Wind turbine generator component life probability prediction and wind power plant energy control method and system

Abstract

The invention discloses a wind turbine generator component life probability prediction method and system, and a wind power plant energy control method and system. The prediction method comprises the steps of constructing a state space model of a degradation process of a wind turbine generator component; acquiring real-time state information of the component; inputting the real-time state information into the state space model, evaluating the degradation state of the component by adopting a sequential data assimilation algorithm, and outputting posterior probability distribution of the degradation state; and based on the posterior probability distribution, forward simulating a future degradation path of the component by adopting Monte Carlo simulation until the future degradation path exceeds a preset failure threshold, and obtaining residual life probability distribution of the component. According to the method, the personalized residual life of each unit can be accurately evaluated, and an innovative wind power plant energy management control strategy is created on the basis, so that the absolute maximization of the full-life-cycle economic benefit of the wind power plant in a complex market environment is realized.

Description

Technical Field

[0001] This invention relates to the field of wind turbine technology, and in particular to a method and system for predicting the lifespan probability of wind turbine components, and a method and system for controlling energy in wind farms. Background Technology

[0002] With the transformation of the global energy structure and the deepening of electricity market reform, the operating goal of wind farms has shifted from maximizing power generation to maximizing power generation revenue, and requires the ability to respond to market price fluctuations in real time.

[0003] Under this new situation, wind farm operation faces unprecedented challenges. On the one hand, although wind turbine generators are standardized products, once installed in a specific wind farm, the micro-topography and wind resource conditions at each turbine site vary significantly. This fixed installation location and the non-uniformity of environmental resources lead to a key issue: even for the same model of turbine, the actual load history and degradation rate of its key components differ greatly. Typically, turbines at sites with excellent wind resources have higher output and higher power generation revenue, but their components also experience greater mechanical stress and faster lifespan degradation; conversely, turbines at sites with poor wind resources have lower output and slower degradation. This phenomenon of "same factory, same model, different fate" makes the traditional "one-size-fits-all" operation and maintenance strategy based on uniform time intervals or average operating hours no longer economical and reasonable.

[0004] On the other hand, the volatility of the electricity market demands that wind farms possess exceptional flexibility. Ideally, during periods of high electricity prices, wind farms should be willing to appropriately increase equipment losses to "maximize power generation," even allowing well-healthy units to "appropriately over-generate" within safety margins, thereby capturing high-value electricity. Conversely, during periods of low or zero / negative electricity prices, the operational strategy should shift to "equipment maintenance," proactively reducing load to delay component degradation and utilizing equipment's "lifespan depletion" effectively. However, achieving such refined and dynamic strategy adjustments cannot be accomplished through traditional manual decision-making and simple control logic.

[0005] Currently, existing technologies have the following main limitations in wind farm operation:

[0006] 1. Component condition prediction lacks individualization and probabilistic approach: Traditional alarm systems based on fixed thresholds cannot cope with the individual differences of wind turbine generators, which are "from the same factory and of the same type but with different fates", and it is difficult to provide accurate early warnings.

[0007] Existing data-driven diagnostic models struggle to quantify prediction uncertainties and cannot effectively integrate component physical models, resulting in an inability to accurately assess individualized Remaining Useful Life (RUL) caused by different operating loads. Consequently, their predictions are difficult to directly apply to economic decisions that require balancing short-term power generation benefits with long-term equipment health.

[0008] The lack of a dynamically updated degradation trajectory tracking mechanism results in limited prediction accuracy.

[0009] 2. Rigid energy management and control strategies for wind farms: Current mainstream power allocation strategies completely isolate market price signals from equipment health status, making it impossible to automatically identify high-price windows and incentivize healthy units to generate more power, or to intelligently reduce load during low-price periods to achieve lifespan equalization.

[0010] The lack of an optimization model that quantifies electricity prices, equipment remaining life probability prediction, power generation revenue, and potential failure losses makes it difficult to automatically generate and execute optimal control commands based on preset economic objectives. It relies heavily on human experience and judgment, resulting in slow response and difficulty in achieving optimal performance.

[0011] In summary, existing technologies suffer from several core defects, including failure to quantify uncertainty in state prediction and inability to reflect individual differences among units, insufficient integration of physical models and real-time data, and rigid control strategies. Summary of the Invention

[0012] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for predicting the lifespan probability of wind turbine components and a method and system for controlling wind farm energy. This method can accurately assess the individual remaining lifespan of each turbine and, based on this, create an innovative wind farm energy management and control strategy to maximize the economic benefits of wind farms throughout their entire life cycle in complex market environments.

[0013] The technical solution of the present invention provides a method for predicting the lifespan probability of wind turbine components, comprising: A state-space model of the degradation process of wind turbine components is constructed. The state-space model includes state equations and observation equations. The state equations represent the evolution of degradation state variables under the drive of operating loads, and the observation equations represent the relationship between the degradation state variables and observable health indicators. Obtain the real-time status information of the component; The real-time state information is input into the state space model, and the degradation state of the component is evaluated using a sequential data assimilation algorithm, and the posterior probability distribution of the degradation state is output. Based on the posterior probability distribution, Monte Carlo simulation is used to simulate the future degradation path of the component forward until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

[0014] In one alternative technical solution, the step of employing a sequential data assimilation algorithm to evaluate the degradation state of the component includes: Based on the state equation, state transition prediction is performed on the initial particle set to obtain the prior probability distribution of the degenerate state; If a new observation is received, the weight of each predicted particle is calculated according to the observation equation. The weights are normalized to obtain the posterior probability distribution.

[0015] In one of the alternative technical solutions, after normalizing the weights to obtain the posterior probability distribution, the method further includes: The normalized particle set is resampled according to the normalized weights to obtain the target particle set, and the target particle set is used as the initial particle set for the next time step.

[0016] The technical solution of the present invention also provides a wind turbine component life probability prediction system, comprising: The model building unit is used to build a state-space model of the degradation process of wind turbine components. The state-space model includes state equations and observation equations. The state equations are the evolution law of the degradation state variables under the drive of operating loads, and the observation equations are the relationship between the degradation state variables and observable health indicators. An acquisition unit is used to acquire the real-time status information of the component; The posterior probability distribution output unit is used to input the real-time state information into the state space model, evaluate the degradation state of the component using a sequential data assimilation algorithm, and output the posterior probability distribution of the degradation state. The remaining lifetime probability distribution acquisition unit is used to simulate the future degradation path of the component forward using Monte Carlo simulation based on the posterior probability distribution, until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

[0017] In one of the alternative technical solutions, the posterior probability distribution output unit is further used for: Based on the state equation, state transition prediction is performed on the initial particle set to obtain the prior probability distribution of the degenerate state; If a new observation is received, the weight of each predicted particle is calculated according to the observation equation. The weights are normalized to obtain the posterior probability distribution.

[0018] One of the alternative technical solutions also includes: The target particle set acquisition unit is used to resample the normalized particle set according to the normalized weights to obtain the target particle set, and use the target particle set as the initial particle set for the next time step.

[0019] The technical solution of the present invention also provides a wind farm energy control method, comprising: The remaining life probability distribution of each wind turbine component, current generating capacity, external environmental information, and real-time electricity price are predicted using the wind turbine component life probability prediction method described above. With the maximization of the expected total revenue throughout the entire life cycle of a wind farm as the core objective, a multi-objective optimization function is constructed. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining life probability distribution of the components. Solving the multi-objective optimization function generates optimized control instructions for each wind turbine, which include the optimal power setpoint and the drive operation strategy based on the real-time electricity price. Issue the aforementioned optimized control command.

[0020] In one of the alternative technical solutions, solving the multi-objective optimization function to generate the optimized control command corresponding to each wind turbine includes: Solve the multi-objective optimization function to obtain the real-time health weight factor for each wind turbine, wherein the real-time health weight factor is inversely proportional to the failure risk of the wind turbine; The optimal power setting value is generated based on the health weighting factor.

[0021] In one of the alternative technical solutions, the real-time electricity price includes a high electricity price and a low electricity price. Solving the multi-objective optimization function to generate optimized control instructions for each wind turbine includes: During the high electricity price period, the wind turbine units are controlled to operate at full capacity or exceed capacity. During periods of low electricity prices, the load on high-risk wind turbines is reduced.

[0022] The technical solution of the present invention also provides a wind farm energy control system, comprising: The acquisition unit is used to acquire the probability distribution of the remaining lifespan of each wind turbine component, the current generating capacity, external environmental information and real-time electricity price, which are predicted by the wind turbine component lifespan probability prediction method described above. The objective optimization function construction unit is used to construct a multi-objective optimization function with the core objective of maximizing the expected total revenue of the wind farm throughout its entire life cycle. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining life probability distribution of the component. An optimized control command generation unit is used to solve the multi-objective optimization function and generate optimized control commands for each wind turbine. The optimized control commands include the optimal power setpoint and the drive operation strategy based on the real-time electricity price. The instruction issuing unit is used to issue the optimization control instruction.

[0023] The above technical solution has the following beneficial effects: 1. By sequentially assimilating physical models with real-time monitoring data, dynamic tracking of component degradation status is achieved, overcoming the individual differences of units that traditional methods cannot handle, and outputting a remaining life prediction with a probability distribution, thereby quantifying prediction uncertainty.

[0024] 2. By directly inputting the remaining lifetime prediction results with probability distribution into the multi-objective optimization function of wind farm energy management and combining it with real-time electricity price, dynamic and automated coupling of equipment health status and real-time electricity price is realized, ultimately achieving the goal of maximizing the economic benefits of the entire wind farm throughout its entire life cycle in a complex market environment. Attached Figure Description

[0025] The disclosure of this invention will become more readily understood by referring to the accompanying drawings. It should be understood that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings: Figure 1 A flowchart illustrating a method for predicting the lifespan probability of wind turbine components according to an embodiment of the present invention; Figure 2 A flowchart illustrating a method for predicting the lifespan probability of wind turbine components, provided in another embodiment of the present invention; Figure 3 A flowchart illustrating a method for predicting the lifespan probability of wind turbine components, provided in another embodiment of the present invention; Figure 4 This is a schematic diagram of a wind turbine component life probability prediction system provided in an embodiment of the present invention; Figure 5 A flowchart illustrating the operation of a wind farm energy control method according to an embodiment of the present invention; Figure 6 A flowchart illustrating a wind farm energy control method according to another embodiment of the present invention; Figure 7This is a schematic diagram of a wind farm energy control system provided in an embodiment of the present invention. Detailed Implementation

[0026] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0027] It is readily understood that, based on the technical solution of this invention, various structural and implementation methods can be interchanged by those skilled in the art without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of the invention.

[0028] The directional terms such as up, down, left, right, front, back, front, back, top, and bottom mentioned or possibly used in this specification are defined relative to the structures shown in the accompanying drawings. They are relative concepts and may therefore vary depending on their location and usage. Therefore, these or other directional terms should not be interpreted as restrictive.

[0029] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting the lifespan probability of wind turbine components, comprising: Step S101: Construct a state-space model of the degradation process of wind turbine components. The state-space model includes state equations and observation equations. The state equations represent the evolution of degradation state variables under the drive of operating loads, and the observation equations represent the relationship between the degradation state variables and observable health indicators. Step S102: Obtain the real-time status information of the component; Step S103: Input the real-time state information into the state space model, and use the sequential data assimilation algorithm to evaluate the degradation state of the component, and output the posterior probability distribution of the degradation state; Step S104: Based on the posterior probability distribution, Monte Carlo simulation is used to simulate the future degradation path of the component forward until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

[0030] Specifically, the controller executes step S101 to construct a state-space model to describe the degradation state of components (including bearings, gearboxes, generators, blades, etc.). This state-space model includes state equations and observation equations, and the degradation state variables include crack depth or cumulative damage.

[0031] 1. Equations of State State equations are used to describe the inherent evolution of a component's degradation state, driven by operational loads. The equations are:

[0032] in, : refers to the state variable, which is defined at discrete time steps. State variables are key variables representing the health status of components. State variables are quantities that cannot be directly measured but can reflect the nature of degradation.

[0033] For example, for bearings: It can be the equivalent crack depth (Unit: mm)

[0034] For example, for gears: It can be the cumulative damage level (Dimensionless, ranging from 0 to 1, where 1 represents failure).

[0035] : A function that describes the state evolution, based on physical mechanisms (such as Paris crack propagation law, Archard wear model) or empirical models.

[0036] : at time step at time step The load vector is derived from SCADA data, such as average torque, speed, and turbulence intensity.

[0037] The physical parameters of the model, such as material constants and geometric parameters, may be partially estimated.

[0038] Process noise represents the uncertainty of the model, and is usually assumed to be Gaussian white noise, i.e. .

[0039] Taking the bearing crack depth model as an example, its state equation is: in, and It is a material constant. It is due to the load The calculated stress range.

[0040] 2. Observation Equation The observation equation establishes the connection between the hidden state variables and the observable monitoring data.

[0041] : For the observed variable, at time step Health indicators related to state variables are extracted from sensor data (such as vibration acceleration and temperature). Examples include: the envelope spectrum amplitude of the vibration signal at the fault characteristic frequency, the temperature rise relative to the ambient temperature, and the concentration of abrasive particles in the oil.

[0042] The observation equation is in the form of: , : is the observation function, which describes how the state variables are mapped to the observation values. It can be linear or nonlinear.

[0043] : These are the parameters of the observation model.

[0044] : Observation noise, representing measurement error and feature extraction error, is usually assumed to be Gaussian white noise, i.e. .

[0045] Taking the vibration observation model as an example, its observation equation is:

[0046] This is an exponential model indicating the crack depth. A tiny increase can lead to vibration characteristics The sharp increase is consistent with the actual situation of the project.

[0047] Then, step S102 is executed to obtain the real-time status information of the component. The real-time status information includes the equivalent crack depth, cumulative damage degree, and health indicators related to the status variables detected by the sensor (such as the envelope spectrum amplitude of the vibration signal at the fault characteristic frequency, the temperature rise relative to the ambient temperature, and the concentration of abrasive particles in the oil).

[0048] Next, step S103 is executed, whereby real-time state information is input into the state-space model, and a sequential data assimilation algorithm is used to evaluate the degradation state of the component, outputting the posterior probability distribution of the degradation state. In another embodiment, a particle filter algorithm is used as the sequential data assimilation algorithm to solve the state estimation problem of nonlinear, non-Gaussian systems. The particle filter uses a set of random samples (particles) to approximate the posterior probability distribution of the state variables.

[0049] Finally, step S104 is executed, where the remaining lifetime (RUL) is defined as the first time the current state reaches the failure threshold. The time frame is used to simulate the future degradation path of each particle in the assimilated set at the current moment, starting from the state of each particle, until the failure threshold is exceeded for the first time. Then, the time steps taken for all simulation paths are counted to obtain the empirical probability distribution of the remaining lifetime (RUL), and the expected remaining lifetime (RUL), confidence interval, and probability of future failure are output.

[0050] In this embodiment, by sequentially assimilating the physical model with real-time monitoring data, dynamic tracking of component degradation status is achieved, overcoming the individual differences of the unit that are difficult to deal with by traditional methods, and outputting a remaining life prediction with a probability distribution, thereby quantifying the uncertainty of prediction.

[0051] like Figure 2 As shown, a preferred embodiment of the present invention provides a method for predicting the lifespan probability of wind turbine components, comprising: Step S201: Construct a state-space model of the degradation process of wind turbine components; Step S202: Obtain the real-time status information of the component; Step S203: Based on the state equation, perform state transition prediction on the initial particle set to obtain the prior probability distribution of the degenerate state; Step S204: If a new observation is received, calculate the weight of each predicted particle according to the observation equation; Step S205: Normalize the weights to obtain the posterior probability distribution; Step S206: Resample the normalized particle set according to the normalized weights to obtain the target particle set, and use the target particle set as the initial particle set for the next time step; Step S207: For the assimilated particle set at the current moment, starting from the state of each particle, use Monte Carlo simulation to simulate the future degradation path of the component, so as to obtain the remaining lifetime probability distribution of the component when it first exceeds the preset failure threshold.

[0052] Specifically, assuming at time... We have a set of particles. ,in, It is the first Individual particle state values It is its weight (representing the credibility of the particle), and the sum of the weights is 1.

[0053] Step S203: Prediction step (state transition); For each particle Prediction is made based on the state equation:

[0054] At this time, the particle set This represents integration Prior probability distribution of the state before the new observation data at time step .

[0055] Step S204: Update steps (incorporate observations); When new observation values Upon arrival, calculate the weight of each predicted particle:

[0056] in, It is called the likelihood function, which represents the likelihood function when the state is Observed at time The probability. It is usually assumed that the observation noise follows a Gaussian distribution, then this probability is proportional to... .

[0057] Step S205: Weight normalization step; Normalize the weights: .

[0058] Normalized particle set The posterior probability distribution representing the state This is currently the best estimate of the component's health status.

[0059] Step S206: Resampling step; To avoid particle degradation (i.e., a few particles have excessively high weights while the rest have nearly zero weights), resampling is necessary. This is based on the weights... Resampling A new set of equally weighted particles is obtained. Particles with higher weights are replicated more often, while particles with lower weights are discarded. After resampling, a new set of equally weighted particles is obtained. It still approximately represents the posterior distribution. This set of particles will be passed to the next time step. .

[0060] Step S207: Remaining life probability prediction step.

[0061] For the current moment Assimilated particle set Each particle represents a possible current state of a component.

[0062] Forward simulation: for each particle In its current state Starting from this point, we use state equations to simulate its future degradation path forward until the path first exceeds a threshold. .

[0063] calculate The number of time steps taken to complete this path represents the remaining lifetime of the particle. .

[0064] Formation distribution: for all individual particles By performing statistical analysis, the remaining lifespan can be obtained. The empirical probability distribution.

[0065] Output: Expected RUL: ; Confidence interval: for example, The 90% confidence interval; Failure probability: Future Probability of failure within 1 hour .

[0066] In this embodiment, by sequentially assimilating the physical model with real-time monitoring data, dynamic tracking of component degradation status is achieved, overcoming the individual differences of the unit that are difficult to deal with by traditional methods, and outputting a remaining life prediction with a probability distribution, thereby quantifying the uncertainty of prediction.

[0067] like Figure 3 As shown, another embodiment of the present invention provides a method for predicting the lifespan probability of wind turbine components, comprising: Step S301: Construct a physical model of component degradation and define an observation model; Step S302: Real-time acquisition of multi-source monitoring data, including vibration, temperature, load, etc.; Step S303: Feature extraction and calculation of health index HI; Step S304: Data assimilation loop; Step S305: Prediction step, predicting state evolution based on the model; Step S306: Update step, compare the prediction with the observation data and update the particle weights; Step S307: Resampling, retaining high-weight particles; Step S308: Output the posterior probability distribution of the current state and parameters; Step S309: Calculate the probability distribution of RUL, including expected value, confidence interval, and failure probability; Step S310: Generate a RUL probability prediction report; Step S311: Determine whether to continue monitoring. If yes, proceed to step S302; otherwise, end.

[0068] like Figure 4 As shown, an embodiment of the present invention provides a wind turbine component lifespan probability prediction system, comprising: Model building unit 401 is used to build a state space model of the degradation process of wind turbine components. The state space model includes state equations and observation equations. The state equations are the evolution law of the degradation state variables under the drive of operating loads, and the observation equations are the relationship between the degradation state variables and observable health indicators. Acquisition unit 402 is used to acquire the real-time status information of the component; The posterior probability distribution output unit 403 is used to input the real-time state information into the state space model, evaluate the degradation state of the component using a sequential data assimilation algorithm, and output the posterior probability distribution of the degradation state. The remaining lifetime probability distribution acquisition unit 404 is used to simulate the future degradation path of the component based on the posterior probability distribution using Monte Carlo simulation until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

[0069] In this embodiment, by sequentially assimilating the physical model with real-time monitoring data, dynamic tracking of component degradation status is achieved, overcoming the individual differences of the unit that are difficult to deal with by traditional methods, and outputting a remaining life prediction with a probability distribution, thereby quantifying the uncertainty of prediction.

[0070] In one embodiment, the sequential data assimilation algorithm is a particle filter algorithm.

[0071] In one embodiment, the posterior probability distribution output unit is further configured to: Based on the state equation, state transition prediction is performed on the initial particle set to obtain the prior probability distribution of the degenerate state; If a new observation is received, the weight of each predicted particle is calculated according to the observation equation. The weights are normalized to obtain the posterior probability distribution.

[0072] In one embodiment, it further includes: The target particle set acquisition unit is used to resample the normalized particle set according to the normalized weights to obtain the target particle set, and use the target particle set as the initial particle set for the next time step.

[0073] like Figure 5 As shown, an embodiment of the present invention provides a wind farm energy control method, comprising: Step S501: Obtain the remaining life probability distribution of each wind turbine component, current generating capacity, external environment information and real-time electricity price predicted using the wind turbine component life probability prediction method described above; Step S502: With the maximization of the expected total revenue of the wind farm throughout its entire life cycle as the core objective, a multi-objective optimization function is constructed. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining life probability distribution of the components. Step S503: Solve the multi-objective optimization function to generate the optimized control command corresponding to each wind turbine. The optimized control command includes the optimal power setpoint and the drive operation strategy based on the real-time electricity price. Step S504: Issue the optimized control command.

[0074] Specifically, when it is necessary to optimize the power allocation of the wind farm, the controller executes step S501 to obtain the remaining life probability distribution, current generating power, external environmental information (such as wind speed) and real-time electricity price of the key components (including bearings, gearboxes, generators, blades, etc.) of each wind turbine using the above-mentioned wind turbine component life probability prediction method.

[0075] Then, step S502 is executed, with the core objective of maximizing the expected total revenue throughout the wind farm's lifecycle. A multi-objective optimization function is constructed, comprehensively considering power generation revenue, expected failure risk costs, and planned maintenance costs. The goal is to maximize the total revenue encompassing these three factors. By incorporating expected failure risk costs into the optimization objective, a direct coupling between equipment health status and economic benefits is achieved. Specifically, the expected failure risk cost is proportional to the real-time failure probability calculated from the component's remaining lifespan probability distribution, while power generation revenue is determined by the product of power generation and real-time electricity price. The expected failure risk cost is proportional to the real-time failure probability and potential failure losses of each component.

[0076] In another embodiment, the multi-objective optimization function is: , in, The expected total return; To optimize the time cycle; For the unit exist The optimal power setting value at any given time; Let be the real-time electricity price at time t; For the unit exist The expected failure risk cost at any given moment; For planned maintenance costs; To achieve a given power setpoint Next, the unit exist The probability of a failure occurring at any given time; For the unit Potential losses due to failure (repair and downtime losses); The total power target issued by the power grid; This represents the total number of wind turbines in the wind farm.

[0077] Next, step S503 is executed. After receiving the grid dispatch instruction or deciding on the power generation target based on the market price, the multi-objective optimization function is solved to generate optimized control instructions for each unit, including the optimal power setpoint and the drive operation strategy based on the real-time electricity price.

[0078] Finally, step S504 is executed, where the optimized optimal power setpoint and pitch reference commands are sent to each unit's actuators. The actuators then use pitch and torque control to achieve the target power output, forming a closed-loop control system. Simultaneously, the system continuously monitors and performs rolling optimizations based on the latest status predictions and market information.

[0079] In this embodiment, by directly inputting the remaining lifetime prediction results with probability distribution into the multi-objective optimization function of wind farm energy management and combining it with real-time electricity price, dynamic and automated coupling of equipment health status and real-time electricity price is realized, ultimately achieving the goal of maximizing the economic benefits of the entire wind farm throughout its entire life cycle in a complex market environment.

[0080] In one embodiment, to achieve balanced unit lifespan, prevent premature failure of a few units due to overuse, and ensure the absolute maximization of the wind farm's life-cycle economic benefits in a complex market environment, the process of solving the multi-objective optimization function to generate optimized control instructions for each wind turbine includes: Solve the multi-objective optimization function to obtain the real-time health weight factor for each wind turbine, wherein the real-time health weight factor is inversely proportional to the failure risk of the wind turbine; The optimal power setting value is generated based on the health weighting factor.

[0081] Specifically, a multi-objective optimization function is solved to calculate a dynamic, real-time health weighting factor for each wind turbine, which is inversely proportional to the failure risk of its components. When allocating power, turbines with good health and low failure risk are given priority to bear more load, while turbines with poor health and high failure risk are appropriately deloaded to slow their degradation. That is, turbines with good health have a high weighting factor and are given priority to bear more load during power allocation; turbines with poor health have a low weighting factor and are appropriately deloaded to slow their degradation, thus achieving a more balanced lifespan for the turbines.

[0082] In one embodiment, to achieve balanced unit lifespan, prevent premature failure of a few units due to overuse, and ensure the absolute maximization of the wind farm's life-cycle economic benefits in a complex market environment, the real-time electricity price includes high and low prices. Solving the multi-objective optimization function to generate the optimized control instructions corresponding to each wind turbine unit includes: During the high electricity price period, the wind turbine units are controlled to operate at full capacity or exceed capacity. During periods of low electricity prices, the load on high-risk wind turbines is reduced.

[0083] Specifically, the multi-objective optimization function is solved to obtain the health weight of each wind turbine, and a market-driven operation strategy is generated. During periods of high electricity prices: Control strategies tend to maximize instantaneous revenue, allowing healthy wind turbines to operate at full capacity or even moderately over-power within the equipment's safety margin to capture high-value electricity. In this situation, short-term revenue carries more weight than long-term health maintenance. Full capacity operation refers to the wind turbine reaching and stably outputting its rated power under specific wind speed conditions. At this point, the wind speed is typically within the turbine's optimal design range (e.g., 11-15 m / s, approximately force 6-7 winds), resulting in the highest power generation efficiency and constant power output. Over-power operation refers to situations where the power output briefly exceeds the rated value.

[0084] During periods of low / negative electricity prices: Control strategies shift to equipment health maintenance, proactively reducing the load on the entire wind farm or high-risk wind turbines, and allocating "lifespan losses" to high-yield periods. At this time, the weight of long-term health maintenance is higher than that of short-term power generation revenue.

[0085] like Figure 6 As shown, another embodiment of the present invention provides a wind farm energy control method, comprising: Step S601: Obtain full field information, including the RUL probability distribution of each wind turbine, real-time electricity price, grid height command, and wind speed / power prediction; Step S602: Construct a multi-objective optimization function with the core objective of maximizing the total revenue from power generation, risk costs, and maintenance costs; Step S603: Electricity price signal determination; Step S604: With high electricity prices, increase the weight of healthy wind turbine units, allow moderate over-generation, and the strategy is biased towards maximizing returns; Step S605: With low / negative electricity prices, reduce the weight of high-risk wind turbine units, proactively reduce load for maintenance, and prioritize equipment health in the strategy; Step S606: Solve the optimization problem and calculate the optimal power setpoint for each wind turbine. Step S607: Generate control commands, including optimal power setpoint, pitch reference command, and maintenance action warning; Step S608: Send instructions to the local controllers of each wind turbine unit; Step S609: The wind turbine executes control commands to achieve the target power output; Step S610: Monitor the execution effect and status changes; Step S611: Determine whether the real-time electricity price has changed. If it has, proceed to step S601; otherwise, proceed to step S610 to continue monitoring.

[0086] like Figure 7 As shown, an embodiment of the present invention provides a wind farm energy control system, comprising: The acquisition unit 701 is used to acquire the remaining life probability distribution of each wind turbine component, current generating power, external environment information and real-time electricity price, which are predicted by the wind turbine component life probability prediction method described above. The objective optimization function construction unit 702 is used to construct a multi-objective optimization function with the core objective of maximizing the expected total revenue of the wind farm throughout its entire life cycle. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining life probability distribution of the component. The optimized control command generation unit 703 is used to solve the multi-objective optimization function and generate optimized control commands corresponding to each wind turbine. The optimized control commands include the optimal power setpoint and the drive operation strategy based on the real-time electricity price. The instruction issuing unit 704 is used to issue the optimization control instruction.

[0087] In this embodiment, by directly inputting the remaining lifetime prediction results with probability distribution into the multi-objective optimization function of wind farm energy management and combining it with real-time electricity price, dynamic and automated coupling of equipment health status and real-time electricity price is realized, ultimately achieving the goal of maximizing the economic benefits of the entire wind farm throughout its entire life cycle in a complex market environment.

[0088] The above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the lifespan probability of wind turbine components, characterized in that, include: A state-space model of the degradation process of wind turbine components is constructed. The state-space model includes state equations and observation equations. The state equations represent the evolution of degradation state variables under the drive of operating loads, and the observation equations represent the relationship between the degradation state variables and observable health indicators. Obtain the real-time status information of the component; The real-time state information is input into the state space model, and the degradation state of the component is evaluated using a sequential data assimilation algorithm, and the posterior probability distribution of the degradation state is output. Based on the posterior probability distribution, Monte Carlo simulation is used to simulate the future degradation path of the component forward until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

2. The wind turbine component life probability prediction method as described in claim 1, characterized in that, The method of evaluating the degradation state of the component using a sequential data assimilation algorithm includes: Based on the state equation, state transition prediction is performed on the initial particle set to obtain the prior probability distribution of the degenerate state; If a new observation is received, the weight of each predicted particle is calculated according to the observation equation. The weights are normalized to obtain the posterior probability distribution.

3. The wind turbine component life probability prediction method as described in claim 2, characterized in that, The process of normalizing the weights to obtain the posterior probability distribution further includes: The normalized particle set is resampled according to the normalized weights to obtain the target particle set, and the target particle set is used as the initial particle set for the next time step.

4. A wind turbine component lifespan probability prediction system, characterized in that, include: The model building unit is used to build a state-space model of the degradation process of wind turbine components. The state-space model includes state equations and observation equations. The state equations are the evolution law of the degradation state variables under the drive of operating loads, and the observation equations are the relationship between the degradation state variables and observable health indicators. An acquisition unit is used to acquire the real-time status information of the component; The posterior probability distribution output unit is used to input the real-time state information into the state space model, evaluate the degradation state of the component using a sequential data assimilation algorithm, and output the posterior probability distribution of the degradation state. The remaining lifetime probability distribution acquisition unit is used to simulate the future degradation path of the component forward using Monte Carlo simulation based on the posterior probability distribution, until the future degradation path exceeds a preset failure threshold, thereby obtaining the remaining lifetime probability distribution of the component.

5. The wind turbine component lifespan probability prediction system as described in claim 4, characterized in that, The posterior probability distribution output unit is also used for: Based on the state equation, state transition prediction is performed on the initial particle set to obtain the prior probability distribution of the degenerate state; If a new observation is received, the weight of each predicted particle is calculated according to the observation equation. The weights are normalized to obtain the posterior probability distribution.

6. The wind turbine component lifespan probability prediction system as described in claim 5, characterized in that, Also includes: The target particle set acquisition unit is used to resample the normalized particle set according to the normalized weights to obtain the target particle set, and use the target particle set as the initial particle set for the next time step.

7. A wind farm energy control method, characterized in that, include: Obtain the probability distribution of remaining lifespan of each wind turbine component and the real-time electricity price using the wind turbine component lifespan probability prediction method as described in any one of claims 1-3; With the core objective of maximizing the expected total revenue of a wind farm throughout its entire life cycle, a multi-objective optimization function is constructed. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining lifetime probability distribution of the components. Solve the multi-objective optimization function to generate optimized control instructions for each wind turbine, the optimized control instructions including the optimal power setpoint and the drive operation strategy based on the real-time electricity price; Issue the aforementioned optimized control command.

8. The wind farm energy control method as described in claim 7, characterized in that, The process of solving the multi-objective optimization function to generate the optimized control command corresponding to each wind turbine includes: Solve the multi-objective optimization function to obtain the real-time health weight factor for each wind turbine, wherein the real-time health weight factor is inversely proportional to the failure risk of the wind turbine; The optimal power setting value is generated based on the health weighting factor.

9. The wind farm energy control method as described in claim 7, characterized in that, The real-time electricity price includes high electricity price and low electricity price. Solving the multi-objective optimization function to generate optimized control instructions for each wind turbine includes: During the high electricity price period, the wind turbine units are controlled to operate at full capacity or exceed capacity. During periods of low electricity prices, the load on high-risk wind turbines is reduced.

10. A wind farm energy control system, characterized in that, include: The acquisition unit is used to acquire the probability distribution of the remaining lifespan of each wind turbine component and the real-time electricity price, which are predicted by the wind turbine component lifespan probability prediction method as described in any one of claims 1-3. The objective optimization function construction unit is used to construct a multi-objective optimization function with the core objective of maximizing the expected total revenue of the wind farm throughout its entire life cycle. The expected total revenue throughout the entire life cycle is inversely proportional to the expected failure risk cost, and the expected failure risk cost is directly proportional to the real-time failure probability calculated from the remaining life probability distribution of the component. An optimized control command generation unit is used to solve the multi-objective optimization function and generate optimized control commands for each wind turbine. The optimized control commands include the optimal power setpoint and the drive operation strategy based on the real-time electricity price. The instruction issuing unit is used to issue the optimization control instruction.

Images