Estimation method and system for nonlinear saturation system of wind turbine variable pitch

Abstract

The embodiment of the application provides an estimation method and system of a wind turbine variable pitch nonlinear saturation system, which introduces a saturation compensation network after a standard EKF prediction step, perceives the saturation depth of an instruction and the initial prediction deviation of a physical model, and further corrects the model prediction error caused by the saturation effect. In this way, the prior state prediction can be actively and accurately compensated before the update step of the EKF, so as to effectively inhibit the impact of large new information caused by saturation on the filtering process, avoid the estimation divergence problem that may occur in the traditional method, and provide reliable state information support for accurate control and safe operation of the wind turbine under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of intelligent estimation technology, specifically to an estimation method and system for a wind turbine pitch nonlinear saturation system. Background Technology

[0002] The pitch control system of a wind turbine is the core actuator for power control and load optimization, and its performance directly affects the efficiency and safety of wind power generation. To achieve precise control of the pitch control system, its key operating states need to be acquired in real time. However, due to the inherent nonlinear characteristics of the pitch control system and the unavoidable physical limitations of the actuator when dealing with extreme conditions—namely, input saturation—accurate state estimation becomes extremely difficult. Therefore, developing a state estimation scheme for wind turbine pitch control systems that can effectively handle nonlinearity and input saturation issues is of great significance for improving the robustness and reliability of wind turbine control systems.

[0003] In existing technologies, model-based estimation algorithms such as Extended Kalman Filter (EKF) are widely used for state estimation of such nonlinear systems. These methods predict the state using a system model and then update it by incorporating actual measurements. However, when a pitch system encounters sudden strong winds or other unforeseen conditions, the controller issues large pitch commands, causing the pitch actuator to enter a deep saturation state. In this case, the actual physical response is much smaller than the commanded value. Traditional EKF methods typically use control commands that do not consider saturation effects as model inputs in the prediction step, leading to a significant deviation between the predicted state and the actual system state. This deviation creates an unusually large piece of information that, in subsequent updates, not only severely contaminates the state estimation results but may also incorrectly correct the estimation error covariance matrix, ultimately causing a degraded performance of the filtering algorithm or even estimation divergence, posing a potential threat to the stable operation of the system.

[0004] Therefore, an optimized estimation method for the nonlinear saturated system of wind turbine pitch control is desired. Summary of the Invention

[0005] The present invention aims to at least solve one of the technical problems existing in the prior art, and provides an estimation method and system for a wind turbine pitch nonlinear saturation system.

[0006] In a first aspect, embodiments of the present invention provide an estimation method for a wind turbine pitch nonlinear saturated system, comprising: Obtain the control input vector for the current cycle, the optimal state estimation vector for the previous time step, and the estimation error covariance matrix for the previous time step; Standard EKF prediction is performed on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step. Saturation depth sensing and residual calculation are performed based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment to obtain the saturation depth index and initial information; Based on the saturation depth index and initial information, the prior state prediction at the current moment is corrected by a saturation compensation network to obtain the compensated prior state prediction. The final EKF update is performed based on the compensated prior state prediction, the prior error covariance at the current time, and the measurement vector at the current period to obtain the optimal state estimate and the estimation error covariance at the current time.

[0007] In some possible embodiments, obtaining the control input vector for the current period, the optimal state estimation vector for the previous time step, and the estimation error covariance matrix for the previous time step includes: The received raw pitch angle signal and raw generator speed signal are conditioned and filtered, and physical units are converted to obtain the measurement vector of the current cycle. Based on the current pitch angle position command and the previous cycle pitch angle position command, the control input vector for the current cycle is calculated, and the control input vector for the current cycle includes the commanded pitch rate for the current cycle.

[0008] In some possible embodiments, standard EKF prediction is performed on the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step, including: Prior state estimation is performed based on the optimal state estimation vector of the previous time step and the control input vector of the current period to obtain the prior state prediction of the current time step. The partial derivative of the optimal state estimation vector at the previous time step is calculated to obtain the state transition Jacobian matrix; Based on the state transition Jacobian matrix, the prior error covariance matrix of the previous time step is propagated to obtain the prior error covariance of the current time step.

[0009] In some possible embodiments, prior error covariance propagation is performed on the estimated error covariance matrix of the previous time step based on the state transition Jacobian matrix to obtain the prior error covariance of the current time step, including: Based on the state transition Jacobian matrix, the estimation error covariance matrix of the previous time step is projected onto the current time step to obtain the projected estimation error covariance matrix. The prior error covariance at the current moment is obtained by adding the projected estimated error covariance matrix to the process noise covariance.

[0010] In some possible embodiments, saturation depth sensing and residual calculation are performed based on the control input vector of the current period, the measurement vector of the current period, and the prior state prediction at the current moment to obtain the saturation depth index and initial information, including: Extract the commanded pitch rate for the current cycle from the control input vector of the current cycle; Calculate the saturation depth index based on the pitch rate commanded in the current cycle; The initial information is calculated based on the measurement vector of the current period and the prior state prediction at the current moment.

[0011] In some possible embodiments, calculating the saturation depth index based on the commanded pitch rate of the current cycle includes: calculating the saturation depth index based on the commanded pitch rate of the current cycle using the following formula:

[0012] in, To preset the pitch rate limit, The pitch rate is the commanded pitch rate for the current cycle. This is the function for finding the maximum value.

[0013] In some possible embodiments, the initial information is calculated based on the measurement vector of the current period and the prior state prediction at the current moment, including: Based on the measurement vector of the current period and the prior state prediction at the current moment, the initial information is calculated using the following formula:

[0014] in, For the observation matrix, The measurement vector for the current period. Predict the prior state at the current moment.

[0015] In some possible embodiments, based on the saturation depth index and initial information, the prior state prediction at the current moment is corrected by a saturation compensation network to obtain a compensated prior state prediction, including: Activation compensation is determined in response to a saturation depth index that is greater than zero. The saturation depth index, initial information, and the prior state prediction at the current time are input into the saturation compensation network to obtain the state correction vector. The compensated prior state prediction is obtained by adding the state correction vector to the prior state prediction at the current time.

[0016] In some possible embodiments, the saturation compensation network is a pre-trained multilayer perceptron model.

[0017] Secondly, embodiments of the present invention provide an estimation system for a wind turbine pitch nonlinear saturation system, comprising: The data acquisition module is used to acquire the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step. The standard EKF prediction module is used to perform standard EKF prediction on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step. The saturation depth sensing and residual calculation module is used to perform saturation depth sensing and residual calculation based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment to obtain the saturation depth index and initial information. The saturation compensation network correction module is used to perform saturation compensation network correction on the prior state prediction at the current moment based on the saturation depth index and initial information to obtain the compensated prior state prediction. The final EKF update module is used to perform the final EKF update based on the compensated prior state prediction, the prior error covariance at the current time, and the measurement vector of the current period to obtain the optimal state estimate and the estimation error covariance at the current time.

[0018] Compared with existing technologies, the present invention provides an estimation method and system for a wind turbine pitch nonlinear saturated system. By innovatively introducing a saturation compensation network after the standard EKF prediction step, this network further corrects the model prediction error caused by saturation effects by sensing the saturation depth of the command and the initial prediction deviation of the physical model. In this way, prior state predictions can be proactively and accurately compensated before the EKF update step, effectively suppressing the impact of large innovations caused by saturation on the filtering process. This avoids the estimation divergence problem that may occur in traditional methods, providing reliable state information support for the precise control and safe operation of wind turbines under complex conditions. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1A flowchart of an estimation method for a wind turbine pitch nonlinear saturated system according to an embodiment of the present invention; Figure 2 This is a data flow diagram illustrating the estimation method for a wind turbine pitch nonlinear saturated system according to an embodiment of the present invention. Figure 3 This is a block diagram of the estimation system for a wind turbine pitch nonlinear saturation system according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0023] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0024] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0025] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0026] In the technical solution of this invention, an estimation method for a nonlinear saturated system of wind turbine pitch is proposed. Figure 1 This is a flowchart of an estimation method for a wind turbine pitch nonlinear saturated system according to an embodiment of the present invention. Figure 2 This is a system architecture diagram of an estimation method for a wind turbine pitch nonlinear saturated system according to an embodiment of the present invention. Figure 1 and Figure 2 As shown, the estimation method for a wind turbine pitch nonlinear saturated system according to an embodiment of the present invention includes the following steps: S1, obtaining the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step; S2, performing standard EKF prediction on the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step; S3, performing saturation depth sensing and residual calculation based on the control input vector of the current period, the measurement vector of the current period, and the prior state prediction of the current time step to obtain the saturation depth index and initial information; S4, performing saturation compensation network correction on the prior state prediction of the current time step based on the saturation depth index and initial information to obtain the compensated prior state prediction; S5, performing final EKF update based on the compensated prior state prediction, the prior error covariance of the current time step, and the measurement vector of the current period to obtain the optimal state estimate and the estimation error covariance of the current time step.

[0027] Specifically, S1 involves acquiring the control input vector for the current cycle, the optimal state estimation vector for the previous moment, and the estimation error covariance matrix for the previous moment. In the technical solution of this invention, firstly, the received raw pitch angle signal and raw generator speed signal are subjected to signal conditioning and filtering, as well as physical unit conversion, to obtain the measurement vector for the current cycle. Signal conditioning aims to adjust the raw electrical signals (such as voltage or current) output by the sensors to a range suitable for subsequent processing. The filtering stage, for example using a low-pass filter, is used to remove high-frequency noise mixed in the measurement signal and extract the effective signal components. Physical unit conversion converts the processed signal into units with clear physical meaning, such as converting the pitch angle signal into "degrees (°)" and the generator speed signal into "revolutions per minute (rpm)". After this series of processing steps, the measurement vector zk for the current cycle is formed. Then, in order to obtain the control input of the drive state prediction model, the control input vector for the current cycle is calculated based on the current pitch angle position command and the pitch angle position command of the previous cycle. The control input vector for the current cycle includes the pitch rate command for the current cycle. This calculation is achieved by differentiating the discrete pitch angle position commands.

[0028] Specifically, S2 involves performing standard EKF prediction on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step. That is, based on the system's dynamic model, the state estimation and its uncertainty from the previous time step are propagated forward to the current time step, thus providing a preliminary, model-based estimate of the system's current state without relying on current actual measurements. This prediction result, namely the prior state prediction and the prior error covariance, will serve as the basis for subsequent state correction by fusing measurement information, acting as a bridge connecting past and current states.

[0029] In practice, firstly, a priori state estimation is performed based on the optimal state estimation vector from the previous moment and the control input vector for the current period to obtain the priori state prediction for the current moment. This step is implemented through a nonlinear state transition function, which describes the physical dynamic characteristics of the wind turbine pitch system. Using the optimal state estimation from the previous moment and the control input vector for the current period as input, it calculates the priori state prediction for the current moment.

[0030] Subsequently, the propagation calculation of the prior error covariance is performed to quantify the changes in estimation uncertainty caused by the uncertainty of the system model (process noise) and the evolution of the state from the previous time step to the current time step. Specifically, firstly, the partial derivatives of the optimal state estimation vector at the previous time step are calculated to obtain the state transition Jacobian matrix. The transition Jacobian matrix is ​​the first-order partial derivative matrix of the system's nonlinear state transition function with respect to the state vector, which provides a local linear dynamic approximation of the system near a specific operating point.

[0031] Furthermore, based on the state transition Jacobian matrix, the prior error covariance matrix of the previous time step is propagated to obtain the prior error covariance of the current time step. This step not only reflects the propagation of deterministic uncertainty caused by the dynamic evolution of the system, but also takes into account the inherent uncertainty of the system model itself (i.e., process noise), thus providing an accurate measure of the reliability of the current state prediction for the subsequent EKF update step. In this process, firstly, based on the state transition Jacobian matrix, the estimation error covariance matrix of the previous time step is projected onto the current time step to obtain the projected estimation error covariance matrix; that is, the estimation error covariance matrix of the previous time step is transformed using the system dynamic model linearized at the prior state estimation point. This projection calculation strictly follows the covariance propagation law. After projection, the projected estimated error covariance matrix is ​​added to the process noise covariance matrix to account for new uncertainties introduced during the evolution from time k-1 to time k, such as dynamics not fully captured by the system model, unmodeled disturbances, or parameter changes. These uncertainties are described by the process noise covariance matrix. By adding the projected covariance matrix to the process noise covariance matrix, the prior error covariance at the current time can be obtained.

[0032] Specifically, S3 involves saturation depth sensing and residual calculation based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment to obtain the saturation depth index and initial information. It should be understood that during wind turbine pitch control, when the commanded pitch rate exceeds the system's physical limits, the system enters a nonlinear saturation state, leading to a decrease in the estimation accuracy of the standard EKF algorithm. By introducing a saturation depth sensing and residual calculation mechanism, the degree of system saturation can be quantified, and initial information reflecting the difference between measured and predicted values ​​can be generated, providing key input parameters for subsequent saturation compensation. That is, before performing the final state update, the degree of model mismatch caused by actuator saturation is actively diagnosed and quantified, and the original prediction error under the standard EKF framework is calculated. The saturation depth index characterizes the degree to which the control command exceeds the physical execution capability; the initial information reflects the original deviation between the model prediction and the actual measurement. These two variables will jointly serve as the core input for the subsequent saturation compensation network correction stage, providing the necessary basis for intelligently correcting the prior state prediction.

[0033] In practice, firstly, the commanded pitch rate for the current cycle is extracted from the control input vector of the current cycle; specifically, this can be achieved by directly accessing the commanded pitch rate data stored in the control input vector. Next, based on the commanded pitch rate of the current cycle, the saturation depth index is calculated; the specific calculation process is expressed by the formula:

[0034] in, To preset the pitch rate limit, The pitch rate is the commanded pitch rate for the current cycle. The function is used to maximize the value. Here, the saturation depth index is a non-negative scalar used to quantitatively describe the extent to which the control command exceeds the physical limits of the actuator; it is a key signal for activating and guiding saturation compensation. Subsequently, based on the measurement vector of the current cycle and the prior state prediction at the current moment, the initial information is calculated. This step follows the framework of standard Kalman filtering, calculating the difference between the measured value and the predicted value without any compensation; the specific calculation process is expressed by the formula:

[0035] in, For the observation matrix, The measurement vector for the current period. This represents the prior state prediction at the current moment. The initial information is a vector representing the difference between the actual measured value and the uncompensated prior state prediction in the measurement space, containing raw information about the model prediction error.

[0036] Specifically, in step S4, based on the saturation depth index and initial information, a saturation compensation network is applied to correct the prior state prediction at the current moment to obtain a compensated prior state prediction. It should be understood that when the wind turbine pitch system enters saturation, the state prediction based on the linearization assumption of the standard EKF algorithm will produce significant deviations, leading to errors in subsequent state updates. By introducing a saturation compensation network to specifically correct the prior state prediction, the model mismatch problem caused by the nonlinear saturation effect can be effectively compensated, improving the accuracy of state estimation.

[0037] In practice, the system first activates compensation in response to a saturation depth index greater than zero. This is achieved by checking if the saturation depth index is greater than zero. A greater-than-zero saturation index indicates the system is in a nonlinear saturated state, requiring compensation activation; otherwise, the standard EKF process remains unchanged. This logic ensures compensation only intervenes when necessary, avoiding unnecessary interference with the estimation process in non-saturated states. After determining the need for compensation activation, the system enters the network inference phase. This involves inputting the saturation depth index, initial information, and the current prior state prediction into the saturation compensation network to obtain the state correction vector. This network is a pre-trained multilayer perceptron model. This model has been trained offline using extensive historical data, enabling it to learn the complex mapping between the system's actual behavior and model predictions in a nonlinear saturated state. The network's structure typically includes an input layer, several hidden layers, and an output layer. The input layer receives three key parameters: the saturation depth index reflects the system's saturation level, the initial information reflects the difference between the measured and predicted values, and the prior state prediction provides the basis for the current state estimation. Then, the state correction vector is added to the prior state prediction at the current moment to obtain the compensated prior state prediction. The state correction vector is essentially an estimate of the prior state prediction error, containing the prediction bias inferred by the saturation compensation network based on the current saturation state and measurement residuals. The system then adds this state correction vector to the prior state prediction at the current moment to obtain the compensated prior state prediction. This addition operation achieves targeted adjustment of the original prediction, making the corrected prediction closer to the system's actual behavior under saturation. The compensated prior state prediction will be used as input to the subsequent EKF update, replacing the original prior prediction, thus incorporating the effect of saturation compensation into the entire estimation process.

[0038] Through this neural network-based intelligent compensation mechanism, the system can effectively overcome the estimation bias caused by nonlinear saturation while maintaining the standard EKF framework. Compared with traditional gain adjustment or model parameter adaptation schemes, this method can better handle complex nonlinear characteristics under saturation conditions and does not require precise modeling of the system's physical model, thus exhibiting greater practicality and robustness.

[0039] Specifically, in step S5, a final EKF update is performed based on the compensated prior state prediction, the prior error covariance at the current moment, and the measurement vector of the current period to obtain the optimal state estimate and the estimation error covariance at the current moment. The optimal state estimate refers to the state variable values ​​obtained under the minimum mean square error criterion, representing the most reliable inference of the system state after considering the system model and measurement information. The estimation error covariance matrix quantifies the uncertainty of the state estimate; its diagonal elements represent the estimation variance of each state variable, and its off-diagonal elements represent the correlation of estimation errors between state variables. Through this complete EKF update process, the system can fully utilize the compensated state prediction and actual measurement information to output a state estimation result with optimal statistical characteristics. This process not only considers the nonlinear characteristics of the system but also maintains a quantitative assessment of the estimation accuracy through the update of the covariance matrix, providing reliable state feedback for the stable operation of the wind turbine pitch control system.

[0040] In practice, the first step is to calculate the Kalman gain. The Kalman gain is a matrix that optimally balances model prediction and sensor measurement. Its calculation process seeks an optimal balance between prediction uncertainty (caused by prior error covariance) and measurement uncertainty (caused by measurement noise covariance).

[0041] Secondly, after calculating the Kalman gain, the state estimate is updated to obtain the optimal state estimate for the current time step. This update is achieved by adding a novel term weighted by the Kalman gain to the compensated prior state prediction. This novel term is the difference between the actual measurement and the measurement prediction calculated based on the compensated prediction.

[0042] Finally, the estimation error covariance matrix is ​​updated to obtain the posterior estimation error covariance at the current time step. During this process, the uncertainty in state estimation is reduced due to the incorporation of new measurement information. This update is calculated using the Kalman gain and the prior error covariance.

[0043] In summary, the estimation method for the nonlinear saturated system of wind turbine pitch control according to embodiments of the present invention is explained. It innovatively introduces a saturation compensation network after the standard EKF prediction step. By sensing the saturation depth of the command and the initial prediction deviation of the physical model, it further corrects the model prediction error caused by the saturation effect. In this way, it can proactively and accurately compensate for the prior state prediction before the EKF update step, thereby effectively suppressing the impact of large innovations caused by saturation on the filtering process, avoiding the estimation divergence problem that may occur in traditional methods, and providing reliable state information support for the precise control and safe operation of wind turbines under complex operating conditions.

[0044] Furthermore, an estimation system for a wind turbine pitch nonlinear saturated system is also provided.

[0045] Figure 3 This is a block diagram of the estimation system for a wind turbine pitch nonlinear saturation system according to an embodiment of the present invention. Figure 3 As shown, the estimation system 300 for a wind turbine pitch nonlinear saturation system according to an embodiment of the present invention includes: a data acquisition module 310, used to acquire the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step; a standard EKF prediction module 320, used to perform standard EKF prediction on the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step; and a saturation depth sensing and residual calculation module 330, used to calculate the control input vector of the current period, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step. The system performs saturation depth sensing and residual calculation on the input vector, the measurement vector of the current period, and the prior state prediction at the current moment to obtain the saturation depth index and initial information; the saturation compensation network correction module 340 is used to perform saturation compensation network correction on the prior state prediction at the current moment based on the saturation depth index and initial information to obtain the compensated prior state prediction; the final EKF update module 350 is used to perform final EKF update based on the compensated prior state prediction, the prior error covariance at the current moment, and the measurement vector of the current period to obtain the optimal state estimate at the current moment and the estimation error covariance at the current moment.

[0046] As described above, the estimation system 300 for the wind turbine pitch nonlinear saturation system according to embodiments of the present invention can be implemented in various wireless terminals, such as servers with estimation algorithms for the wind turbine pitch nonlinear saturation system. In one possible implementation, the estimation system 300 for the wind turbine pitch nonlinear saturation system according to embodiments of the present invention can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the estimation system 300 for the wind turbine pitch nonlinear saturation system can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the estimation system 300 for the wind turbine pitch nonlinear saturation system can also be one of many hardware modules of the wireless terminal.

[0047] Alternatively, in another example, the estimation system 300 of the wind turbine pitch nonlinear saturation system and the wireless terminal can also be separate devices, and the estimation system 300 of the wind turbine pitch nonlinear saturation system can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0048] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. An estimation method for a wind turbine pitch nonlinear saturated system, characterized in that, include: Obtain the control input vector for the current cycle, the optimal state estimation vector for the previous time step, and the estimation error covariance matrix for the previous time step; Standard EKF prediction is performed on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step. Saturation depth sensing and residual calculation are performed based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment to obtain the saturation depth index and initial information; Based on the saturation depth index and initial information, the prior state prediction at the current moment is corrected by a saturation compensation network to obtain the compensated prior state prediction. The final EKF update is performed based on the compensated prior state prediction, the prior error covariance at the current time, and the measurement vector at the current period to obtain the optimal state estimate and the estimation error covariance at the current time.

2. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 1, characterized in that, Obtain the control input vector for the current cycle, the optimal state estimation vector from the previous time step, and the estimation error covariance matrix from the previous time step, including: The received raw pitch angle signal and raw generator speed signal are conditioned and filtered, and physical units are converted to obtain the measurement vector of the current cycle. Based on the current pitch angle position command and the previous cycle pitch angle position command, the control input vector for the current cycle is calculated, and the control input vector for the current cycle includes the commanded pitch rate for the current cycle.

3. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 2, characterized in that, Standard EKF prediction is performed on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step, including: Prior state estimation is performed based on the optimal state estimation vector of the previous time step and the control input vector of the current period to obtain the prior state prediction of the current time step. The partial derivative of the optimal state estimation vector at the previous time step is calculated to obtain the state transition Jacobian matrix; Based on the state transition Jacobian matrix, the prior error covariance matrix of the previous time step is propagated to obtain the prior error covariance of the current time step.

4. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 3, characterized in that, Based on the state transition Jacobian matrix, prior error covariance propagation is performed on the estimation error covariance matrix of the previous time step to obtain the prior error covariance of the current time step, including: Based on the state transition Jacobian matrix, the estimation error covariance matrix of the previous time step is projected onto the current time step to obtain the projected estimation error covariance matrix. The prior error covariance at the current moment is obtained by adding the projected estimated error covariance matrix to the process noise covariance.

5. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 2, characterized in that, Based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment, saturation depth sensing and residual calculation are performed to obtain the saturation depth index and initial information, including: Extract the commanded pitch rate for the current cycle from the control input vector of the current cycle; Calculate the saturation depth index based on the pitch rate commanded in the current cycle; The initial information is calculated based on the measurement vector of the current period and the prior state prediction at the current moment.

6. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 5, characterized in that, Calculating the saturation depth index based on the commanded pitch rate of the current cycle includes: calculating the saturation depth index based on the commanded pitch rate of the current cycle using the following formula: in, To preset the pitch rate limit, The pitch rate is the commanded pitch rate for the current cycle. This is the function for finding the maximum value.

7. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 5, characterized in that, Based on the measurement vector of the current period and the prior state prediction at the current moment, the initial information is calculated, including: Based on the measurement vector of the current period and the prior state prediction at the current moment, the initial information is calculated using the following formula: in, For the observation matrix, The measurement vector for the current period. Predict the prior state at the current moment.

8. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 6, characterized in that, Based on the saturation depth index and initial information, the prior state prediction at the current moment is corrected using a saturation compensation network to obtain a compensated prior state prediction, including: Activation compensation is determined in response to a saturation depth index that is greater than zero. The saturation depth index, initial information, and the prior state prediction at the current time are input into the saturation compensation network to obtain the state correction vector. The compensated prior state prediction is obtained by adding the state correction vector to the prior state prediction at the current time.

9. The estimation method for a wind turbine pitch nonlinear saturation system according to claim 8, characterized in that, The saturation compensation network is a pre-trained multilayer perceptron model.

10. An estimation system for a wind turbine pitch nonlinear saturated system, characterized in that, include: The data acquisition module is used to acquire the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step. The standard EKF prediction module is used to perform standard EKF prediction on the control input vector of the current cycle, the optimal state estimation vector of the previous time step, and the estimation error covariance matrix of the previous time step to obtain the prior state prediction and the prior error covariance of the current time step. The saturation depth sensing and residual calculation module is used to perform saturation depth sensing and residual calculation based on the control input vector of the current cycle, the measurement vector of the current cycle, and the prior state prediction at the current moment to obtain the saturation depth index and initial information. The saturation compensation network correction module is used to perform saturation compensation network correction on the prior state prediction at the current moment based on the saturation depth index and initial information to obtain the compensated prior state prediction. The final EKF update module is used to perform the final EKF update based on the compensated prior state prediction, the prior error covariance at the current time, and the measurement vector of the current period to obtain the optimal state estimate and the estimation error covariance at the current time.

Images