A performance degradation prediction method based on wind turbine independent pitch control system

Abstract

The present invention relates to the field of wind power generation, and specifically to a performance degradation prediction method based on an independent variable pitch system of a wind turbine, comprising: step 1: using the pitch angle control error as the core degradation characterization quantity, establishing a performance degradation model using the Wiener process, and analyzing the degradation law of key components of the variable pitch system in real time; step 2: defining a state vector including the control error and its changing speed, introducing the deterministic drift and random fluctuation caused by degradation, and constructing a state space equation; step 3: designing an objective function, and dynamically adjusting the weight coefficient according to the degradation state; step 4: implementing model predictive control, adjusting the control strategy in real time according to typical working conditions, and balancing control accuracy and component life extension. By comprehensively analyzing the complex relationship between external loads and the degradation of the variable pitch system, a more accurate performance degradation model is established using the Wiener process, fully considering the impact of external loads on the degradation law of key components.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation, and in particular to a performance degradation prediction method based on an independent pitch control system of a wind turbine. Background Art

[0002] In the field of wind power generation, the pitch system of a wind turbine plays a critical role in ensuring stable operation and efficient power generation. However, long-term operation in complex and harsh environments can lead to increasingly noticeable performance degradation of the pitch system, seriously impacting the efficiency of wind power generation and the lifespan of the turbine. Therefore, a deeper understanding of the degradation mechanisms of the pitch system and the development of appropriate control strategies are key requirements for the development of wind power technology.

[0003] CN117028169A collects data by installing sensors, cameras, and microphones on wind turbines, then transmits the data to cloud servers for processing and storage. Data analysis and machine learning methods are used to establish performance and life prediction models, allowing for status diagnosis and health assessment of wind turbines. Based on the assessment results, the operating parameters of the wind turbines are automatically adjusted. Finally, a graphical interface and voice interaction technology are used to display operating status and provide maintenance recommendations to operation and maintenance personnel. This method not only improves the performance and lifespan of wind turbines but also simplifies operation and maintenance through intelligent means. However, this method carries high data acquisition costs, and data accuracy and resolution are difficult to guarantee.

[0004] CN119209749A proposes a method for coordinated active power control of a wind turbine with optimized pitch system load. This method involves designing a fuzzy controller to update the pitch range limit parameter α and using α and the speed range limit parameter to coordinately control the pitch angle and torque to ensure the turbine accurately tracks the active power command issued by the wind farm and reduce the pitch system load. This method uses the speed and speed variation as input variables of the fuzzy controller, establishes fuzzy rules to dynamically update the pitch range limit parameter α, and further introduces speed range limit parameters b and γ to redefine the speed regulation error. Based on this speed regulation error, active power control coordinated with pitch angle and torque is designed. This method is simple and easy to implement, effectively expanding the range of speed-dependent active power regulation at any pitch angle and reducing the pitch system load. However, this method requires high initial investment, system complexity, data security risks, and technical dependence.

[0005] Faced with the degradation problem of variable pitch systems, traditional control strategies and methods are gradually unable to meet the needs. In the existing technology, although some progress has been made in the monitoring and evaluation of the performance degradation of variable pitch systems, there are still obvious deficiencies in how to closely integrate the degradation process for targeted control to achieve system performance optimization and life extension. For example, some data-driven methods mainly focus on using sensor data for status monitoring and fault diagnosis, fail to fully explore the intrinsic connection between the degradation process and the control strategy, and cannot dynamically adjust the control parameters according to the real-time degradation state. Some optimization control strategies, such as simple load optimization or power regulation methods, do not fully consider the degradation characteristics of the variable pitch system. When dealing with complex and changeable operating conditions and degradation states, it is difficult to achieve ideal control effects. Summary of the Invention

[0006] In response to the problems existing in the prior art, the purpose of the present invention is to provide a performance degradation prediction method based on an independent pitch system of a wind turbine, which is committed to ensuring that the system response characteristics meet the actual operating requirements of the wind turbine, while realizing intelligent optimization control of the pitch system, effectively extending its service life, improving the overall performance and operating stability of the wind turbine, and providing more efficient and reliable technical support for the field of wind power generation.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a performance degradation prediction method based on a wind turbine independent pitch control system, comprising the following steps:

[0008] Step 1: Using the pitch angle control error as the core degradation representation, a performance degradation model is established using the Wiener process. By defining the target pitch angle, actual pitch angle, and control error, a stochastic differential equation with drift and diffusion terms is constructed. After discretization, maximum likelihood-based parameter estimation, and timely updates, the degradation patterns of key components of the variable pitch system are analyzed in real time.

[0009] Step 2: Construct a joint degradation-control model, combining the Wiener process degradation law with the pitch system dynamics. Define a state vector containing the control error and its changing rate, introduce the deterministic drift and random fluctuation caused by degradation, and construct the state space equation.

[0010] Step 3: Design a "life extension-oriented" objective function that integrates the three major requirements of control accuracy, energy consumption, and component load. Dynamically adjust the weight coefficient based on the degradation state. When degradation worsens, the component load weight is increased and the control accuracy weight is reduced.

[0011] Step 4: Implement model predictive control and adopt a rolling optimization strategy. Based on the latest degradation parameters and state estimates, solve the optimal control sequence for the next N steps and implement only the first control variable. Process hard constraints and use Kalman filtering to fuse measured data for feedback correction. Adjust the control strategy in real time according to typical operating conditions to balance control accuracy and component life extension.

[0012] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 1 includes:

[0013] Step 1-1: Using the pitch angle control error as the core degradation representation, by defining the target pitch angle, actual pitch angle, and control error, an input system for the variable pitch system performance degradation model is established. This is used to quantitatively analyze the degradation state of key components and drive the control strategy.

[0014] Step 1-2: Based on the Wiener process, the degradation of the variable pitch system is abstracted as the superposition of deterministic trends and random fluctuations. A continuous-time model is established to describe the dynamic degradation process of the pitch angle control error.

[0015] Steps 1-3: Discretization processing: Based on the continuous model, by setting the sampling period, the Wiener process in the continuous time domain is converted into sequence data at discrete time points, and the actual collected sensor data is used to estimate the model parameters and perform real-time control;

[0016] Steps 1-4: Based on the collected error increments and the joint probability density, the likelihood function is constructed by taking the logarithm, and the drift coefficient and diffusion coefficient are solved based on the degradation of the maximum likelihood method to obtain the average rate of change of the error sequence and the variance of the error increment;

[0017] Steps 1-5: The recursive maximum likelihood method is used to handle non-stationary degradation, the forgetting factor λ is used to weight the historical data, and the parameters are updated online to make the model more sensitive to the current degradation state.

[0018] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 1-1 includes:

[0019] Step a: Define core variables, including the target pitch angle calculated in real time by the unit's main control system based on wind speed and power optimization algorithm , the actual pitch angle measured by the pitch system angle sensor , and the control error e(t), the control error g ;

[0020] Step b: Apply the core variables and use the control error e(t) as the state variable of the Wiener process model. The degradation dynamics are described by the stochastic differential equation SDE: Based on time series data, the deterministic drift μ and random fluctuation σ are estimated using the maximum likelihood method to quantitatively analyze the component degradation rate and volatility;

[0021] Step c: Update the parameters in real time through the forgetting factor λ to enhance the sensitivity to accelerated degradation and obtain its rate of change by taking the derivative of the control error e(t) , forming the state vector x(t) as the input of the degradation-control joint model, which is embedded in the state space equation to drive the control strategy optimization.

[0022] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, steps 1 to 3 include:

[0023] Step d: Discrete time series construction, assuming the sampling period is , discrete moments , corresponding to the error sequence , error increment: ,in are independent normal random variables;

[0024] Step e: Through the incremental property of the Wiener process , indicating that Brownian motion in any time interval The increment on Obey the normal distribution derivation logic with mean 0 and variance Δt, Δt represents the length of the time interval, by combining the continuous diffusion coefficient σ with Δt and multiplying it by the standard normal distribution random variable , obtain the diffusion term discretization , using the variance formula The fluctuation characteristics of the error increment in the discrete model are constrained to ensure that the discretized model is consistent with the continuous-time model in describing the randomness of the degradation process, where Δek represents the error increment after discretization.

[0025] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 2 includes:

[0026] Step 2-1: Define the state vector, take over the output of the Wiener process, integrate the degradation and control input, simplify the variable pitch system into a second-order linear system, and establish the basic dynamic model without degradation based on the core dynamic characteristics determined by inertia-damping-control gain. , where d represents the damping coefficient, k represents the control gain, represents the pitch angular acceleration, represents the damping term, u(t) represents the control voltage of the pitch motor;

[0027] Step 2-2: Define the derivative of the control error and the acceleration of the target angle ,get , substituting the non-degenerate dynamic equation into it, we get: ;

[0028] Step 2-3: Deterministic drift μ and process noise due to degradation Introducing random perturbation σ and embedding the degradation term into the error dynamic equation, we obtain: , complete the fusion of Wiener process degradation terms;

[0029] Step 2-4: Construct the state space equation and define the state vector , then the state derivative is: , the standard linear stochastic differential equation SDE form is: , where A is the system matrix, B is the control matrix, and η(t) is the degraded input vector.

[0030] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 3 includes:

[0031] Step 3-1: Control target decomposition, including control accuracy λ1 for minimizing the square integral of the error and ensuring power tracking accuracy, energy consumption λ2 for suppressing excessive control input, and component load λ3 for penalizing the pitch angular velocity. , reducing dynamic loads on bearings and / or gears;

[0032] Step 3-2: Dynamically adjust the weights of the control degradation process. When the Wiener process parameters show that the degradation is getting worse, automatically increase λ3 to reduce the pitch speed and / or reduce λ1, allowing a slight decrease in accuracy in exchange for an increase in life.

[0033] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 4 includes:

[0034] Step 4-1: Rolling optimization strategy, at the current moment , based on the latest degradation parameters and state estimation , solve the optimal control sequence for the next N steps , only the first control quantity is implemented , perform optimization problem and state transfer, the optimization problem is: , the state transition is: using the state space equation to predict the future state: ,in is the random noise generated by Monte Carlo simulation to simulate the uncertainty of degradation;

[0035] Step 4-2: Perform hard constraint processing while avoiding physical damage, including limiting the pitch angular velocity, mechanically limiting the pitch angle range, and controlling the motor voltage input saturation;

[0036] Step 4-3: Feedback correction, sensor noise and model error, real-time update of state estimation through Kalman filter: , the corrected state estimate As the initial condition for the next cycle optimization;

[0037] Step 4-4: Deep coupling of the degradation model and the control model. Within the time of executing a rolling optimization strategy, a closed loop of data collection → degradation modeling → control calculation → execution is completed, responding to degradation changes in real time.

[0038] In the above-mentioned performance degradation prediction method based on the wind turbine independent pitch control system, step 4-4 includes:

[0039] When the wind speed is stable, the control goal is to prioritize the control accuracy λ1=1.0 and allow the medium pitch speed component load λ3=0.5. The control effect at this time is: the motor torque fluctuation is small and the error is stable at ±0.2°;

[0040] When the wind speed changes drastically and bearing wear occurs, the control target is to switch to life extension mode, set component load λ3 to 1.5, and limit the pitch speed to ≤3° / s. The control effect at this time is: the error temporarily expands to ±0.5°, but the bearing load is reduced by 30%, and the life is extended by 25%.

[0041] The beneficial effects of the present invention's performance degradation prediction method for an independent wind turbine pitch system include: focusing on optimizing control strategies during the degradation of key pitch system components under the dynamic effects of external loads. By comprehensively analyzing the complex relationship between external loads and pitch system degradation, a more accurate performance degradation model is established using the Wiener process, fully accounting for the impact of external loads on the degradation patterns of key components. On this basis, a joint degradation-control model is constructed, combined with a model predictive control algorithm, to dynamically adjust the control strategy based on the real-time degradation state and external load conditions.

[0042] The present invention uses the Wiener process to establish a performance degradation model for the variable pitch system, capturing in real time the deterministic degradation trends (drift terms) and random fluctuations (diffusion terms) of key components such as bearing wear and gear damage, and embedding them into the control model. This allows the control algorithm to be dynamically adjusted according to the real-time degradation status, achieving a balance between "control accuracy" and "component life extension", filling the technical gap in the coordinated optimization of degradation processes and control strategies.

[0043] Dynamic Weight Adjustment Mechanism for the Objective Function: A "life extension-oriented" objective function is designed, integrating the three key objectives of control accuracy, energy consumption, and component load. Fuzzy logic or neural networks are used to establish a mapping relationship between degradation parameters (μ / σ) and weight coefficients. As degradation intensifies, the component load weight is automatically increased (for example, λ3 from 0.5 to 1.5), while the control accuracy weight is reduced (for example, λ1 from 1.0 to 0.8). The pitch speed is prioritized to reduce dynamic bearing and gear loads (for every 10% increase in load fluctuation, life is shortened by 20%). This significantly extends component life while maintaining basic system response characteristics, surpassing the limitations of fixed-weight or single-objective optimization in existing technologies.

[0044] Efficient application of model predictive control (MPC): Utilizing a rolling optimization strategy (executed every 100ms), this system predicts future control sequences based on the latest degradation parameters and state estimates, implementing only the first control variable. Kalman filtering is combined with other techniques to address sensor noise and model errors, enabling robust control in complex wind conditions and degradation processes. Compared to existing approaches that rely on cloud-based data processing or fuzzy rules, this system significantly improves control timeliness and accuracy through a real-time closed-loop process (data acquisition → degradation modeling → control calculation → execution). This system is particularly effective in addressing nonlinear degradation and sudden load changes in variable pitch systems, particularly in harsh offshore environments such as high humidity and salt spray. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the overall technical solution flow of an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the performance degradation model process according to an embodiment of the present invention;

[0047] Figure 3 Schematic diagram of the pitch angle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is described below in conjunction with specific implementation methods and drawings.

[0049] Example 1

[0050] The wind turbine pitch system is a key component to ensure efficient energy acquisition. Under complex working conditions, the performance of the core components of the pitch system (such as bearings) will gradually degrade, making the original pitch angle no longer the optimal pitch angle (the pitch angle diagram is shown in the figure). Figure 3This can lead to two problems. First, when wind speeds exceed the rated wind speed of the wind turbine, the turbine generates excessive power, causing damage. Second, the wind turbine load is unbalanced. During operation, uneven wind speed distribution, changes in wind direction, and other factors can cause the blades to bear unbalanced aerodynamic loads, leading to a rapid decrease in their lifespan or performance, and potentially causing various accidents.

[0051] During operation, the degradation of a pitch system is a complex process driven by the combined effects of multiple factors. From an environmental perspective, offshore wind turbine pitch systems are subject to long-term exposure to high humidity and salt spray, which can lead to corrosion of the pitch bearings, degradation of lubricant performance, and seal failure, accelerating system performance degradation. In cold regions, low temperatures significantly increase lubricant viscosity, making metal materials brittle and affecting component function. High temperatures, on the other hand, thin the lubricant and cause electrical components to overheat, similarly leading to performance issues. From a mechanical component perspective, pitch bearings, motors, and gearboxes are subject to long-term pulsating loads, making them susceptible to fatigue cracks, increased wear, and insufficient lubrication. For example, wear on pitch bearings can increase transmission clearance, causing the actual pitch angle to lag behind the target pitch angle. From an electrical system perspective, components such as motors, drives, and batteries can degrade due to aging, overload, and high temperatures. This can even cause the emergency retraction function to fail, posing a serious safety hazard to the turbine. In addition, unreasonable control strategies and parameter settings, such as incorrect overspeed protection parameters, improper pitch control gain, insufficient maintenance, and early design defects are also important factors that lead to the degradation of pitch system performance.

[0052] As the performance of the variable pitch system continues to deteriorate, a series of problems have ensued. In terms of power generation efficiency, the decline in the performance of the variable pitch system makes it impossible for the blades to be accurately adjusted to the optimal pitch angle, making it difficult to effectively capture wind energy. In low wind speed areas, the response delay of the variable pitch system will cause the blades to miss the optimal power generation opportunity; at high wind speeds, failure of the variable pitch system may cause the unit to shut down due to overload protection, greatly reducing the power generation time. In terms of mechanical load and fatigue damage, due to the inability to effectively balance the loads between blades, complex wind conditions such as wind shear, tower shadow effect and turbulence will cause key components such as blades, hubs, main shafts and towers to bear uneven mechanical stress, accelerating fatigue damage of components, shortening their service life, and increasing maintenance costs and downtime.

[0053] To this end, this embodiment proposes an intelligent optimization control method and strategy for the performance degradation process of the wind turbine pitch system. The flow chart is as follows: Figure 1As shown in the figure, the degradation patterns of key components of a variable pitch system are analyzed and modeled using the Wiener process. Based on these patterns, a control model for system life extension is constructed. A model predictive control algorithm is then used to intelligently optimize its operation. First, the Wiener process is used to establish a performance degradation model, with the pitch angle control error as the core degradation indicator. By defining the target pitch angle, actual pitch angle, and control error, a stochastic differential equation with drift and diffusion terms is constructed. Through discretization, maximum likelihood-based parameter estimation, and timely updating, the degradation patterns of key components are analyzed in real time. Next, a joint degradation-control model is constructed, combining the Wiener process degradation pattern with the dynamics of the variable pitch system. A state vector containing the control error and its rate of change is defined, and the deterministic drift and random fluctuations caused by degradation are introduced to construct a state-space equation. A "life extension-oriented" objective function is designed, integrating the three key requirements of control accuracy, energy consumption, and component load. The weight coefficients are dynamically adjusted based on the degradation state, increasing the component load weight and decreasing the control accuracy weight as degradation worsens. Finally, model predictive control is implemented, and a rolling optimization strategy is adopted. Every 100ms, the optimal control sequence for the next N steps is solved based on the latest degradation parameters and state estimation, and only the first control quantity is implemented. Hard constraints such as variable pitch angular velocity, pitch angle range, and control input are processed, and Kalman filtering is used to fuse measured data for feedback correction, forming a closed loop of "data acquisition → degradation modeling → control calculation → execution". The control strategy is adjusted in real time according to typical operating conditions such as stable wind speed, sudden change in wind speed + bearing wear, to achieve a balance between "control accuracy" and "component life extension".

[0054] The specific plan is as follows:

[0055] Step 1: Analyze the degradation patterns of key components of the pitch system.

[0056] First, the performance degradation model of the wind turbine pitch system is established using the Wiener process (structure as Figure 3 (as shown in Figure 2), this model describes the degradation patterns of key components (such as bearings and gears). The key principle is to establish a stochastic differential equation for the pitch angle control error based on the Wiener process. The drift term characterizes the deterministic degradation trend, while the diffusion term characterizes random fluctuations. This allows for real-time monitoring of component degradation and analysis of the decline patterns of key components in the variable pitch system.

[0057] The process is as follows:

[0058] 1. Definition of state variables: The pitch angle control error is used as the core degradation representation.

[0059] (1) Core variable definition.

[0060] Target pitch angle: , calculated in real time by the unit's main control system based on wind speed and power optimization algorithm, unit: ° Actual pitch angle: , measured by the pitch system angle sensor (such as an absolute encoder), with an accuracy of ±0.1°, unit: °.

[0061] Control error: , directly reflects the tracking accuracy of the pitch system and is the comprehensive output of multi-component degradation.

[0062] (2) Application of core variables.

[0063] Input for degradation modeling: e(t) is used as the state variable of the Wiener process model, and its degradation dynamics is described by stochastic differential equations (SDE): .

[0064] Parameter estimation and updating: Based on time series data, the maximum likelihood method is used to estimate μ and σ, and quantitatively analyze the component degradation rate and volatility.

[0065] The parameters are updated in real time through the forgetting factor λ, which enhances the sensitivity to accelerated degradation (such as bearing crack growth).

[0066] Inputs to the joint degradation-control model: e(t) and its rate of change Construct the state vector x(t) and embed it into the state space equation to drive the control strategy optimization.

[0067] (3) The role of the physical meaning of core variables and their degradation association.

[0068] Physical meaning:

[0069] e(t) mean drift: reflects systematic degradation, such as bearing wear leading to increased transmission clearance and actual angle lag.

[0070] Expansion of the e(t) variance: reflects random degradation, such as periodic meshing errors caused by gear damage.

[0071] Abnormal dynamic characteristics: Motor torque attenuation causes response delay, which manifests as cumulative error in e(t).

[0072] 2. Establish a continuous-time model: stochastic differential equation SDE based on Wiener process.

[0073] Model assumptions and construction.

[0074] The degradation of the pitch system meets the following conditions:

[0075] Independent incrementality: The error changes are independent of each other.

[0076] Normal distribution characteristics: Error changes in a short period of time obey .

[0077] Construct a Wiener process with drift: .

[0078] Drift term : Deterministic degradation trend, unit: ° / s, reflecting the average rate of component wear accumulation.

[0079] Diffusion term : Random degradation fluctuation, unit: °, reflecting random factors such as load mutation and measurement noise.

[0080] 3. Discrete processing: from continuous model to data-driven form.

[0081] Discretization involves converting a random process (Wiener process) in the continuous time domain into a sequence of data at discrete time points. This allows for model parameter estimation and real-time control using actual sensor data (such as a time series of pitch angle error). The core of this process is to convert continuous error changes into a discrete sequence of error increments by setting a sampling period based on a continuous model.

[0082] (1) Discrete time series construction.

[0083] Assume the sampling period is (such as 0.1s), discrete moments , corresponding to the error sequence , error increment: in are independent normal random variables.

[0084] (2) Deductive logic.

[0085] Using the incremental properties of the Wiener process: .

[0086] Discretization of diffusion term , maintaining variance consistency: .

[0087] 4. Parameter estimation: Degenerate parameter solution based on maximum likelihood method.

[0088] (1) Likelihood function construction.

[0089] N error increments are collected ,

[0090] Its joint probability density is: ,

[0091] Take the log-likelihood function: .

[0092] (2) Parameter solution process.

[0093] Calculate the drift coefficient :right Taking the derivative and setting it to zero, we get: ,

[0094] Physical meaning: average rate of change of error sequence,

[0095] Find the diffusion coefficient :right Taking the derivative and setting it to zero, we get: .

[0096] Physical meaning: The variance of the error increment reflects the uncertainty of the degradation process.

[0097] 5. Time-varying update: Recursive maximum likelihood method to deal with non-stationary degradation.

[0098] When component degradation accelerates (such as bearing crack growth), and As time goes by, parameters need to be updated online: ,

[0099] , where λ is the forgetting factor (0 < λ ≤ 1). The forgetting factor is a coefficient that weights historical data in parameter estimation. Its core idea is to give more weight to recent data and less weight to older data, allowing the model to more sensitively track the current degradation state and avoid interference from earlier healthy data on real-time degradation characteristics.

[0100] Step 2: Construct a joint degradation-control model.

[0101] The core of the degradation-control model is to combine the degradation law (μ / σ) described by the Wiener process with the dynamics of the pitch system. This allows the control algorithm to dynamically adjust its strategy based on the current degradation state, achieving a balance between "control accuracy" and "component life extension." Next, completing model predictive control for pitch system performance degradation requires three steps: state-space modeling → objective function design → model predictive control (MPC) implementation. Each step is closely dependent on the degradation parameters output by the Wiener process.

[0102] The process of step 2 is as follows.

[0103] 1. Definition of state vector (following the output of Wiener process).

[0104] Core status variables:

[0105] The state variables directly from the Wiener process (target and actual angle deviation),

[0106] Reflects the error change rate and is used to characterize the dynamic response characteristics of the system (such as the decrease in the rate of change when the motor torque is insufficient).

[0107] 2. Construction of dynamic equations (integration of degradation and control input).

[0108] First, the basic dynamic model (system characteristics without degradation) is established. The pitch system can be simplified to a second-order linear system, whose core dynamic characteristics are determined by inertia-damping-control gain. Ignoring higher-order nonlinear factors (such as gear clearance nonlinearity and motor saturation characteristics), the dynamic equation without degradation is: .

[0109] Where d is the damping coefficient, which is the decay rate caused by bearing or gear friction.

[0110] k is the control gain, which is the pitch acceleration corresponding to unit voltage.

[0111] : pitch angular acceleration (unit: ° / s²), reflecting the system inertia, : Damping term (unit: ° / s), including transmission chain friction damping, aerodynamic damping, etc. : control input (unit: ° / s²), k is the motor control gain (reflecting the conversion efficiency from voltage to acceleration), and u(t) is the pitch motor control voltage.

[0112] 3. Introduce the state space description of the control error.

[0113] Defining Control Error , whose derivative is , the acceleration of the target angle (The main control system smoothly outputs the target angle), then: .

[0114] Substituting the non-degenerate dynamics equation into the equation, we obtain: , after sorting, we get: .

[0115] 4. Integrate the Wiener process degradation term.

[0116] Considering the deterministic drift (μ) and random fluctuation (σ) caused by degradation, the drift term μ reflects the error change rate trend caused by component wear, which is equivalent to a constant offset of the error change rate, such as the bearing clearance causing Average increase μ, diffusion term σ: through process noise Introduce random perturbations.

[0117] By embedding the degradation term into the error dynamic equation, we obtain: , physical meaning: Degradation causes the system to produce an additional average drift μ and random fluctuation w (t) in addition to the control input.

[0118] 5. Construct the state space equation.

[0119] Define the state vector ,

[0120] Then the state derivative is: ,

[0121] It can be simplified into the standard linear stochastic differential equation SDE form:

[0122] , where A is the system matrix, B is the control matrix, and η(t) is the degraded input vector.

[0123] Step 3: After the state space equation is established, design the “life extension-oriented” objective function.

[0124] 1. Control target decomposition, including the three core demands of control accuracy λ1, energy consumption λ2, and component load λ3.

[0125] Control accuracy (λ1): Minimizes the squared integral of the error to ensure power tracking accuracy. An error exceeding 0.5° will result in a 1% drop in power generation efficiency. Energy consumption (λ2): Suppresses excessive control inputs, such as excessive motor voltage that increases energy consumption and heat generation.

[0126] Component load (λ3): The core of the core, by penalizing the pitch angular velocity ,Right now , reduce the dynamic load of bearings and / or gears. For every 10% increase in load fluctuation, the life is shortened by 20%.

[0127] 2. Dynamic adjustment of weights based on degradation state.

[0128] When the Wiener process parameters show increasing degradation, such as μ↑ or σ↑:

[0129] Automatically increase λ3, for example, from 0.5 to 1.0, and give priority to reducing the pitch speed.

[0130] Appropriately reducing λ1, such as from 1.0 to 0.8, allows for a slight decrease in accuracy in exchange for improved life.

[0131] Engineering implementation: Through fuzzy logic or neural network, a mapping table of "μ / σ→weight coefficient" is established, as shown in Table 1.

[0132] Table 1: μ / σ→weight coefficient mapping table

[0133] .

[0134] Step 4: Model predictive control MPC implementation - rolling optimization and constraint processing.

[0135] 1. Rolling optimization strategy, executed every 100ms.

[0136] At the current moment , based on the latest degradation parameters and state estimation , solve the optimal control sequence for the next N steps, such as N=10 , only the first control quantity is implemented .

[0137] Optimization problem: .

[0138] State transition: Use state space equations to predict future states: ,in is the random noise generated by Monte Carlo simulation to simulate the uncertainty of degradation.

[0139] 2. Hard constraint processing to avoid physical damage.

[0140] Pitch angular velocity limit: , such as -15° / s ≤ ≤ +15° / s.

[0141] Pitch angle range: , use mechanical limit.

[0142] Control input saturation: , motor voltage safety range.

[0143] 3. Feedback correction: Kalman filter integrates measured data.

[0144] Due to sensor noise, such as the angle sensor accuracy of ±0.1° and model errors, the state estimate needs to be updated in real time through the Kalman filter: .

[0145] Corrected state estimate It is used as the initial condition for the next optimization cycle to ensure control robustness.

[0146] 4. Deep coupling of degradation model and control model.

[0147] (1) Data flow is closed loop.

[0148] The closed loop of "data collection → degradation modeling → control calculation → execution" is completed every 100ms, responding to degradation changes in real time.

[0149] (2) Strategy adjustment under typical working conditions.

[0150] Condition 1: Stable wind speed, early degradation stage, μ=0.01° / s, σ=0.1°.

[0151] Control objective: Prioritize the control accuracy λ1=1.0 and allow the medium pitch speed component load λ3=0.5.

[0152] Control effect: The error is stable at ±0.2°, and the motor torque fluctuation is small.

[0153] Working condition 2: sudden change in wind speed + bearing wear, μ=0.06° / s, σ=0.4°.

[0154] Control objective: Switch to "life extension mode", component load λ3=1.5, and limit pitch speed ≤3° / s.

[0155] Control effect: The error is temporarily expanded to ±0.5°, but the bearing load is reduced by 30% and the life is extended by 25%.

[0156] Compared with the prior art, the advantages of the present invention are mainly reflected in the following aspects.

[0157] Deep coupling of degradation modeling and control strategies: Existing technologies often focus on state diagnosis or load optimization, without directly integrating component degradation patterns with control strategies. This invention utilizes the Wiener process to establish a performance degradation model for the variable pitch system. This model captures the deterministic degradation trends (drift terms) and random fluctuations (diffusion terms) of key components, such as bearing wear and gear damage, in real time. This information is embedded in the control model, enabling the control algorithm to dynamically adjust based on the real-time degradation state, achieving a balance between control accuracy and component life extension. This approach fills a gap in the collaborative optimization of degradation processes and control strategies.

[0158] Dynamic weight adjustment mechanism for the objective function: A "life extension-oriented" objective function is designed, integrating the three major objectives of control accuracy, energy consumption, and component load. A mapping relationship between degradation parameters (μ / σ) and weight coefficients is established through fuzzy logic or neural networks. When degradation intensifies, the component load weight is automatically increased (for example, λ3 increases from 0.5 to 1.5), while the control accuracy weight is reduced (for example, λ1 decreases from 1.0 to 0.8). The pitch speed is preferentially reduced to reduce dynamic bearing and gear loads (for every 10% increase in load fluctuation, life is shortened by 20%). This significantly extends component life while maintaining basic system response characteristics, surpassing the limitations of fixed weights or single-objective optimization in existing technologies.

[0159] Efficient application of model predictive control (MPC): Utilizing a rolling optimization strategy (executed every 100ms), this system predicts future control sequences based on the latest degradation parameters and state estimates, implementing only the first control variable. Kalman filtering is combined with other techniques to address sensor noise and model errors, enabling robust control for complex wind conditions and degradation processes. Compared to existing approaches that rely on cloud-based data processing or fuzzy rules, this system significantly improves control timeliness and accuracy through a real-time closed-loop process (data acquisition → degradation modeling → control calculation → execution). This system is particularly effective in addressing nonlinear degradation and sudden load changes in variable pitch systems, particularly in harsh offshore environments such as high humidity and salt spray.

[0160] Cost and Benefit Optimization: For independent variable pitch systems, precise adjustment of individual blade pitch angles balances aerodynamic loads, reducing fatigue damage to key components like the hub and main shaft, and lowering maintenance costs and downtime. Leveraging angle sensors (±0.1° accuracy) and the proven Wiener process and MPC algorithms, this system significantly reduces data acquisition costs and system complexity while maintaining control performance, delivering both practical engineering and economic efficiency.

[0161] Improved Adaptability and Reliability: A time-varying update mechanism handles non-stationary degradation (such as accelerated degradation caused by bearing crack propagation), updating model parameters in real time. This allows the control strategy to dynamically adapt to different degradation stages (healthy, mildly degraded, severely degraded) and typical operating conditions (stable wind speed, drastic wind speed fluctuations and bearing wear). For example, in severely degraded conditions, limiting the pitch speed to ≤3° / s can reduce bearing load by 30% and extend bearing life by 25%. This effectively addresses the load imbalance problem faced by traditional unified pitch systems in complex wind conditions, improving unit operational stability and safety.

[0162] Example 2

[0163] This embodiment addresses the performance degradation problem of wind turbine pitch control systems and proposes an optimization strategy that integrates degradation modeling and intelligent control.

[0164] First, a performance degradation model is established using the Wiener process, with pitch angle control error as the core degradation parameter. By defining the target pitch angle, actual pitch angle, and control error, a stochastic differential equation with drift and diffusion terms is constructed. Combining discretization, maximum likelihood parameter estimation, and a time-varying update mechanism, this model captures the degradation patterns of key components (such as bearings, gears, and motors) in real time, addressing the issue of insufficient dynamic utilization of degradation processes in existing technologies. Secondly, a joint degradation-control model is constructed, embedding the degradation parameters (μ / σ) output by the Wiener process into the dynamics of the pitch system. The deterministic drift and random fluctuations caused by degradation are integrated through state-space equations. A "life extension-oriented" objective function is designed, dynamically adjusting the weighting coefficients of control accuracy, energy consumption, and component load to achieve a balance between "control accuracy" and "component life extension." In particular, when degradation intensifies, the pitch speed is prioritized to reduce load. Finally, model predictive control (MPC) is implemented. Through a rolling optimization strategy, hard constraint processing, and Kalman filter feedback correction, a real-time closed loop of "data acquisition → degradation modeling → control calculation → execution" is formed, dynamically adjusting the strategy based on operating conditions. This embodiment integrates a degradation modeling method based on the Wiener process, the construction of a joint degradation-control model, a dynamic weight adjustment mechanism, and a model predictive control algorithm. Its core approach is to optimize the control strategy using degradation laws to extend system life. This approach differs from existing approaches that are simply data-driven or load-based, filling a gap in the deep coupling of degradation processes and control strategies.

[0165] like Figure 1-Figure 2 As shown, a performance degradation prediction method based on a wind turbine independent pitch control system includes the following steps.

[0166] Step 1: Taking the pitch angle control error as the core degradation characterization quantity, the Wiener process is used to establish a performance degradation model. By defining the target pitch angle, actual pitch angle and control error, a stochastic differential equation containing drift and diffusion terms is constructed. After discretization processing, parameter estimation based on the maximum likelihood method and timely update, the degradation law of the key components of the variable pitch system is analyzed in real time.

[0167] include:

[0168] Step 1-1: Taking the pitch angle control error as the core degradation characterization quantity, by defining the target pitch angle, actual pitch angle and control error, establish the input system of the variable pitch system performance degradation model, which is used for quantitative analysis of the degradation state of key components and driving the control strategy.

[0169] include:

[0170] Step a: Define core variables, including the target pitch angle calculated in real time by the unit's main control system based on wind speed and power optimization algorithm , the actual pitch angle measured by the pitch system angle sensor , and the control error e(t), the control error .

[0171] Step b: Apply the core variables and use the control error e(t) as the state variable of the Wiener process model. The degradation dynamics are described by the stochastic differential equation SDE: ,Based on time series data, the deterministic drift μ and random fluctuation σ are estimated using the maximum likelihood method to quantitatively analyze the component degradation rate and volatility.

[0172] Deterministic drift μ (drift term) and random fluctuation σ (diffusion term) are the core parameters of the Wiener process model, characterizing the deterministic trend and random fluctuation of the degradation of key pitch system components, respectively. The drift term μ is the "average trend" of the degradation process, revealing the cumulative effects of component wear and used to predict the degradation path. The diffusion term σ is the "uncertainty bound" of the degradation process, reflecting the interference intensity of random factors and used to assess degradation risk.

[0173] Step c: Update parameters in real time through the forgetting factor λ to enhance the sensitivity to accelerated degradation and use the control error e(t) and its rate of change The state vector x(t) is constructed as the input of the degradation-control joint model and embedded in the state-space equations to drive the control strategy optimization.

[0174] The rate of change of the control error is obtained by differentiating the control error. Its physical meaning is the speed at which the error changes over time, reflecting the dynamic response characteristics of the variable pitch system.

[0175] Step 1-2: Based on the Wiener process, the degradation of the variable pitch system is abstracted as the superposition of deterministic trends and random fluctuations. A continuous-time model is established to describe the dynamic degradation process of the pitch angle control error.

[0176] The continuous-time model, based on the Wiener process, describes the dynamic degradation of pitch angle control error. This model abstracts the degradation of the variable pitch system as a superposition of deterministic trends and random fluctuations, providing a mathematical framework for real-time monitoring of component degradation and analyzing its patterns. This model not only serves as the foundation for subsequent discretization and parameter estimation but also serves as the core link between degradation modeling and control strategies, ensuring that the control algorithm dynamically adjusts to the real-time degradation state, achieving a balance between control accuracy and component lifespan.

[0177] Model assumptions and construction.

[0178] The pitch system degradation satisfies:

[0179] Independent incrementality: The error changes of are independent of each other.

[0180] Normal distribution characteristics: Error changes in a short period of time obey ,

[0181] Construct a Wiener process with drift: ,

[0182] Drift term : Deterministic degradation trend, unit: ° / s, reflecting the average rate of component wear accumulation,

[0183] Diffusion term : Random degradation fluctuation, unit: °, reflecting random factors such as load mutation and measurement noise,

[0184] Steps 1-3: Discretization processing: Based on the continuous model, by setting the sampling period, the Wiener process in the continuous time domain is converted into sequence data at discrete time points, and the actual collected sensor data is used to estimate the model parameters and perform real-time control.

[0185] include:

[0186] Step d: Discrete time series construction, assuming the sampling period is , discrete moments , corresponding to the error sequence , error increment: ,in are independent normal random variables.

[0187] Step e: Through the incremental property of the Wiener process , indicating that Brownian motion in any time interval The increment on Obey the normal distribution derivation logic with mean 0 and variance Δt, Δt represents the length of the time interval, by combining the continuous diffusion coefficient σ with Δt and multiplying it by the standard normal distribution random variable , obtain the diffusion term discretization , using the variance formula The fluctuation characteristics of the error increment in the discrete model are constrained to ensure that the discretized model is consistent with the continuous-time model in describing the randomness of the degradation process, where Δek represents the error increment after discretization.

[0188] Step f: Through the incremental properties of the Wiener process and the variance consistency constraint, an unbiased conversion of the diffusion term of the continuous model to the discrete model is achieved. The core is to ensure that the random characteristics of the degradation process remain unchanged after discretization. It shows that Brownian motion in any time interval The increment on Obey the normal distribution with mean 0 and variance Δt (i.e. the length of the time interval). Diffusion term discretization , the diffusion term in the continuous time model is discretized by combining the continuous diffusion coefficient σ with the sampling period Δt and multiplying it by the standard normal distribution random variable , to simulate the random fluctuations of the degradation process at discrete time points, ensuring that the discretized model can reasonably reflect the random factors and is consistent with the continuous model in terms of statistical characteristics such as variance. Var(): is the mathematical symbol of variance, which represents the degree of discreteness of a random variable and is used to measure the degree of deviation of a set of data from its mean. The variance of the discretized error increment Δek is equal to the product of the square of the diffusion coefficient σ2 in the continuous-time model and the sampling period Δt.

[0189] Its significance lies in ensuring that the discretized model is consistent with the continuous-time model in describing the randomness of the degradation process. This equation is used to constrain the fluctuation characteristics of the error increment in the discrete model, so that the degradation model constructed based on discrete data can accurately reflect the influence of random factors, providing a reliable basis for subsequent parameter estimation and control strategy formulation.

[0190] Steps 1-4: Based on the collected error increments and joint probability density, take the logarithm to construct the likelihood function, solve the drift coefficient and diffusion coefficient based on the degradation of the maximum likelihood method, and obtain the average change rate of the error sequence and the variance of the error increment.

[0191] Steps 1-5: The recursive maximum likelihood method is used to handle non-stationary degradation, the forgetting factor λ is used to weight the historical data, and the parameters are updated online to make the model more sensitive to the current degradation state.

[0192] Step 2: Construct a degradation-control joint model, combining the Wiener process degradation law with the pitch system dynamics, defining a state vector containing the control error and its changing rate, introducing the deterministic drift and random fluctuation caused by degradation, and constructing the state space equation.

[0193] include:

[0194] Step 2-1: Define the state vector, take over the output of the Wiener process, integrate the degradation and control input, simplify the variable pitch system into a second-order linear system, and establish the basic dynamic model without degradation based on the core dynamic characteristics determined by inertia-damping-control gain. , where d represents the damping coefficient, k represents the control gain, represents the pitch angular acceleration, represents the damping term, and u(t) represents the control voltage of the pitch motor.

[0195] Step 2-2: Define the derivative of the control error and the acceleration of the target angle ,get , substituting the non-degenerate dynamic equation into it, we get: .

[0196] To ensure stable and smooth wind turbine operation, the main control system typically outputs a target pitch angle that does not experience abrupt changes. Based on current operating conditions (such as wind speed), the target pitch angle is adjusted relatively smoothly and continuously to avoid sudden changes that could impact turbine components. The target angle acceleration is approximately zero. Secondly, in practice, performance degradation of the variable pitch system (such as bearing wear and gear damage) and the dynamic changes in control errors are key areas of research. Setting the target angle acceleration to zero simplifies the mathematical model, reducing unnecessary variables and complexity. This allows for a more focused study of the errors and their variations caused by system degradation and control response as the actual pitch angle tracks the target pitch angle. This highlights the key influencing factors and facilitates subsequent model derivation, parameter estimation, and control strategy design. The non-degradation dynamic equation describes the system's motion under ideal conditions, ignoring performance degradation factors such as component wear and aging. This equation takes into account the reality that actual variable pitch systems exhibit control errors, which are related to the target pitch angle, the actual pitch angle, and its derivatives. Substituting the non-degenerate dynamics equations into the equations is intended to combine the idealized laws of motion with factors such as actual control errors. This results in an equation that better reflects actual conditions, accounting for both the system's inherent laws of motion and the errors introduced by various factors.

[0197] Step 2-3: Deterministic drift μ and process noise due to degradation Introducing random perturbation σ and embedding the degradation term into the error dynamic equation, we obtain: , completing the fusion of the Wiener process degradation term.

[0198] Step 2-4: Construct the state space equation and define the state vector , then the state derivative is: , the standard linear stochastic differential equation SDE form is: , where A is the system matrix, B is the control matrix, and η(t) is the degraded input vector.

[0199] Step 3: Design a "life extension-oriented" objective function that integrates the three major demands of control accuracy, energy consumption, and component load. Dynamically adjust the weight coefficient according to the degradation state. When the degradation worsens, increase the component load weight and reduce the control accuracy weight.

[0200] include:

[0201] Step 3-1: Control target decomposition, including control accuracy λ1 for minimizing the square integral of the error and ensuring power tracking accuracy, energy consumption λ2 for suppressing excessive control input, and component load λ3 for penalizing the pitch angular velocity. , reducing dynamic loads on bearings and / or gears.

[0202] Step 3-2: Control the dynamic adjustment of the weights during the degradation process. When the Wiener process parameters show that the degradation is intensifying, automatically increase λ3 to preferentially reduce the pitch speed and / or appropriately reduce λ1, allowing a slight decrease in accuracy in exchange for an increase in life.

[0203] Step 4: Implement model predictive control and adopt a rolling optimization strategy. Based on the latest degradation parameters and state estimates, solve the optimal control sequence for the next N steps and implement only the first control variable. Process hard constraints and use Kalman filtering to fuse measured data for feedback correction. Adjust the control strategy in real time according to typical operating conditions to balance control accuracy and component life extension.

[0204] include:

[0205] Step 4-1: Rolling optimization strategy, at the current moment , based on the latest degradation parameters and state estimation , solve the optimal control sequence for the next N steps , only the first control quantity is implemented , perform optimization problem and state transfer, the optimization problem is: , the state transition is: using the state space equation to predict the future state: ,in is the random noise generated by Monte Carlo simulation to simulate the uncertainty of degradation.

[0206] Step 4-2: Perform hard constraint processing while avoiding physical damage, including limiting the pitch angular velocity, mechanically limiting the pitch angle range, and controlling the motor voltage input saturation.

[0207] Step 4-3: Feedback correction, sensor noise and model error, real-time update of state estimation through Kalman filter: , the corrected state estimate As the initial condition for the next cycle optimization.

[0208] Step 4-4: Deep coupling of the degradation model and the control model. Within the time of executing a rolling optimization strategy, a closed loop of data collection → degradation modeling → control calculation → execution is completed, responding to degradation changes in real time.

[0209] include:

[0210] When the wind speed is stable, the control goal is to prioritize the control accuracy λ1=1.0 and allow the medium pitch speed component load λ3=0.5. The control effect at this time is: the motor torque fluctuation is small and the error is stable at ±0.2°;

[0211] When the wind speed changes drastically and bearing wear occurs, the control target is to switch to life extension mode, set component load λ3 to 1.5, and limit the pitch speed to ≤3° / s. The control effect at this time is: the error temporarily expands to ±0.5°, but the bearing load is reduced by 30%, and the life is extended by 25%.

[0212] The above embodiments are intended only to illustrate the structural concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the present invention and implement it accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention should be included in the scope of protection of the present invention.

Claims

1. A performance degradation prediction method based on a wind turbine independent pitch control system, characterized in that: The following steps are involved: Step 1: Using the pitch angle control error as the core degradation representation, a performance degradation model is established using the Wiener process. By defining the target pitch angle, actual pitch angle, and control error, a stochastic differential equation with drift and diffusion terms is constructed. After discretization, maximum likelihood-based parameter estimation, and timely updates, the degradation patterns of key components of the variable pitch system are analyzed in real time. Step 2: Construct a joint degradation-control model, combining the Wiener process degradation law with the pitch system dynamics. Define a state vector containing the control error and its changing rate, introduce the deterministic drift and random fluctuation caused by degradation, and construct the state space equation, including: Step 2-1: Define the state vector, take over the output of the Wiener process, integrate the degradation and control input, simplify the variable pitch system into a second-order linear system, and establish the basic dynamic model without degradation based on the core dynamic characteristics determined by inertia-damping-control gain. , where d represents the damping coefficient, k represents the control gain, represents the pitch angular acceleration, represents the damping term, u(t) represents the control voltage of the pitch motor; Step 2-2: Define the derivative of the control error and the acceleration of the target angle ,get , substituting the non-degenerate dynamic equation into it, we get: ; Step 2-3: Deterministic drift μ and process noise due to degradation Introducing random perturbation σ and embedding the degradation term into the error dynamic equation, we obtain: , complete the fusion of Wiener process degradation terms; Step 2-4: Construct the state space equation and define the state vector , then the state derivative is: , the standard linear stochastic differential equation SDE form is: , where A is the system matrix, B is the control matrix, and η(t) is the degraded input vector; Step 3: Design a "life extension-oriented" objective function that integrates the three major requirements of control accuracy, energy consumption, and component load. Dynamically adjust the weight coefficient based on the degradation state. When degradation worsens, the component load weight is increased and the control accuracy weight is reduced. Step 4: Implement model predictive control and adopt a rolling optimization strategy. Based on the latest degradation parameters and state estimates, solve the optimal control sequence for the next N steps and implement only the first control variable. Process hard constraints and use Kalman filtering to fuse measured data for feedback correction. Adjust the control strategy in real time according to typical operating conditions to balance control accuracy and component life extension.

2. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 1 is characterized in that: The step 1 comprises: Step 1-1: Using the pitch angle control error as the core degradation representation, by defining the target pitch angle, actual pitch angle, and control error, an input system for the variable pitch system performance degradation model is established. This is used to quantitatively analyze the degradation state of key components and drive the control strategy. Step 1-2: Based on the Wiener process, the degradation of the variable pitch system is abstracted as the superposition of deterministic trends and random fluctuations. The degradation dynamics are described by the stochastic differential equation (SDE). A continuous-time model is established to describe the dynamic degradation process of the pitch angle control error. Steps 1-3: Discretization processing: Based on the continuous model, by setting the sampling period, the Wiener process in the continuous time domain is converted into sequence data at discrete time points, and the actual collected sensor data is used to estimate the model parameters and perform real-time control; Steps 1-4: Based on the collected error increments and the joint probability density, the likelihood function is constructed by taking the logarithm, and the drift coefficient and diffusion coefficient are solved based on the degradation of the maximum likelihood method to obtain the average rate of change of the error sequence and the variance of the error increment; Steps 1-5: The recursive maximum likelihood method is used to handle non-stationary degradation, the forgetting factor λ is used to weight the historical data, and the parameters are updated online to make the model more sensitive to the current degradation state.

3. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 2 is characterized in that: The step 1-1 includes: Step a: Define core variables, including the target pitch angle calculated in real time by the unit's main control system based on wind speed and power optimization algorithm , the actual pitch angle measured by the pitch system angle sensor , and the control error e(t), the control error g ; Step b: Apply the core variables and use the control error e(t) as the state variable of the Wiener process model through degradation dynamics: Based on time series data, the deterministic drift μ and random fluctuation σ are estimated using the maximum likelihood method to quantitatively analyze the component degradation rate and volatility; Step c: Update the parameters in real time through the forgetting factor λ to enhance the sensitivity to accelerated degradation and obtain its rate of change by taking the derivative of the control error e(t) , forming the state vector x(t) as the input of the degradation-control joint model, which is embedded in the state space equation to drive the control strategy optimization.

4. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 3 is characterized in that: Steps 1-3 include: Step d: Discrete time series construction, assuming the sampling period is , discrete moments , corresponding to the error sequence , error increment: ,in are independent normal random variables; Step e: Through the incremental property of the Wiener process , indicating that Brownian motion in any time interval The increment on Obey the normal distribution derivation logic with mean 0 and variance Δt, Δt represents the length of the time interval, by combining the continuous diffusion coefficient σ with Δt and multiplying it by the standard normal distribution random variable , obtain the diffusion term discretization , using the variance formula The fluctuation characteristics of the error increment in the discrete model are constrained to ensure that the discretized model is consistent with the continuous-time model in describing the randomness of the degradation process, where Δek represents the error increment after discretization.

5. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 1 is characterized in that: The step 3 comprises: Step 3-1: Control target decomposition, including control accuracy λ1 for minimizing the square integral of the error and ensuring power tracking accuracy, energy consumption λ2 for suppressing excessive control input, and component load λ3 for penalizing the pitch angular velocity. , reducing dynamic loads on bearings and / or gears; Step 3-2: Dynamically adjust the weights of the control degradation process. When the Wiener process parameters show that the degradation is getting worse, automatically increase λ3 to reduce the pitch speed and / or reduce λ1, allowing a slight decrease in accuracy in exchange for an increase in life.

6. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 1 is characterized in that: The step 4 comprises: Step 4-1: Rolling optimization strategy, at the current moment , based on the latest degradation parameters and state estimation , solve the optimal control sequence for the next N steps , only the first control quantity is implemented , perform optimization problem and state transfer, the optimization problem is: , the state transition is: using the state space equation to predict the future state: ,in is the random noise generated by Monte Carlo simulation to simulate the uncertainty of degradation; Step 4-2: Perform hard constraint processing while avoiding physical damage, including limiting the pitch angular velocity, mechanically limiting the pitch angle range, and controlling the motor voltage input saturation; Step 4-3: Feedback correction, sensor noise and model error, real-time update of state estimation through Kalman filter: , the corrected state estimate As the initial condition for the next cycle optimization; Step 4-4: Deep coupling of the degradation model and the control model. Within the time of executing a rolling optimization strategy, a closed loop of data collection → degradation modeling → control calculation → execution is completed, responding to degradation changes in real time.

7. The performance degradation prediction method based on the wind turbine independent pitch control system according to claim 6 is characterized in that: The step 4-4 includes: When the wind speed is stable, the control goal is to prioritize the control accuracy λ1=1.0 and allow the medium pitch speed component load λ3=0.

5. The control effect at this time is: the motor torque fluctuation is small and the error is stable at ±0.2°; When the wind speed changes drastically and bearing wear occurs, the control target is to switch to life extension mode, set component load λ3 to 1.5, and limit the pitch speed to ≤3° / s. The control effect at this time is: the error temporarily expands to ±0.5°, but the bearing load is reduced by 30%, and the life is extended by 25%.

Images