An effective wind speed soft measurement method

Abstract

The present invention provides an effective wind speed soft measurement method, comprising a wind turbine SCADA data acquisition module, an offline optimization calculation module based on an improved gray wolf optimized IGWO algorithm, an effective wind speed soft measurement modeling module based on a kernel extreme learning machine (KELM) algorithm, and the like. The signal output end of the wind turbine SCADA data acquisition module sequentially passes through a SCADA data preprocessing and normalization module, an offline optimization calculation module based on an improved gray wolf optimized IGWO algorithm, an effective wind speed soft measurement modeling module based on a kernel extreme learning machine (KELM) algorithm, an effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and an effective wind speed soft measurement model performance evaluation module, and finally outputs the effective value of the wind speed. The present invention establishes an effective wind speed soft measurement model, which does not require wind tunnel experiments or virtual simulations to obtain the data required for effective wind speed soft measurement modeling, but only requires SCADA data of the wind turbines actually operating in the wind farm. The model has the characteristics of low cost, high prediction accuracy, and ease of implementation.

Description

Technical Field

[0001] The present invention relates to an effective wind speed soft measurement method, belonging to the technical field of soft measurement of effective wind speed of wind power generation system, and particularly to an effective wind speed soft measurement method based on an improved grey wolf algorithm optimized kernel extreme learning machine. Background Art

[0002] Accurate wind speed measurement is fundamental to wind turbine operation and control. Wind turbine output power models are typically established through wind tunnel experiments under ideal conditions. However, when operating in wind farms with harsh wind conditions (such as plateau and mountainous wind farms and offshore wind farms), the wind speed distribution across the entire wind turbine's rotational plane varies due to factors such as turbulence, tower structure, wind shear differences, and surface roughness. This results in a significant deviation between the wind turbine's actual output power curve and the theoretical power design curve, posing a significant challenge to wind turbine operation and control. The wind speed measured by the anemometer described above is only the wind speed at a point on top of the nacelle (except for PIV speed measurement), which differs significantly from the effective wind speed experienced by the entire wind turbine's rotational plane. Therefore, the effective wind speed cannot be measured directly, but it can be estimated using soft sensing methods.

[0003] The secondary variables for effective wind speed soft sensing are wind turbine generator speed, actual generator output power, and pitch angle, while the dominant variable is effective wind speed. Accurate measurement of effective wind speed is crucial for wind turbine control and optimization. Effective wind speed acquisition not only improves the wind turbine's ability to capture wind energy but also provides real-time reference input for wind turbine control and optimization, ensuring safe and stable operation. Wind speed soft sensing models can utilize Kalman filters, neural networks, adaptive fuzzy neural networks, extreme learning machines, or support vector machines for offline identification and online self-calibration.

[0004] The Kalman filter is an algorithm for optimally estimating random signals. While it can produce wind speed predictions when used for wind speed estimation, it struggles to operate in real time when wind speed varies widely. This is because the Kalman filter estimates the state of a nonlinear system based on a linearized model at the operating point. When the operating point changes, the linearized model needs to be updated in real time, placing a significant computational burden on the computer. Support vector machines (SVMs), a learning method based on statistical learning theory, are effective in addressing problems such as small sample sizes, nonlinearity, high dimensionality, and local minima, and possess strong generalization capabilities. However, standard SVMs are slow to compute and suffer from robustness, sparsity, and large-scale computational challenges. The least-squares SVM algorithm (LS-SVM) addresses these issues, simplifying computational complexity and accelerating solution speed. However, when processing large datasets, LS-SVMs suffer from slow training, high computational complexity, and difficulty in online training. Neural networks have the ability to rapidly model nonlinear data. They adjust their network structure and connection weights through repeated learning of training sets, and can then predict and estimate unknown data. However, neural networks are based on the principle of empirical minimization, resulting in complex structures, prone to local optimal solutions, and prone to overlearning problems. The wind speed estimate from the adaptive TS fuzzy neural network effectively tracks wind speed sample values, accelerating convergence and improving local convergence. It also possesses advantages such as strong learning and generalization capabilities. This method is able to effectively track wind speed trends over a wide range of wind speed variations, making it an effective wind speed measurement method. The Kernel-Based Extreme Learning Machine (KELM) is an improved algorithm based on the Extreme Learning Machine (ELM) and combined with a kernel function. KELM improves the model's predictive performance while retaining the advantages of the ELM.

[0005] However, the above soft-sensing methods for wind speed RMS do not consider the following two issues:

[0006] (1) The generator output power is smooth and optimal;

[0007] (2) Minimize the control changes of generator speed and pitch angle. Summary of the Invention

[0008] To solve the above technical problems, the present invention provides an effective wind speed soft measurement method. The effective wind speed soft measurement method proposes a multi-objective optimization model and an improved grey wolf optimization (IGWO) algorithm. The optimal solution of the optimization model is used as a training sample and a test sample. The IGWO algorithm is used to optimize the kernel extreme learning machine to establish a soft measurement model of effective wind speed.

[0009] The present invention is achieved through the following technical solutions.

[0010] The present invention provides an effective wind speed soft measurement method, which includes a wind turbine SCADA data acquisition module, a SCADA data preprocessing and normalization module, an offline optimization calculation module based on an improved gray wolf optimization (IGWO) algorithm, an effective wind speed soft measurement modeling module based on a kernel extreme learning machine (KELM) algorithm, an effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and an effective wind speed soft measurement model performance evaluation module. The signal output end of the wind turbine SCADA data acquisition module sequentially passes through the SCADA data preprocessing and normalization module, the offline optimization calculation module based on the improved gray wolf optimization (IGWO) algorithm, the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm, the effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and the effective wind speed soft measurement model performance evaluation module, and finally outputs the effective value of the wind speed; and

[0011] The wind turbine SCADA data acquisition module collects data on wind turbine output power, generator speed, pitch angle and wind speed;

[0012] The SCADA data preprocessing and normalization module preprocesses the SCADA data collected in the wind turbine SCADA data acquisition module to remove abnormal data;

[0013] The offline optimization calculation module based on the improved Grey Wolf Optimization (IGWO) algorithm applies the IGWO algorithm to solve the established multi-objective optimization model to obtain the optimal predicted values ​​of the wind turbine generator output power, pitch angle, and generator speed;

[0014] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm establishes an effective wind speed soft-sensing model according to the prediction data provided by the offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm;

[0015] The effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm uses the IGWO algorithm to optimize the parameters of the effective wind speed soft sensor model;

[0016] The effective wind speed soft-sensing model performance evaluation module evaluates the performance of the effective wind speed soft-sensing model after parameter optimization, and provides real-time reference input for maximum wind power tracking MPPT control.

[0017] The offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm is a first microprocessor. The first microprocessor performs offline optimization calculations based on the data output by the SCADA data preprocessing and normalization module, and the obtained optimal solution is used as training sample data and test sample data for the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm.

[0018] The first microprocessor includes a multi-objective optimization model and an improved grey wolf optimization (IGWO) algorithm, specifically constructed as follows:

[0019] a. Construct a multi-objective optimization model with wind turbine output power, pitch angle, and generator speed as optimization targets at all wind speeds:

[0020] min F(x)

[0021]

[0022] Where F(x) is a multi-objective function, h(x) is an equality vector constraint function, and g(x) is an inequality vector constraint function;

[0023] Among them, the multi-objective function is:

[0024] F(x)=μJ P +(1-μ)J g ,μ∈(0,1),

[0025]

[0026]

[0027]

[0028] Where, J P is the objective function for the actual output power of the generator to track the expected predicted value at all wind speeds, J g is the objective function for minimizing the control variation of generator speed and pitch angle, x is the decision variable; P gk 、ω gk and β k are the actual output power, generator rotor angular velocity and pitch angle at time k respectively; and are the expected predicted values ​​of generator output power, generator rotor angular velocity and pitch angle at time k+1; μ, w P and w g is the weight coefficient; n is the optimization process, P N 、P m They are the rated output power and theoretical design output power of the wind turbine respectively.

[0029] a.1. Set the constraints of the multi-objective optimization model:

[0030] a.1.1. Aerodynamic power balance constraints

[0031]

[0032]

[0033] Where ρ is the air density in kg / m 3 , v is the wind speed, the unit is m / s, R is the radius of the wind wheel, the unit is m, P m is the wind turbine output power, β is the wind turbine pitch angle, λ is the wind turbine tip speed ratio, n g is the gearbox speed ratio, C p (ω g ,β) is the wind energy utilization coefficient of the wind turbine, and the relationship between them is expressed as follows:

[0034]

[0035] a.1.2. Weight coefficient constraints

[0036] 0<μ<1,

[0037] 0<w p <1,

[0038] 0<w g <1.

[0039] a.1.3. Output power constraints

[0040]

[0041] a.1.4. Pitch angle and generator speed constraints

[0042]

[0043]

[0044] Where, β H and β L are the upper and lower limits of the pitch angle, ω H and ω L are the upper and lower limits of the generator speed;

[0045] b. Improved Grey Wolf Optimization IGWO algorithm:

[0046] b.1. Improved nonlinear convergence factor

[0047]

[0048] Where λ, τ and n are adjustment parameters, t is the current number of iterations, and t max is the maximum number of iterations;

[0049] b.2. Improved adaptive position update equation:

[0050]

[0051] in,

[0052] d max =max{|X1-X2|,|X2-X3|,|X3-X1|},

[0053] d min =min{|X1-X2|,|X2-X3|,|X3-X1|},

[0054] Where X(t+1) is a triangle The center of the inscribed circle, w1, w2 and w3 are weight coefficients; X best is the historical best position vector of the gray wolf individual ω, X worst (t) is the position vector of the gray wolf individual with the worst fitness in the gray wolf population at the tth iteration, b1 and b2 are two random numbers in the interval [0,1], c1 is the learning factor of the gray wolf individual ω, c2 is the reverse search factor, d max It is a triangle The maximum side length, d min It is a triangle The minimum side length of

[0055] c1 and c2 are defined as follows:

[0056]

[0057] The improved nonlinear convergence factor maintains its maximum value in the first half of the iterative search and its minimum value in the second half of the iterative search.

[0058] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm is a second microprocessor, and a kernel function is provided in the second microprocessor, and the kernel function is a Gaussian kernel function:

[0059]

[0060] Where σ is the control parameter, x and x′ are any two points in the state space under consideration;

[0061] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm calls the training sample data stored in the offline optimization calculation module based on the improved gray wolf optimization (IGWO) algorithm: the optimal predicted value of the generator output power Optimal predicted value of generator speed Optimal predicted value of pitch angle Establish a soft sensor model for effective wind speed and send the output signal to the parameter optimization module of the soft sensor model for effective wind speed based on the IGWO algorithm.

[0062] The parameter optimization module of the soft sensor model for effective wind speed based on the IGWO algorithm is the third microprocessor, which is used to call the output signal of the soft sensor modeling module for effective wind speed based on the kernel extreme learning machine (KELM) algorithm to optimize the parameters of the established soft sensor model. The optimization steps are as follows:

[0063] Step 1: Preprocessing and normalization of SCADA data;

[0064] Step 2: Parameter initialization and normalization;

[0065] Step 3: Calculate the fitness function value;

[0066] Step 4: Initialize the position vectors of alpha, beta, and delta;

[0067] Step 5: Set the iteration number k = 1;

[0068] Step 6: Update the parameter nonlinear convergence factor a, coefficient A, and coefficient C, where A = a(2r1 - 1), C = 2r2^2, and r1 and r2 are random numbers in the interval [0, 1];

[0069] Step 7: Update the fitness function values and corresponding positions of alpha, beta, and delta;

[0070] Step 8: Determine whether k is greater than or equal to Max_iter, where Max_iter is the maximum number of iterations. If k ≥ Max_iter, go to Step 9; if k < Max_iter, go to Step 6;

[0071] Step 9: Return the current position of alpha to obtain the optimal solution;

[0072] Step 10: Calculate the kernel matrix Ω ELM =(Ω ELMij ) n×n :Ω ELMij =K(x i ,x j ), where i = 1, 2,..., n; j = 1, 2,..., n, x i and x j are training samples, and n is the total number of samples;

[0073] Step 11: Calculate the weight matrix of the hidden layer output;

[0074] Step 12: Return the regression values ​​σ and C1, where σ is the control parameter of the Gaussian function and C1 is the adjustment parameter.

[0075] In the effective wind speed soft measurement model performance evaluation module, if the effective wind speed soft measurement model after parameter optimization passes the performance evaluation, the soft measurement model parameters are output; if the performance evaluation fails, the module returns to the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm, retrains the samples, and reoptimizes the parameters until the performance evaluation criteria are met.

[0076] The beneficial effects of the present invention are as follows: a multi-objective optimization model with generator output power, generator speed and pitch angle as optimization targets is established, so that the generator output power is smooth and optimal; the nonlinear convergence factor and adaptive position update equation of the standard grey wolf optimization algorithm are improved, so that the control variation of the generator speed and pitch angle is minimized; an effective wind speed soft measurement model is established, which does not require wind tunnel experiments or virtual simulations to obtain the data required for effective wind speed soft measurement modeling, but only requires SCADA data of the actual operation of the wind turbine in the wind farm, and has the characteristics of low cost, high prediction accuracy and easy implementation. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a modeling flow chart of the present invention;

[0078] Figure 2 is the modeling module of the present invention;

[0079] Figure 3 It is the model parameter optimization module of the present invention;

[0080] Figure 4 The triangle of the present invention Schematic diagram of the center position of the inscribed circle. DETAILED DESCRIPTION

[0081] The technical solution of the present invention is further described below, but the scope of protection claimed is not limited to the description.

[0082] like Figure 1As shown, an effective wind speed soft measurement method includes a wind turbine SCADA data acquisition module, a SCADA data preprocessing and normalization module, an offline optimization calculation module based on an improved Grey Wolf Optimization (IGWO) algorithm, an effective wind speed soft measurement modeling module based on a kernel extreme learning machine (KELM) algorithm, an effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and an effective wind speed soft measurement model performance evaluation module. The signal output end of the wind turbine SCADA data acquisition module sequentially passes through the SCADA data preprocessing and normalization module, the offline optimization calculation module based on the improved Grey Wolf Optimization (IGWO) algorithm, the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm, the effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and the effective wind speed soft measurement model performance evaluation module, and finally outputs the effective value of the wind speed; and

[0083] The wind turbine SCADA data acquisition module collects data on wind turbine output power, generator speed, pitch angle and wind speed;

[0084] The SCADA data preprocessing and normalization module preprocesses the SCADA data collected by the wind turbine SCADA data acquisition module, removes abnormal data, and avoids the abnormal data from having a significant impact on the optimal solution of the multi-objective optimization model;

[0085] The offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm solves the established multi-objective optimization model based on the data input by the SCADA data preprocessing and normalization module, and the optimal solution of the multi-objective optimization model is used as the training sample and test sample of the effective wind speed soft sensor modeling module based on the kernel extreme learning machine (KELM) algorithm.

[0086] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm establishes an effective wind speed soft-sensing model according to the prediction data provided by the offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm;

[0087] The effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm calls the IGWO algorithm to optimize the parameters of the effective wind speed soft sensor model established by the effective wind speed soft sensor modeling module based on the kernel extreme learning machine KELM algorithm;

[0088] The effective wind speed soft measurement model performance evaluation module evaluates the performance of the effective wind speed soft measurement model after parameter optimization, provides real-time reference input for maximum wind power tracking MPPT control, and improves the accuracy and effectiveness of the soft measurement model in predicting effective wind speed.

[0089] The offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm is a first microprocessor. The first microprocessor performs offline optimization calculations based on the data output by the SCADA data preprocessing and normalization module, and the obtained optimal solution is used as training sample data and test sample data for the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm.

[0090] The first microprocessor includes a multi-objective optimization model and an improved grey wolf optimization (IGWO) algorithm, specifically constructed as follows:

[0091] a. Construct a multi-objective optimization model with wind turbine output power, pitch angle, and generator speed as optimization targets at all wind speeds:

[0092] min F(x)

[0093]

[0094] Where F(x) is a multi-objective function, h(x) is an equality vector constraint function, and g(x) is an inequality vector constraint function;

[0095] Among them, the multi-objective function is:

[0096] F(x)=μJ P +(1-μ)J g ,μ∈(0,1),

[0097]

[0098]

[0099]

[0100] Where, J P is the objective function for the actual output power of the generator to track the expected predicted value at all wind speeds, J g is the objective function for minimizing the control variation of generator speed and pitch angle, x is the decision variable; P gk 、ω gk and β k are the actual output power, generator rotor angular velocity and pitch angle at time k respectively; and are the expected predicted values ​​of generator output power, generator rotor angular velocity and pitch angle at time k+1; μ, wP and w g is the weight coefficient; n is the optimization process, P N 、P m are the rated output power and theoretical design output power of the wind turbine respectively;

[0101] a.1. Set the constraints of the multi-objective optimization model:

[0102] a.1.1. Aerodynamic power balance constraints

[0103]

[0104]

[0105] Where ρ is the air density in kg / m 3 , v is the wind speed, the unit is m / s, R is the radius of the wind wheel, the unit is m, P m is the wind turbine output power, β is the wind turbine pitch angle, λ is the wind turbine tip speed ratio, n g is the gearbox speed ratio, C p (ω g ,β) is the wind energy utilization coefficient of the wind turbine, and the relationship between them is expressed as follows:

[0106]

[0107] a.1.2. Weight coefficient constraints

[0108] 0<μ<1,

[0109] 0<w p <1,

[0110] 0<w g <1.

[0111] a.1.3. Output power constraints

[0112]

[0113] a.1.4. Pitch angle and generator speed constraints

[0114]

[0115]

[0116] Where, β H and β L are the upper and lower limits of the pitch angle, ω H and ω L are the upper and lower limits of the generator speed;

[0117] b. Improved Grey Wolf Optimization IGWO algorithm:

[0118] b.1. Improved nonlinear convergence factor

[0119]

[0120] Where λ, τ and n are adjustment parameters, t is the current number of iterations, and t max is the maximum number of iterations;

[0121] b.2. Improved adaptive position update equation:

[0122]

[0123] in,

[0124] d max =max{|X1-X2|,|X2-X3|,|X3-X1|},

[0125] d min =min{|X1-X2|,|X2-X3|,|X3-X1|},

[0126] First consider X(t+1) as a triangle The center of the inscribed circle (such as Figure 4 As shown in Figure 2, the weight coefficients w1, w2, and w3 reflect the leadership role and function of the three wolves alpha, beta, and delta in the process of capturing prey. The greater the weight of their position in the position update, the faster they approach the prey and the algorithm can quickly find the optimal solution; X best is the historical best position vector of the gray wolf individual ω, X worst (t) is the position vector of the gray wolf individual with the worst fitness in the gray wolf population at the tth iteration, b1 and b2 are two random numbers in the interval [0,1], c1 is the learning factor of the gray wolf individual ω, c2 is the reverse search factor, d max It is a triangle The maximum side length, d min It is a triangle The minimum side length of

[0127] c1 and c2 are defined as follows:

[0128]

[0129] The IGWO algorithm is used to solve the established multi-objective model and optimize the kernel extreme learning machine, and a soft measurement model of effective wind speed is established. This model does not require wind tunnel data and virtual simulation data, but only SCADA data, which is low in cost, high in prediction accuracy and easy to implement.

[0130] The improved nonlinear convergence factor maintains the maximum value in the first half of the iterative search, improving the global search ability of the gray wolf algorithm, and maintains the minimum value in the second half of the iterative search, improving the local search ability of the gray wolf algorithm, and can quickly search for the optimal solution.

[0131] Based on the SCADA data preprocessed and normalized by the SCADA data module, the IGWO algorithm is used to solve the established multi-objective optimization model to obtain the optimal predicted value of the generator output power. Prediction value of optimal generator speed Optimal predicted value of pitch angle

[0132] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm is a second microprocessor, and a kernel function is provided in the second microprocessor, and the kernel function is a Gaussian kernel function:

[0133]

[0134] Where σ is the control parameter, and x and x′ are any two points in the state space under consideration.

[0135] The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm calls the training sample data stored in the offline optimization calculation module based on the improved gray wolf optimization (IGWO) algorithm: the optimal predicted value of the generator output power Prediction value of optimal generator speed Optimal predicted value of pitch angle Establish a soft sensor model for effective wind speed (such as Figure 2 , is the estimated value of the effective wind speed), and the output signal is transmitted to the effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm.

[0136] The effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm is a third microprocessor, which is used to call the output signal of the effective wind speed soft sensor modeling module based on the kernel extreme learning machine KELM algorithm to optimize the parameters of the established soft sensor model. Figure 3 As shown, the optimization steps are as follows:

[0137] Step 1: Preprocessing and normalization of SCADA data;

[0138] Step 2: parameter initialization and normalization;

[0139] Step 3: Calculate the fitness function value;

[0140] Step 4: Initialize the position vectors of alpha, beta, and delta;

[0141] Step 5: Set the iteration number k = 1;

[0142] Step 6: Update the parameter non-linear convergence factor a, coefficient A, and coefficient C, where A = a(21r - 1), C = 2r2, and r1 and r2 are random numbers in the interval [0, 1];

[0143] Step 7: Update the fitness function values and corresponding positions of alpha, beta, and delta;

[0144] Step 8: Determine whether k is greater than or equal to Max_iter, where Max_iter is the maximum number of iterations. If k ≥ Max_iter, go to Step 9; if k < Max_iter, go to Step 6;

[0145] Step 9: Return the current position of alpha to obtain the optimal solution;

[0146] Step 10: Calculate the kernel matrix Ω ELM =(Ω ELMij ) n×n :Ω ELMij =K(x i ,x j ), where i = 1, 2,..., n; j = 1, 2,..., n, x i and x j are training samples, and n is the total number of samples;

[0147] Step 11: Calculate the weight matrix of the output of the hidden layer;

[0148] Step 12: Return the regression values σ and C1, where σ is the control parameter of the Gaussian function and C1 is the adjustment parameter.

[0149] In the performance evaluation module of the effective wind speed soft sensor model, if the effective wind speed soft sensor model after parameter optimization passes the performance evaluation, the soft sensor model parameters are output; if the performance evaluation fails, return to the effective wind speed soft sensor modeling module based on the kernel extreme learning machine KELM algorithm, retrain the samples, and re-optimize the parameters until the performance evaluation criteria are met.

[0150] As described above, to address the problem of the inability to directly measure the effective wind speed of a wind power generation system, the present invention, based on wind turbine SCADA data, constructs a multi-objective optimization model with the optimization objectives of smoothing and optimizing the output power of the wind turbine generator and minimizing the control changes of the generator speed and pitch angle. A multi-objective optimization model is proposed, and an improved Gray Wolf optimization algorithm is used to solve the multi-objective optimization model and optimize the parameters of the kernel extreme learning machine. The optimal solution of the optimization model is used as the training sample and test sample of the kernel extreme learning machine, and an effective wind speed soft-sensing model based on the improved Gray Wolf optimization algorithm is established. The model uses the optimal solution of the multi-objective optimization model, i.e., the optimal predicted values ​​of the generator output power, pitch angle, and generator speed, as the model input, and the effective wind speed as the model output. The method can well predict the changing trend of the effective wind speed with high accuracy, providing an important reference indicator for the operation and control of the wind turbine. The method does not require wind tunnel experiments or virtual simulations to obtain the data required for the effective wind speed soft-sensing model, but only requires SCADA data of the actual operation of the wind turbine in the wind farm. In practical applications, the method is low-cost and easy to implement.

[0151] In summary, the present invention can perform offline identification and online self-correction, which not only has high accuracy and effectiveness, but also does not require wind tunnel experiments or virtual simulations, is low-cost, and is easy to implement.

Claims

1. An effective wind speed soft measurement method, characterized by: The signal output end of the wind turbine SCADA data acquisition module passes through the SCADA data preprocessing and normalization module, the offline optimization calculation module based on the improved gray wolf optimization IGWO algorithm, the effective wind speed soft measurement modeling module based on the kernel extreme learning machine KELM algorithm, the effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm, and the effective wind speed soft measurement model performance evaluation module, and finally outputs the effective value of the wind speed; and, The wind turbine SCADA data acquisition module collects data on wind turbine output power, generator speed, pitch angle and wind speed; The SCADA data preprocessing and normalization module preprocesses the SCADA data collected in the wind turbine SCADA data acquisition module to remove abnormal data; The offline optimization calculation module based on the improved Grey Wolf Optimization (IGWO) algorithm applies the IGWO algorithm to solve the established multi-objective optimization model to obtain the optimal predicted values ​​of the wind turbine generator output power, pitch angle, and generator speed. The specific composition is as follows: a. Construct a multi-objective optimization model with wind turbine output power, pitch angle, and generator speed as optimization targets at all wind speeds: Where, is a multi-objective function, is an equality vector constraint function, is an inequality vector constraint function; Among them, the multi-objective function is: Where, is the objective function for the actual output power of the generator to track the expected predicted value at all wind speeds, is the objective function for minimizing the control variation of generator speed and pitch angle, is the decision variable; 、 and are the actual output power, generator rotor angular velocity and pitch angle at time k respectively; 、 and They are the generator output power, generator rotor angular velocity and pitch angle in k Expected forecast value at time +1; 、 and is the weight coefficient; n is the optimization process, 、 are the rated output power and theoretical design output power of the wind turbine respectively; b. Improved Grey Wolf Optimization (IGWO) algorithm: b.

1. Improved nonlinear convergence factor Where, 、 and n are adjustment parameters, t is the current number of iterations, is the maximum number of iterations; b.

2. Improved adaptive position update equation: in, ， ， Where, It is a triangle The center of the inscribed circle of 、 and is the weight coefficient; Gray wolf individual The historical best position vector of is the position vector of the individual gray wolf with the worst fitness in the gray wolf population in the tth iteration, and are two random numbers in the interval [0,1], Gray wolf individual The learning factor, is the reverse search factor, It is a triangle The maximum side length of It is a triangle The minimum side length of and The definition is as follows: ； The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm establishes an effective wind speed soft-sensing model according to the prediction data provided by the offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm; The effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm uses the IGWO algorithm to optimize the parameters of the effective wind speed soft sensor model; The effective wind speed soft-sensing model performance evaluation module evaluates the performance of the effective wind speed soft-sensing model after parameter optimization, and provides real-time reference input for maximum wind power tracking MPPT control.

2. The effective wind speed soft-sensing method according to claim 1, wherein: The offline optimization calculation module based on the improved grey wolf optimization (IGWO) algorithm is a first microprocessor. The first microprocessor performs offline optimization calculations based on the data output by the SCADA data preprocessing and normalization module, and the obtained optimal solution is used as training sample data and test sample data for the effective wind speed soft measurement modeling module based on the kernel extreme learning machine (KELM) algorithm.

3. The effective wind speed soft-sensing method according to claim 1, wherein: The improved nonlinear convergence factor maintains its maximum value in the first half of the iterative search and its minimum value in the second half of the iterative search.

4. The effective wind speed soft-sensing method according to claim 1, wherein: The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm is a second microprocessor, and a kernel function is provided in the second microprocessor, and the kernel function is a Gaussian kernel function: Where, are the control parameters, x and are any two points in the state space under consideration.

5. The effective wind speed soft-sensing method according to claim 1, wherein: The effective wind speed soft-sensing modeling module based on the kernel extreme learning machine (KELM) algorithm calls the training sample data stored in the offline optimization calculation module based on the improved gray wolf optimization (IGWO) algorithm: the optimal predicted value of the generator output power , generator optimal speed prediction value , optimal predicted value of pitch angle , a soft measurement model of effective wind speed is established, and the output signal is transmitted to the effective wind speed soft measurement model parameter optimization module based on the IGWO algorithm.

6. The effective wind speed soft-sensing method according to claim 1, wherein: The effective wind speed soft sensor model parameter optimization module based on the IGWO algorithm is a third microprocessor, which is used to call the output signal of the effective wind speed soft sensor modeling module based on the kernel extreme learning machine (KELM) algorithm to optimize the parameters of the established soft sensor model. The optimization steps are as follows: Step 1: Preprocessing and normalization of SCADA data; Step 2: parameter initialization and normalization; Step 3: Calculate the fitness function value; Step 4: Initialize the position vectors of alpha, beta, and delta; Step 5: Set the iteration number k = 1; Step 6: Update the parameter nonlinear convergence factor a, coefficient A and coefficient C, where , , and is a random number in the interval [0, 1]; Step 7: Update the fitness function values and corresponding positions of alpha, beta, and delta; Step 8: Judge whether k is greater than or equal to Max_iter, where Max_iter is the maximum number of iterations. If k ≥ Max_iter, go to Step 9; if k < Max_iter, go to Step 6; Step 9: Return the position of the current alpha to obtain the optimal solution; Step 10: Calculate the kernel matrix based on the optimized parameters ,in , and is the training sample, n is the total number of samples; Step 11: Calculate the weight matrix of the hidden layer output; Step 12: Return the regression value and ,in, is the control parameter of the Gaussian function, is a tuning parameter.

7. The effective wind speed soft-sensing method according to claim 1, wherein: In the performance evaluation module of the effective wind speed soft sensor model, if the effective wind speed soft sensor model after parameter optimization passes the performance evaluation, output the soft sensor model parameters; if the performance evaluation fails, return to the effective wind speed soft sensor modeling module based on the Kernel Extreme Learning Machine (KELM) algorithm, retrain the samples, and re-optimize the parameters until the performance evaluation criteria are met.

Images