-up / shutdown control signal, S=1: Grid dispatch requires unit start-up; S=0: Grid dispatch requires unit shutdown. E∈{0,1}: Emergency shutdown signal, E=1: Grid fault or extreme weather conditions require emergency shutdown of the unit; E=0: Unit is operating normally, T response Response time constraint.

[0105] The safety domain constraint dimensions are as follows: Safety domain A is the safety domain for the unit under rated operating conditions; Safety domain B is the safety domain for the unit under sub-rated operating conditions; Safety domain C is the safety domain for the unit under standby conditions; and Safety domain D is the safety constraint for the unit under shutdown conditions.

[0106] The S200 performs the first-level trigger judgment based on short-term wind and solar power prediction and safety domain, the second-level judgment based on grid dispatch instructions, and arbitration using a three-dimensional comparison matrix in case of conflict. Finally, it merges the two-level signals to generate a wind turbine start-up and shutdown strategy that takes into account both safety and command requirements.

[0107] Please refer to Figure 3 The diagram illustrates a flowchart of an example of a safety domain-aware intelligent start-stop decision method for wind turbines, S200, which includes:

[0108] S210: The first-level trigger judgment signal is based on the short-term wind and solar power prediction results and the safety domain.

[0109] The first-level trigger judgment signal is to dynamically compare the short-term wind and solar power prediction results with the safety domain-command three-dimensional comparison matrix generated by S130 to identify the potential risks of wind and solar power fluctuations to the safe operation of the unit.

[0110] Specifically, short-term wind and solar power forecasts provide power output trends for the next 15-30 minutes, while the safety domain defines the safety thresholds for various operating conditions. Combining the two allows for the quantification of the degree to which wind and solar power fluctuations intrude into the unit's safety domain, triggering start-up and shutdown actions to mitigate risks.

[0111] Short-term wind and solar power forecasting is a prerequisite for the first-level trigger judgment signal. In one possible implementation, the prediction model is mathematically expressed as:

[0112]

[0113] in, To predict power, W history For historical wind speed series, M weather For meteorological data, F turbulence G represents the intensity characteristics of turbulence. solar For the contribution of photovoltaic power, α is the model weighting coefficient (taken as 0.3-0.7, f LSTM f is the LSTM model function. RF This is the function for the random forest model.

[0114] The first-level trigger judgment signal will predict the short-term result. Real-time comparison is performed with the corresponding security domain in the security domain-instruction three-dimensional comparison matrix M.

[0115] In one possible implementation, the process of comparing the short-term prediction results with the corresponding security domain in the security domain-instruction three-dimensional comparison matrix is ​​as follows:

[0116] Calculate the degree of matching between the power change rate within the future time window and the allowable fluctuation range of the safety domain. If the power change rate exceeds the threshold, a start / stop decision is triggered.

[0117] The relationship between predicted power and safety domain parameters is obtained from the power curve of the wind turbine, and the values ​​of safety domain parameters are obtained.

[0118] Determine whether the security domain parameter values ​​conform to the security domain and generate a trigger signal for switching the first-level operating state.

[0119] S220: The second-level trigger judgment signal is based on the response to the power grid dispatch command.

[0120] The second-layer triggering and judgment signal focuses on the real-time parsing and response to grid dispatching commands, overcoming the limitations of the first layer which only focuses on wind and solar power fluctuations. It transforms external grid commands into executable start-stop signals, ensuring that unit operation meets grid stability requirements. The second-layer triggering and judgment signal is responsible for parsing command types, assessing response urgency, and linking with the security domain-command three-dimensional comparison matrix.

[0121] The parsing of power grid dispatch instructions is the starting point for the second-level trigger judgment signal. According to the security domain-instruction three-dimensional comparison matrix, instruction types include:

[0122] Standard command: Target power P target S∈{0,1}: Start-up and shutdown control signals; S=1: Power grid dispatching requires the unit to start.

[0123] Emergency Command: E∈{0,1}: Emergency shutdown signal; E=1: Requires emergency shutdown of the unit under power grid failure or extreme weather conditions; E=0: The unit is operating normally. response Response time constraint.

[0124] In one possible implementation, the specific process for determining the second-level trigger signal is as follows:

[0125] Receive and identify parameters such as target power value, start / stop signal, emergency stop signal and response time constraint in the scheduling command;

[0126] Execution priority is set according to the command type. In one possible implementation, command priority is divided according to the power grid security level: emergency shutdown (highest level) > state switching command > routine power adjustment (lowest level).

[0127] Based on the priority order, the scheduling instruction parameters are matched with the security domains in the security domain-instruction three-dimensional comparison matrix to determine whether the instruction meets the safe operation conditions of the unit.

[0128] Based on the matching result between the instruction and the security domain, a trigger signal for switching the second-level running state is generated.

[0129] S230: When the first-level trigger judgment signal and the second-level trigger judgment signal conflict, priority arbitration is performed with the security domain-instruction three-dimensional comparison matrix as a constraint.

[0130] When a conflict arises between the first-level trigger judgment signal and the second-level trigger judgment signal, priority arbitration must be performed based on the constraints between the operating state, safety domain, and grid dispatch commands defined in the safety domain-command three-dimensional comparison matrix to ensure that the unit operates within a safe and controllable range. Conflicts typically manifest as inconsistencies between the state switching recommendations triggered by wind and solar power fluctuations and the state requirements of the grid dispatch commands.

[0131] In one possible implementation, the conflict types and priority arbitration rules are as follows:

[0132] Conflict type 1: Security priority conflict.

[0133] Scenario: The first-level trigger judgment signal is due to wind and solar fluctuations, such as a predicted sudden drop in power, suggesting shutdown or power reduction, while the second-level trigger judgment signal is due to grid dispatch instructions requiring power to be maintained or increased.

[0134] Arbitration rule: The first-level trigger judgment signal is executed first to ensure unit safety. This is based on the fact that the safety domain in the safety domain-command three-dimensional comparison matrix is ​​a hard constraint; grid commands must be executed within the safety domain. If a grid command requires execution to exceed the safety domain, it will be rejected.

[0135] Conflict type 2: Emergency command priority conflict.

[0136] Scenario: The second-level trigger signal is an emergency shutdown command, such as a power grid failure or extreme wind speed, while the first-level trigger signal does not indicate any risk.

[0137] Arbitration rule: The second-level trigger judgment signal is enforced, and the first-level result is ignored. This is based on the fact that the emergency shutdown signal has the highest priority and requires immediate response to avoid grid collapse or equipment damage.

[0138] Conflict type 3: Conflict between regular instructions and security domains.

[0139] Scenario: The second-level trigger signal is a regular power adjustment command, but the first-level trigger signal predicts that wind and solar fluctuations will cause the power to exceed the safe range.

[0140] Arbitration rule: Refusal or delay of instruction execution prioritizes unit safety. This is based on the fact that the safety domain parameters in the safety domain-instruction three-dimensional comparison matrix represent insurmountable physical limitations, requiring power limiting to mitigate risks.

[0141] Conflict type 4: State transition suggestion conflict.

[0142] Scenario: The first-level trigger signal suggests switching from rated operation to sub-rated operation, while the second-level trigger signal requires maintaining rated operation.

[0143] Arbitration rule: Based on the overlapping area of ​​the security domains between states in the security domain-command three-dimensional comparison matrix, select the target state that is closest to the current state and has the lowest risk. The basis is that if the current state is rated operation, and there is partial overlap between the security domains of sub-rated operation and rated operation, then the sub-rated operation state with the smallest fluctuation is selected.

[0144] In one possible implementation, the arbitration process includes:

[0145] Input the first-level trigger judgment signal and the second-level trigger judgment signal.

[0146] Conflict detection is performed by comparing the target state and security domain constraints of the two-layer signals.

[0147] Priority determination: If the second-level signal is an emergency shutdown, the shutdown is executed directly; if the first-level signal involves a safety domain breach (such as vibration exceeding limits), the first level is executed first; for other conflict scenarios, the optimal state is selected based on the overlapping area of ​​the safety domain.

[0148] Output the target state after arbitration.

[0149] S240: Integrates dual-layer trigger judgment signals to generate start / stop strategies.

[0150] After conflict arbitration, the results of the first and second layer trigger judgment signals are fused to generate a start-up and shutdown strategy for the wind turbine. The start-up and shutdown strategy integrates the effects of wind and solar power fluctuation prediction and grid dispatch instructions, and has been corrected by safety domain constraints in the event of a conflict, thus possessing both executability and safety.

[0151] In one possible implementation, the fusion process is as follows:

[0152] Extract the state switching signal output from the first-level trigger judgment signal and the instruction response signal output from the second-level trigger judgment signal;

[0153] If the two layers of judgment are consistent, the consistent state suggestion is directly output; if the two layers of judgment conflict, the signal after arbitration by S230 is used as the final target state.

[0154] If multiple feasible states still exist after arbitration, the state with the lowest overall risk and the highest return will be selected as the preliminary recommendation.

[0155] The start / stop policy is output in the form of a status code, including the target status, expected duration, power setting value, and safety domain verification flag.

[0156] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0157] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0158] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0159] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0160] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0161] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart start-stop decision-making method for wind turbines based on safety domain awareness, characterized in that, include: Based on multi-source heterogeneous data from wind turbine generators, a multi-state operation profile table is constructed. The kernel density estimation algorithm is used to dynamically calculate the security domain of each state in the multi-state operation profile table. The security domain and power grid dispatch instructions are combined to form a three-dimensional comparison matrix of security domain and instruction. Based on the security domain-command three-dimensional comparison matrix, a two-layer trigger judgment signal for wind and solar fluctuations and grid demand is used. When the first-layer trigger judgment signal and the second-layer trigger judgment signal conflict, priority arbitration is performed with the security domain-command three-dimensional comparison matrix as a constraint. A dual-layer trigger judgment signal generation start / stop strategy is implemented; Among them, the first-level trigger judgment signal is based on the short-term wind and solar power prediction results and the safety domain, and the second-level trigger judgment signal is based on the response to the grid dispatch command.

2. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 1, characterized in that, The multi-source heterogeneous data includes historical operating data of wind turbine units, real-time meteorological information, and equipment health status data. The historical operating data includes SCADA data on power output, wind speed, generator speed, and gearbox temperature. The real-time meteorological information includes wind speed, wind direction, ambient temperature, and air pressure data from wind towers or weather forecasts. The equipment health status data includes status monitoring data on gearbox vibration, bearing temperature, and hydraulic system pressure.

3. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 1, characterized in that, The multi-state operation profile table divides the wind turbine's operating state into rated operating state, sub-rated operating state, standby state, and shutdown state. The rated operating state is defined as the wind speed being between the rated wind speed and the cut-out wind speed and the power output fluctuation range being within ±5% of the rated power. The sub-rated operating state is defined as a wind speed lower than the rated wind speed and a power output lower than 95% of the rated power but higher than the minimum power generation power corresponding to the cut-in wind speed. The standby state is defined as the wind speed being lower than the cut-in wind speed, the power grid dispatching command requiring the suspension of operation, or the equipment being in maintenance mode. The shutdown state is defined as when the equipment health monitoring data exceeds the safety threshold, the wind speed exceeds the safety range, or the power grid commands a forced shutdown.

4. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 3, characterized in that, The multi-state operation profile table: The key coupling variables for rated operating conditions include wind speed, power, generator speed, and gearbox vibration acceleration; The key coupled variables for sub-rated operation include wind speed, pitch angle, power, and blade root bending moment. The key set of coupled variables in standby mode includes grid frequency, voltage, device standby time, and ambient temperature. The key set of coupled variables in the shutdown state includes extreme wind speed, peak vibration, fault codes, and safety chain status.

5. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 1, characterized in that, The step of dynamically calculating the security domain using the kernel density estimation algorithm includes: applying a multivariate kernel density estimation formula to each operating state: in, This represents the state distribution probability density estimate of the input variable x, where x represents the input variable. i Let represent the variable value of the i-th data point, K(·) represent the kernel function, h is the bandwidth parameter used to control the smoothness of the estimation, and n represents the number of variables.

6. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 5, characterized in that, The security domain is determined by setting a probability threshold α: in, It is the probability density of the state distribution obtained by kernel density estimation, and α is a preset probability threshold; like Then x is considered a value within the security region; if Then x is considered to be outside the security domain.

7. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 1, characterized in that, The mathematical expression of the safety domain-command three-dimensional comparison matrix is ​​a three-dimensional tensor structure. Its operating state dimension includes safety domain parameters for rated operating state, sub-rated operating state, standby state, and shutdown state. The safety domain parameter dimension includes thresholds for wind speed, power output, generator speed, and gearbox temperature. The power grid dispatch command parameter dimension includes target power values, start / stop control signals, emergency shutdown signals, and response time constraints. Furthermore, the safety domain-command three-dimensional comparison matrix maps the feasibility boundary of state switching through the safety domain constraint dimension. The mathematical expression of the safety domain-command three-dimensional comparison matrix is ​​as follows: Where: M: Security domain-instruction three-dimensional comparison matrix; under rated operating conditions, v rated Rated wind speed; P rated Rated power; ω rated Rated generator speed; T gearbox : Gearbox temperature threshold; under sub-rated operating conditions, v sub Sub-rated wind speed; P sub Subrated power; ω sub Sub-rated generator speed; in standby mode, v cut-in Cut-in wind speed, P standby Standby power, ω standby Standby speed; in the off state, v cut-out Cut off the wind speed, P off Shutdown power, ω off Stop speed; Grid dispatch command dimension, P target Target power value, S∈{0,1}; Start-stop control signal, S=1: Grid dispatch requires unit to start; S=0: Grid dispatch requires unit to stop; E∈{0,1}: Emergency stop signal, E=1: Grid fault or extreme weather conditions require emergency stop of unit; E=0: Unit is operating normally, T response Response time constraints; The safety domain constraint dimension is as follows: safety domain A is the safety domain of the unit under rated operating conditions; safety domain B is the safety domain of the unit under sub-rated operating conditions; safety domain C is the safety domain of the unit under standby conditions; and safety domain D is the safety constraint of the unit under shutdown conditions.

8. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 1, characterized in that, Its features are, The step of determining the first-level triggering signal based on short-term wind and solar power prediction results and the safety domain includes: The short-term wind and solar power is predicted using a predictive model, expressed as follows: in, To predict power, W history For historical wind speed series, M weather For meteorological data, F turbulence G represents the intensity characteristics of turbulence. solar For the contribution of photovoltaic power, α is the model weighting coefficient (taken as 0.3-0.7, f LSTM f is the LSTM model function. RF This is the function for the random forest model; Short-term forecast results Real-time comparison is performed with the corresponding security domain in the security domain-instruction three-dimensional comparison matrix M; the comparison process is as follows: Calculate the matching degree between the power change rate within the future time window and the allowable fluctuation range of the safety domain; if the power change rate exceeds the threshold, trigger the start / stop judgment. The relationship between predicted power and safety domain parameters is obtained from the power curve of the wind turbine, and the values ​​of safety domain parameters are obtained. Determine whether the security domain parameter values ​​conform to the security domain and generate a trigger signal for switching the first-level operating state; The process for determining the second-level trigger signal based on the response to power grid dispatch instructions is as follows: Receive and identify parameters such as target power value, start / stop signal, emergency stop signal and response time constraint in the scheduling command; The execution priority is set according to the command type. The command priority is divided into the following categories according to the power grid security level from high to low: emergency shutdown > state switching command > normal power adjustment. Based on the priority order, the scheduling instruction parameters are matched with the security domains in the security domain-instruction three-dimensional comparison matrix to determine whether the instruction meets the safe operation conditions of the unit. Based on the matching result between the instruction and the security domain, a trigger signal for switching the second-level running state is generated.

9. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 8, characterized in that, When the first-level trigger judgment signal and the second-level trigger judgment signal conflict, the priority arbitration is performed based on the security domain-instruction three-dimensional comparison matrix: if the second-level trigger judgment signal contains an emergency stop signal, then the second-level trigger judgment signal is forced to be executed first. If the first-level trigger judgment signal indicates that the critical coupling variable is outside the security domain, then the first-level trigger judgment signal will be executed first. For regular command conflicts, the target state with the lowest risk and the least fluctuation from the current state is selected based on the overlapping area of ​​the security domain, and the arbitrated target state code is output, including the target state, expected duration, power setting value and security domain verification flag.

10. The intelligent start-stop decision-making method for wind turbines based on safety domain awareness according to claim 9, characterized in that, The generation of the start / stop strategy includes the following steps: The results of the first and second layer trigger judgment signals are combined. If there is a conflict, arbitration is performed based on the security domain-instruction three-dimensional comparison matrix. The start / stop policy is output in the form of a status code, including the target status, expected duration, power setting value, and safety domain verification flag; The start-stop strategy achieves coordinated optimization of wind and solar power fluctuations and grid demand by dynamically adjusting the safety domain constraint parameters.

Images