A short-term wind power prediction method and system for the ice-covered scenario of wind turbines

Abstract

The present invention belongs to the technical field of wind power prediction, and specifically discloses a short-term wind power prediction method and system for a wind turbine icing scenario. The method includes: constructing an icing model applicable to wind turbine blades, and assisting in solving key parameters in the icing model under set meteorological conditions through icing simulation; obtaining meteorological data and icing data in the wind turbine icing scenario within a set time range in the area to be measured, inputting the meteorological data into a trained basic prediction model to obtain a basic wind power prediction result, inputting the icing data and wind speed data into a trained power loss prediction model to obtain a power loss prediction result; and obtaining a final short-term wind power prediction result for the wind turbine icing scenario based on the basic wind power prediction result and the power loss prediction result. The present invention can improve the calculation accuracy of icing data while saving computing resources, and further improve the reliability of the power prediction result.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power prediction, and particularly to a short-term wind power prediction method and system for wind turbine icing scenarios. Background Art

[0002] The statements in this part only provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Wind energy is a clean and high-quality renewable energy source, playing an irreplaceable role in aspects such as economic development, energy supply, and reduction of greenhouse gas emissions. The healthy development of wind power is of great significance for promoting the adjustment of the energy structure.

[0004] Accurate prediction of wind power is an effective measure to cope with its randomness and volatility, and is of great significance for supporting power grid dispatching, improving the operating stability of the power system, and increasing the consumption of new energy. However, the current meteorological environment is becoming increasingly complex, and extreme weather such as freezing rain and cold snaps occurs frequently, which easily leads to icing of wind turbine blades and affects the output fluctuation characteristics. In addition, affected by factors such as latitude or altitude, the icing of wind turbine blades is also common at low temperatures, affecting the power output of wind turbines and the normal operation of wind farms.

[0005] In the above scenarios, traditional wind power prediction models are often constructed based on numerical weather forecasts and historical operation data, without considering the impact of icing scenarios, so there may be large prediction deviations and threaten the safe and reliable operation of the power system. Summary of the Invention

[0006] To solve the above problems, the present invention proposes a short-term wind power prediction method and system for wind turbine icing scenarios. By quantifying and calculating icing data and considering wind power data under the condition of wind turbine blade icing, the accuracy of wind power prediction can be improved, and the ability of the new power system to cope with wind power volatility can be enhanced.

[0007] In some embodiments, the following technical solutions are adopted:

[0008] A short-term wind power prediction method for wind turbine icing scenarios, comprising:

[0009] Considering the ablation and shedding effects of ice, constructing an icing model applicable to wind turbine blades, and solving the key parameters in the icing model under set meteorological conditions through icing simulation assistance;

[0010] Converting discrete icing data into continuous icing data corresponding to meteorological conditions at different times through data interpolation;

[0011] Train the basic prediction model for wind power prediction and the power loss prediction model considering the ice-covered scenario of the wind turbine respectively;

[0012] Obtain the meteorological data and ice-covered data under the ice-covered scenario of the wind turbine within a set time range in the area to be measured. The meteorological data is input into the trained basic prediction model to obtain the basic wind power prediction result, and the ice-covered data and wind speed data are input into the trained power loss prediction model to obtain the power loss prediction result;

[0013] Based on the basic wind power prediction result and the power loss prediction result, obtain the final short-term wind power prediction result for the ice-covered scenario of the wind turbine.

[0014] Furthermore, considering the ablation and shedding effects of ice, construct an ice-covered model applicable to the wind turbine blade, specifically:

[0015] ;

[0016] ;

[0017] ;

[0018] Among them, is the cumulative ice-covered degree of a single wind turbine when considering the ablation and shedding effects of ice, is the increase in the maximum ice thickness at the i moment when the ablation and shedding effects of ice are not considered, is the proportionality coefficient of ice reduction quantifying the ablation and shedding effects of ice, represents the increase in the maximum ice thickness at the i moment after considering the ablation and shedding effects of ice, represents i the maximum ice-covered length of the airfoil at the ; c represents the blade chord length, and n represents the set time duration; are the collision coefficient and the freezing coefficient respectively, is the mass fraction of water droplets in the air, is the movement speed of the water droplets, is the unit time, is the density of the ice accretion, is the proportionality coefficient when solving the ice accretion volume.

[0019] Furthermore, the mass fraction of water droplets in the air is specifically:

[0020] ;

[0021] Among them, Q and d represent the cloud water mixing ratio and the air density respectively.

[0022] Furthermore, the key parameters in the icing model under set meteorological conditions are solved by means of icing simulation assistance, specifically as follows:

[0023] Construct a physical model of the wind turbine blade, use Fluent software to calculate the air flow field around the wind turbine blade, import the flow field calculation results into the icing simulation software, preset the median diameter of water droplets and the liquid water content in the air, and calculate the collision coefficient respectively through the simulation of the water droplet movement trajectory and icing simulation , freezing coefficient , the proportional coefficient when solving the ice accretion volume and the proportional coefficient of ice accretion reduction .

[0024] Furthermore, the discrete icing data is converted into continuous icing data corresponding to meteorological conditions at different times through data interpolation, specifically as follows:

[0025] Estimate the values of the icing parameters α1 and α3 under any meteorological data through the Kriging interpolation method, and then calculate the icing data under meteorological data at different times.

[0026] Furthermore, a basic prediction model for wind power prediction and a power loss prediction model considering the wind turbine icing scenario are trained respectively, specifically as follows:

[0027] Construct datasets D1 containing conventional meteorological data and corresponding wind power data, datasets D2 and D3 containing meteorological data and corresponding wind power data under icing scenarios respectively;

[0028] Use dataset D1 to train the basic prediction model to obtain a trained basic prediction model;

[0029] Use the trained basic prediction model to predict the wind power of dataset D2, and take the difference between the obtained wind power prediction value and the actual value, which is recorded as the power loss; the power loss and the corresponding wind speed and icing data form a power loss dataset D4;

[0030] Use dataset D4 to train the power loss prediction model to obtain a trained power loss prediction model;

[0031] Use dataset D3 to test the trained basic prediction model and power loss prediction model.

[0032] Furthermore, the basic prediction model selects the LightGBM model, and the power loss prediction model selects the Few-shot model.

[0033] In some other embodiments, the following technical solutions are adopted:

[0034] A short-term wind power prediction system for the icing scenario of wind turbines, comprising:

[0035] An icing model simulation and solution module, which is used to consider the ablation and shedding effects of ice, construct an icing model applicable to wind turbine blades, and assist in solving the key parameters in the icing model under set meteorological conditions through icing simulation;

[0036] An icing data interpolation module, which is used to convert discrete icing data into continuous icing data corresponding to meteorological conditions at different times through data interpolation;

[0037] A model training module, which is used to train the basic prediction model for wind power prediction and the power loss prediction model considering the icing scenario of wind turbines respectively;

[0038] A power prediction module, which is used to obtain the meteorological data and icing data under the icing scenario of wind turbines within a set time range in the area to be measured. The meteorological data is input into the trained basic prediction model to obtain the basic wind power prediction result. The icing data and wind speed data are input into the trained power loss prediction model to obtain the power loss prediction result; based on the basic wind power prediction result and the power loss prediction result, the final short-term wind power prediction result for the icing scenario of wind turbines is obtained.

[0039] In some other embodiments, the following technical solutions are adopted:

[0040] A terminal device, which includes a processor and a memory. The processor is used to implement instructions; the memory is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to perform the above-mentioned short-term wind power prediction method for the icing scenario of wind turbines.

[0041] In some other embodiments, the following technical solutions are adopted:

[0042] A computer-readable storage medium, in which multiple instructions are stored, and the instructions are suitable for being loaded and executed by the processor of the terminal device to perform the above-mentioned short-term wind power prediction method for the icing scenario of wind turbines.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] (1) The present invention fully considers processes such as ice ablation and shedding, constructs a calculation formula for the icing degree of wind turbine blades, and assists in obtaining the key icing parameters through icing simulation; uses Kriging interpolation to expand the icing parameters and establish a mapping relationship from meteorological data to icing parameters, so as to realize the acquisition of icing data under meteorological conditions at any time; it can improve the calculation accuracy of icing data and then improve the reliability of the power prediction result while saving computing resources.

[0045] (2) The present invention separately constructs a basic prediction model for wind power considering the influence of icing and a power loss prediction model. According to the icing degree data, the power loss of wind power output under the icing scenario is obtained, and the conventional wind power prediction power is corrected, improving the prediction accuracy. Thus, the ability of the new energy system to cope with the uncertainty of wind power under extreme icing scenarios is enhanced, ensuring the stable operation of the system.

[0046] Other features and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of this aspect. Brief Description of the Drawings

[0047] Figure 1 It is a flowchart of the short-term wind power prediction method for the wind turbine icing scenario in the embodiment of the present invention;

[0048] Figure 2 It is a schematic diagram of the definition of icing degree in the embodiment of the present invention;

[0049] Figure 3 It is a flowchart of the icing simulation of the wind turbine blade in the embodiment of the present invention;

[0050] Figure 4 It is a simulation result diagram of the collision coefficient under different meteorological conditions in the embodiment of the present invention;

[0051] Figure 5 It is a simulation result diagram of the freezing coefficient under different meteorological conditions in the embodiment of the present invention;

[0052] Figure 6 It is a schematic diagram of the change of icing degree in the icing event in the embodiment of the present invention;

[0053] Figure 7 It is the wind power prediction result of icing event A in the embodiment of the present invention;

[0054] Figure 8 It is the wind power prediction result of icing event B in the embodiment of the present invention;

[0055] Figure 9 It is the wind power prediction result of icing event C in the embodiment of the present invention. Detailed Embodiments

[0056] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.

[0057] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should also be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0058] Embodiment 1

[0059] In one or more embodiments, a short-term wind power prediction method for a wind turbine icing scenario is disclosed, combined with Figure 1 , and specifically includes the following process:

[0060] S101: Considering the ablation and shedding effects of ice, construct an icing model applicable to wind turbine blades, and assist in solving the key parameters in the icing model under set meteorological conditions through icing simulation.

[0061] In this embodiment, first, the ratio of the maximum icing thickness of the airfoil to the chord length of the wind turbine is defined as the icing degree of the wind turbine blade, and a specific expression form applicable to the icing of the wind turbine blade is derived based on the Makkonen icing model.

[0062] In a low-temperature and humid environment, the air will contain more supercooled water droplets. Supercooled water droplets refer to water droplets with a temperature below 0°C but still not frozen. The supercooled water droplets move with the wind. When they collide with the wind turbine blade, some of the supercooled water droplets are captured by the blade, and the captured water droplets release heat and freeze, thus generating an ice layer on the surface of the wind turbine blade. The icing of the wind turbine blade is affected by meteorological conditions. The meteorological factors with greater influence include temperature, air humidity (in this embodiment, it is reflected by the median volume diameter MVD of supercooled water droplet particles and the liquid water content LWC in the air), and the speed and direction of the wind. Existing literature shows that the necessary conditions for the blade to generate icing are: the environmental temperature is below 0°C, and the air humidity reaches more than 85%.

[0063] The adverse effects of wind turbine blade icing on wind power prediction are mainly reflected in: ① Icing changes the blade shape, affects its aerodynamic performance, and the output power of the wind turbine decreases accordingly. Therefore, the actual power under the icing state is lower than the power predicted by the conventional prediction model; ② Icing increases the load on the wind turbine blade, but due to the irregularity of the icing, the increased load is unbalanced, resulting in changes in the amplitude and frequency of the blade, increasing the operation risk of the wind turbine, and in severe cases, causing the wind turbine to shut down, further expanding the prediction error.

[0064] The prior art has proposed an icing model for the surface of an object, which decomposes the icing process into three stages: the collision of supercooled water droplets in the air with the object, the capture of supercooled water droplets by the object surface, and the freezing of liquid water droplets. The Makkonen icing model is as follows:

[0065] (1)

[0066] In the formula, α1, α2, and α3 are coefficients representing the behavior of supercooled water droplets in the three stages of the icing model, respectively. Specifically, the collision coefficient α1 is the ratio of the amount of water droplets actually colliding with the wind turbine blade to the amount of all water droplets that can collide. This coefficient is related to the wind speed and the liquid water content. The capture rate α2: When water droplets collide with the wind turbine blade, they may bounce or splash and leave the blade surface. Therefore, a parameter needs to be introduced to represent the amount of water droplets remaining on the blade surface, that is, the capture rate, which represents the ratio of the amount of water droplets remaining on the blade surface to the amount of all colliding water droplets. This coefficient is related to the object material. The freezing coefficient α3 represents the ratio of the amount of water droplets frozen on the wind turbine blade surface to the amount of captured water droplets. This coefficient is directly affected by temperature and is also related to the liquid water content. is the mass fraction of water droplets in the air, that is, the liquid water content LWC in the air; is the movement speed of the water droplets, and A is the cross-sectional area of the object that undergoes collision.

[0067] The Makkonen icing model can be applied to the scenarios of conductor icing and aircraft wing icing. However, the calculation formula for the scenario of wind turbine blade icing is still blank.

[0068] This embodiment will start from the Makkonen model and derive its specific expression in the scenario of wind turbine blade icing.

[0069] Combined with the numerical weather prediction (NWP) data with a resolution of = 15 min, this time is defined as a unit time. It is approximately considered that the values of each meteorological factor and model parameter remain unchanged within a unit time. Then, on the area of unit area the increased ice mass within a unit time is:

[0070] (2)

[0071] During the icing process, supercooled water droplets generally do not bounce back. Therefore, α2 is usually considered to be equal to 1. The numerical solution methods for α1 and α3 are complex processes. This embodiment uses simulation to assist in obtaining the values of the above parameters under different meteorological conditions. In the application scenario of this embodiment, v is approximately equal to the wind speed. Since the liquid water content in the air Outside the NWP forecast range, it is necessary to calculate and solve using the output physical quantities of the WRF forecast model. The relevant formulas are as follows:

[0072] (3)

[0073] In the formula, Q and d represent the cloud water mixing ratio and air density respectively, both of which can be obtained from the WRF output.

[0074] As Figure 2 shown, the ratio of the maximum ice accretion thickness of the fan airfoil to the chord length is defined as the ice accretion degree, that is:

[0075] (4)

[0076] In the formula, I represents the ice accretion degree of the blade, represents the maximum ice accretion length of the airfoil, and c represents the blade chord length.

[0077] Existing research shows that the position with the most serious ice accretion appears at the leading edge of the blade tip. When dealing with it, it can be approximately considered that the ice accretion thickness is the largest at 95% of the blade span, and only the ice accretion condition at this position is analyzed.

[0078] Assume that the ice accretion thickness is approximately equal within the area near it, then there is the following relationship:

[0079] (5)

[0080] In the formula, is the density of the ice accretion, is = the increase in the ice accretion thickness at the position under study within 15 minutes, is the proportionality coefficient when solving the ice accretion volume, which is determined by the shape of the ice accretion. Due to the irregular shape of the ice accretion, in this embodiment the value of is obtained from the statistical experience of a large number of ice accretion simulations. Combining formula (2) and formula (5), an expression for the increase in ice accretion thickness per unit time can be obtained as:

[0081] (6)

[0082] Then the cumulative ice accretion degree is:

[0083] (7)

[0084] Among them, is the increase in ice accretion thickness at the maximum ice accretion position per unit time, and n is the total duration.

[0085] The numerical results of the ice accretion degree of a single wind turbine at each moment can be obtained through the above formula. In this embodiment, the following simplification is made: it is considered that within a wind farm, each wind turbine faces the same meteorological conditions. Therefore, the calculation result of the ice accretion degree of a single wind turbine can be used as the average ice accretion degree of all wind turbines in the wind farm.

[0086] As can be seen from the above analysis, the key to calculating the ice accretion degree lies in obtaining the key parameters α1 and α3. In this embodiment, the ice accretion simulation method is used to assist in solving the key ice accretion parameters.

[0087] The ice accretion process is a complex physical process that needs to consider fluid motion and heat exchange. Two stages are considered during the simulation:

[0088] (1) Supercooled water droplets move under the action of wind, hit the surface of the wind turbine blade and are captured, and heat exchange occurs with the blade surface and the surrounding air.

[0089] (2) Some of the captured water droplets are frozen on the blade surface to form ice accretion, while the unfrozen water droplets will move on the blade surface or be evaporated along with the flow field. As an alternative to the experimental method, modeling and simulation analysis provide a data basis for ice accretion analysis by reproducing the ice accretion process on the wind turbine blade.

[0090] FENSAP-ICE is a commonly used ice accretion simulation software based on the Navier-Stokes (RANS) equation, which can obtain relatively accurate results. However, since it is more suitable for analyzing the final ice accretion results and ignores the ice accretion process, in this embodiment, when using this software for ice accretion simulation, it is only used to implement the ice accretion parameters in the Makkonen ice accretion model α 1 and α 3 for auxiliary calculation, study the main influencing factors of ice accretion - factors such as temperature, wind speed, liquid water content, and median diameter of water droplets, etc. on the ice accretion parameters, and obtain the calculation results of the ice accretion parameters under different combinations of meteorological element values. Follow the steps in Figure 3 to conduct the simulation of ice accretion.

[0091] To simulate the icing process, it is necessary to first calculate the air flow field around the wind turbine blade, and then calculate the movement trajectory of water droplets in the flow field to determine the amount of supercooled water droplets collected on the blade surface. Since the flow field calculation module of FENSAP-ICE has a relatively single function and a slow calculation speed, Fluent software is first used for the flow field calculation. After the calculation is completed, the results are imported into FENSAP-ICE for subsequent calculation and analysis of the water droplet movement trajectory and icing. When performing icing simulation, considering that the liquid water droplets in the air in reality do not have a single particle size distribution, the median volume diameter (MVD) of the water droplets and the liquid water content (LWC) in the air are set in advance. The Custom distribution model provided in FENSAP-ICE is used to customize the median volume diameter of the liquid water droplets and their proportion, making it follow a normal distribution. After setting the values of other remaining meteorological factors and the icing time, the software is run to obtain the simulation results of the corresponding icing parameters. It should be noted that the airfoils of different wind turbines will vary, and the specific wind turbine should be selected according to the actual situation during icing simulation. In this embodiment, the wind turbine model selected for the target wind farm is GW115 / 2.0MW, and the airfoil used for icing simulation is NACA4412.

[0092] In addition, a complete icing process should also consider processes such as ice ablation and shedding. However, the Makkonen model does not consider the above processes. In this embodiment, by introducing a proportionality coefficient to characterize the reduction process of icing, it is considered that the rate of ice shedding is proportional to the existing ice amount, and is taken as a fixed value.

[0093] Furthermore, when considering the ice ablation and shedding effects, the increase in ice amount per unit time can be expressed by the following formula:

[0094] (8)

[0095] Correspondingly, when considering the ice ablation and shedding effects, the formula for calculating the icing degree at each moment is:

[0096] (9)

[0097] Icing simulation is used to assist in solving and , and a series of simulation results of icing thickness are obtained, forming an overdetermined system of equations. By using the least squares method, the approximate values of and that minimize the sum of the squares of the errors of all equations are found.

[0098] S102: Convert the discrete icing data into continuous icing data corresponding to meteorological conditions at different times through data interpolation.

[0099] As described above, by setting specific values of icing meteorology, the icing model parameters under the corresponding meteorological environment can be obtained through the FENSAP-ICE software, and then the icing thickness and the degree of icing can be calculated. However, only the icing results under discrete conditions can be obtained in this way. In fact, meteorological factors change continuously. When predicting, obtaining the icing results under different predicted meteorological conditions through simulation requires a large amount of time and computing resources, and it is difficult to meet the requirements of prediction in terms of real-time performance and usability. Therefore, in this embodiment, the Kriging interpolation algorithm is used to realize the transformation from discrete results to continuous space, and further combined with meteorological data, the time series of the degree of icing at different times in the icing event can be calculated.

[0100] The Kriging interpolation algorithm is a spatial interpolation method based on the theory of variograms and is used to achieve the optimal error estimation of variables within a limited area. This interpolation method is not only based on the spatial positions of the points to be interpolated and adjacent sampling points, but also takes into account the mutual positional relationships of these sampling points, makes full use of the distribution characteristics of existing observations, and realizes a more accurate and practical estimation than traditional interpolation methods. By extending the concept of "space" to "numerical space", Kriging interpolation can be used to achieve the interpolation estimation of high-dimensional data.

[0101] The formula of the Kriging interpolation regression model is as follows:

[0102] (10)

[0103] In the formula, x0 is the point to be interpolated, and Z * (x0) is the result of its interpolation estimation, x i (i = 1, 2, ……, n) are the known sampling points within the study area, and Z(x i ) is its sampling value, and λ i is the undetermined weight coefficient. This formula indicates that the Kriging interpolation result is a weighted combination of the sampling values of other known sampling points within the study area.

[0104] The weight coefficient can be calculated by the following formula:

[0105] (11)

[0106] In the formula, C (x i , x j ) is the covariance function of the sampling values at the sampling points x i , x j , and μ is the Lagrange multiplier introduced when minimizing the estimated variance.

[0107] In this embodiment, a high-dimensional numerical space is composed of meteorological factors such as wind speed, temperature, and air humidity, and Kriging interpolation is used to estimate the values of the icing parameters α1 and α3 in this data space.

[0108] After the above Kriging interpolation, the ice accretion parameters α1 and α3 under any meteorological value can be obtained. Furthermore, the corresponding ice accretion data can be calculated based on the meteorological data at different times, providing data support for the wind power prediction under the subsequent wind turbine ice accretion scenario.

[0109] S103: Train the basic prediction model for wind power prediction and the power loss prediction model considering the wind turbine ice accretion scenario respectively.

[0110] In this embodiment, historical real data is collected and three data sets, namely the training set D1, the validation set D2, and the prediction set D3, are constructed respectively. The training set D1 contains conventional meteorological data and wind power data; both the validation set D2 and the prediction set D3 contain wind power data under the ice accretion scenario and their corresponding meteorological data.

[0111] Training of the basic prediction model: Taking meteorological data as input and wind power data as output, train the LightGBM model on the data set D1 as the basic prediction model for wind power prediction.

[0112] Use the trained LightGBM model to predict the wind power of the validation set D2. The difference between the predicted power and the actual power is recorded as the power loss, which forms the power loss sample set D4 together with the corresponding wind speed data and ice accretion data.

[0113] Taking ice accretion data and wind speed as input and power loss as output, train the Few-shot model on the power loss sample set D4 to construct the mapping relationship between the ice accretion degree, wind speed, and power loss; obtain the trained power loss prediction model.

[0114] First, use the trained LightGBM model to predict the wind power in the prediction set D3 to obtain the basic wind power prediction result; at the same time, use the trained Few-shot model to predict the power loss of the prediction set D3; finally, subtract the power loss prediction result from the basic wind power prediction result to obtain the final wind power prediction result considering the influence of ice accretion.

[0115] After the above training and prediction processes, the trained basic prediction model and power loss prediction model are obtained.

[0116] In this embodiment, LightGBM (Light Gradient Boosting Machine) is an efficient ensemble learning algorithm based on the gradient boosting framework. It is an improved version of XGBoost, and its design goal is to improve the training speed and efficiency while maintaining high accuracy. The principle of LightGBM in regression tasks is mainly based on the algorithm of Gradient Boosting Decision Tree (GBDT). It approximates the minimization process of the objective function by gradually adding decision trees. Each newly added tree aims to reduce the residuals left by the previous tree, thereby improving the accuracy of the model.

[0117] The total loss of LightGBM consists of two parts: the loss function and the regularization term. Suppose the training dataset is , where, y i refers to wind power, x i refers to meteorological elements related to wind power (such as wind speed, temperature, humidity, etc.); the predicted value of the model in the t-th round is .

[0118] Then the total loss can be expressed as:

[0119] (12)

[0120] (13)

[0121] In the formula, loss is the loss function, and the square loss can be used. is the part for regularizing each tree, mainly used to prevent overfitting and ensure the generalization ability of the model. N L is the number of leaf nodes of the tree, w j is the predicted value of the j-th leaf node, and are regularization parameters, which control the complexity of the tree and the sum of squares of leaf weights respectively.

[0122] In each iteration, the model adds a new tree f t to reduce the residuals. The new tree is fitted according to the negative gradient, and this step is similar to updating the parameters in the gradient descent method:

[0123] (14)

[0124] In the formula, is the learning rate, is the predicted value of the new tree for the sample x i .

[0125] The tree f tThe goal is to fit the residual between the prediction of the previous-round model and the true value. For the squared loss function, the negative gradient is:

[0126] (15)

[0127] Using the negative gradient r it as the target value, learn f t on the training data through the decision tree algorithm. The construction of each tree focuses on maximizing the gain after data splitting, while considering the regularization term to avoid overfitting. Repeat the above process until the set maximum number of trees is reached.

[0128] In this way, each tree is trying to correct the error left by the previous tree, and the accuracy of the entire model gradually improves with the number of iterations. The optimization and efficient implementation of LightGBM enable this process to run quickly even on large datasets.

[0129] In this embodiment, the Few-shot few-sample learning algorithm is implemented based on Meta-Learning. Meta-Learning is an important branch in machine learning, aiming to improve the generalization ability and adaptability of algorithms by learning how to learn. Its core idea is to use the existing task experience to enhance the learning efficiency and effect of new tasks. The goal of Meta-Learning is to enable the model to quickly adapt to new tasks and reduce the dependence on large-scale data and long-time training. Since there is less icing data, the sample of the constructed error dataset is small, and it is difficult for traditional artificial intelligence algorithms to play their advantages. Therefore, it is considered to use Meta-Learning to construct a few-sample learning model to achieve relatively accurate error prediction based on a small number of icing samples.

[0130] Meta-Learning can be divided into Model-Based Methods and Optimization-Based Methods. Model-Agnostic Meta-Learning (MAML) is an optimization-based meta-learning method. It can optimize the initial parameters of the model so that when the model encounters a new task, it can quickly adapt and perform well with only a small amount of training data and very few gradient updates. Since MAML does not depend on a specific model architecture, it is applicable to a wide range of task types and can effectively share knowledge among multiple tasks, thus greatly improving the generalization ability of the model on new tasks, especially in the case of scarce data.

[0131] Specifically, first divide the dataset into multiple tasks, each task contains k groups of samples, and the task data is denoted as . Among them, represents the power loss, and here i= 1 because it is one-dimensional, and j represents the group number of the samples; represents the data of wind speed and icing degree (affecting power loss). Here, i = 2 because it is two-dimensional, and j represents the group number of the samples.

[0132] Define an LSTM model for implementing non-linear regression, and at the same time use the MAML algorithm for model training to achieve few-shot learning.

[0133] The training process is divided into two parts: the inner loop (Inner Loop) and the outer loop (Outer Loop).

[0134] The inner loop can be understood as performing a small-scale training on each task to improve the performance of the model on that task. Specifically, select a task T from the task set i , and its training data is , calculate the loss function of the model on the training data of this task.

[0135] (16)

[0136] In the formula, θ is the initial model parameter.

[0137] According to the gradient of the loss function, update the model parameters to optimize the performance of this task:

[0138] (17)

[0139] In the formula, is the inner loop learning rate, is the updated model parameter on task T i .

[0140] The goal of the inner loop is to enable the model to quickly adapt to the training data of each specific task. After the inner loop ends, the model parameter θ i ' has undergone a task-specific update and the performance on this task has been improved.

[0141] The purpose of the outer loop is to aggregate information across multiple tasks, thereby adjusting the initial model parameter θ so that the model can achieve good performance with a small number of inner loop updates when facing new tasks. This is equivalent to finding a good initial point so that fast learning can be achieved on different tasks.

[0142] Specifically, for each task T i , use the parameter θ i ' updated by the inner loop to calculate the validation loss on the validation set of this task:

[0143] (18)

[0144] The verification losses of all tasks are summed to obtain the global meta-loss, and based on this, the initial model parameters θ are updated as follows:

[0145] (19)

[0146] where is the outer-loop learning rate.

[0147] Due to the different sizes of the processed data samples, the number of iterations of the inner loop and the outer loop also varies. Generally, the number of inner-loop iterations can be set to 1 - 5 times, while the outer loop is set to dozens to hundreds of times.

[0148] S104: When making an actual prediction, obtain the meteorological data and icing data under the fan icing scenario within a set time range for the area to be measured. The meteorological data is input into the trained basic prediction model to obtain the basic wind power prediction result. The icing data and wind speed data are input into the trained power loss prediction model to obtain the power loss prediction result;

[0149] S105: Based on the basic wind power prediction result and the power loss prediction result, obtain the final short-term wind power prediction result for the fan icing scenario.

[0150] In this embodiment, a certain wind farm in the Northeast region is taken as the research object, and a case study is set up to verify the effectiveness of the proposed method. The data used includes NWP data from October 1, 2020 to February 28, 2021, measured data of the wind farm's wind speed and wind power, and icing records provided by the wind farm. The dataset is divided according to Table 1.

[0151] Table 1 Dataset division

[0152]

[0153] First, the key parameters of the icing model are solved by assisting with icing simulation. The settings of relevant variables in the icing simulation are shown in Table 2.

[0154] Table 2 Input meteorological factor values

[0155]

[0156] The collision coefficient is jointly affected by the wind speed and the liquid water content, while the freezing coefficient is mainly affected by the temperature and the liquid water content. The icing simulation is carried out according to the set meteorological factor values in Table 2, and the simulation results of the obtained icing parameters are as shown in Figure 4 and Figure 5 shown. From Figure 4It can be seen that the collision coefficient is approximately proportional to both the wind speed and the liquid water content. At the same liquid water content, the higher the wind speed, the more water droplets will collide with the fan blades, and correspondingly, the greater the collision coefficient. And it can be seen from Figure 5 that the freezing coefficient does not have a simple linear relationship with meteorological variables. When the temperature is low enough, such as -10°C and below, almost all the water droplets remaining on the blade surface can freeze, so the freezing coefficient can reach a level close to 1; while when the temperature is higher, due to the existence of a "saturation" of liquid water droplets on the blade that cannot freeze all the water droplets, as the liquid water content in the air increases, the proportion of water droplets that can be frozen decreases, so the freezing coefficient decreases accordingly.

[0157] To reduce time resources and computing resources, the present invention proposes to use Kriging interpolation to obtain the icing parameter results in the continuous meteorological numerical space. Taking the example of obtaining the icing parameters at 96 time points in a day, the comparison of the computing time required by the interpolation method and the traditional simulation method is shown in Table 3.

[0158] Table 3 Comparison of computing time

[0159]

[0160] The comparison between the predicted values obtained by interpolation and the "actual values" obtained by simulation in the high-dimensional data space is given in Table 4. The results show that the error of the icing parameters obtained by the interpolation method is small compared with the simulation results, that is, the Kriging interpolation method can better realize the expansion of the icing parameters from discrete values to the continuous numerical space, thereby replacing the simulation work and effectively saving time and computing resources.

[0161] Table 4 Comparison between interpolation results and actual values

[0162]

[0163] Using the method proposed by the present invention to calculate the icing degree at each moment, Figure 6 shows the changes in the icing degree of several icing events that occurred in February 2021. According to the characteristics of the curve, it can be further divided into three types: rapid icing growth, slow icing growth, and icing first and then melting.

[0164] It can be seen from the figure that the icing change processes on different dates have great differences in terms of growth rate, change trend, etc., and the method proposed by the present invention can reflect the phenomena of icing ablation and shedding. The change process of icing is mainly affected by meteorological elements. Combining the aforementioned icing parameter interpolation results, after calculating the icing on different dates and comparing the results, the following influencing laws can be summarized:

[0165] (1)Effect of temperature: As the temperature decreases, the leaf freezing coefficient increases, the amount of water droplets freezing after impact increases, and the increase in ice accretion within the same time period becomes larger. However, as the temperature decreases, a lower temperature will turn the water droplets with a smaller median volume in the air into ice crystals, resulting in fewer water droplets hitting the leaf surface, and the increase in ice accretion within the same time period will instead decrease. Combining the calculation results of the ice accretion parameters, the turning point appears at around -10°C.

[0166] (2)Effect of wind speed: The ice accretion mass on the leaf surface increases with the increase in wind speed. The main reason is that the wind speed directly affects the water droplet collection amount on the leaf surface: when the wind speed is small, the inertial force of the water droplets is relatively small compared to the viscous force they receive, so fewer water droplets hit the leaf surface; as the wind speed increases, the inertial force gradually increases, and more water droplets hit the leaf surface, that is, the greater the wind speed, the more supercooled water droplets are blown towards the leaf, and when the temperature is low enough, the ice accretion amount on the leaf is more.

[0167] (3)Effect of liquid water content: When other ice accretion conditions are the same, the larger the LWC, the more water droplets hit the leaf surface per unit time. If the environmental temperature is low enough at this time, the ice accretion amount on the leaf surface increases faster.

[0168] To further verify the effectiveness of the prediction method proposed in this embodiment, three ice accretion events A, B, and C in Figure 6 are selected, and the power prediction and correction are carried out using the method proposed in the present invention. The wind power prediction results of the three ice accretion events are as shown in Figures 7 - 9 . It can be seen from this that after error correction, the predicted power value decreases, but it is closer to the true power in most time periods, that is, the method proposed in the present invention can reflect the influence of the power loss caused by ice accretion on the fan blades when predicting wind power.

[0169] Combined with the ice accretion calculation results for analysis, the calculated ice accretion degree in ice accretion event A generally shows a trend of first increasing and then decreasing, and the corresponding change trend of the power loss value is generally the same. When the ice accretion degree is low, the power loss is low, and when the ice accretion degree is high, the power loss also reaches a relatively high level. Although the calculated ice accretion degree in ice accretion event B shows a continuous increasing trend, its power loss does not strictly change in a gradually increasing trend. This is because the size of the power loss value is not only determined by the ice accretion degree, but also affected by factors such as wind speed. In this ice accretion event, when the ice accretion degree is low in the early stage, the size of the power loss is mainly determined by the wind speed; while in the later stage, although the ice accretion degree increases to a relatively high level, due to the low wind speed at this time and the low wind power itself, the power loss is not large. The calculated ice accretion degree in ice accretion event C has an obvious trend of rapid increase, and at this time the ice accretion degree plays a greater role in the power loss, so the power loss also generally conforms to the gradually increasing trend.

[0170] Based on the above analysis, the impact of icing on wind turbine blades on wind power can be summarized as follows: ① Blade icing will cause the output power of the wind turbine to decrease, resulting in "power loss". Under the same wind speed conditions, the greater the icing degree, the greater the power loss value; ② The magnitude of the power loss value is affected by both the icing degree and the wind speed conditions. When the wind speed is relatively high, a relatively low icing degree will also cause a large power loss. Under the same icing degree, the higher the wind speed, the greater the percentage of the power loss value.

[0171] The normalized root mean square error (NRMSE) and the normalized mean absolute error (NMAE) are selected as the prediction evaluation indicators, and the prediction errors are given in Table 5.

[0172] Table 5 Comparison of prediction errors

[0173]

[0174] It can be seen from Table 5 that after the impact of icing is added to the prediction model by using the method proposed in the present invention, the prediction accuracy of wind power has been improved, and the NRMSE and NMAE are reduced by 0.0397 and 0.0439 respectively compared with the case without considering the impact of icing.

[0175] The method of this embodiment improves the prediction accuracy, thereby enhancing the ability of the new energy system to cope with the uncertainty of wind power under extreme icing scenarios and ensuring the stable operation of the system.

[0176] Embodiment 2

[0177] In one or more embodiments, a short-term wind power prediction system for wind turbine icing scenarios is disclosed, including:

[0178] An icing model simulation and solution module, which is used to consider the ablation and shedding effects of ice, construct an icing model applicable to wind turbine blades, and assist in solving the key parameters in the icing model under set meteorological conditions through icing simulation;

[0179] An icing data interpolation module, which is used to convert discrete icing data into continuous icing data corresponding to meteorological conditions at different times through data interpolation;

[0180] A model training module, which is used to train the basic prediction model for wind power prediction and the power loss prediction model considering wind turbine icing scenarios respectively;

[0181] A power prediction module is used to obtain meteorological data and icing data in the scenario of wind turbine icing within a set time range in the area to be measured. The meteorological data is input into a trained basic prediction model to obtain a basic wind power prediction result. The icing data and wind speed data are input into a trained power loss prediction model to obtain a power loss prediction result. Based on the basic wind power prediction result and the power loss prediction result, a final short-term wind power prediction result for the wind turbine icing scenario is obtained.

[0182] It should be noted that the specific implementation manners of the above modules are exactly the same as those in Embodiment 1 and will not be elaborated here.

[0183] Embodiment 3

[0184] In one or more embodiments, a terminal device is disclosed, which includes a processor and a memory. The processor is used to implement instructions; the memory is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to perform the short-term wind power prediction method for the wind turbine icing scenario described in Embodiment 1.

[0185] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0186] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0187] In the implementation process, each step of the above method may be completed by the integrated logic circuit in the hardware of the processor or the instructions in the form of software.

[0188] Embodiment 4

[0189] In one or more embodiments, a computer-readable storage medium is disclosed, in which multiple instructions are stored, and the instructions are suitable for being loaded and executed by the processor of the terminal device to perform the short-term wind power prediction method for the wind turbine icing scenario described in Embodiment 1.

[0190] Although the specific implementation manners of the present invention are described above in conjunction with the drawings, it is not a limitation to the protection scope of the present invention. Those skilled in the art should understand that based on the technical solutions of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A short-term wind power prediction method for wind turbine icing scenarios, characterized in that: include: Considering the melting and shedding effects of ice, an icing model suitable for wind turbine blades is constructed, and the key parameters of the icing model under set meteorological conditions are solved through icing simulation. Combined with numerical weather forecast (NWP) data =15min resolution, It is defined as a unit time. It is assumed that the values ​​of meteorological factors and model parameters remain unchanged within a unit time. In the area of ​​, the increased ice mass per unit time is: The ice melting and shedding effects are considered to construct an ice covering model suitable for wind turbine blades, specifically: ； ； ； in, To consider the ice melting and shedding effect, the accumulated ice coverage of a single wind turbine is When the effects of ice melting and shedding are not considered i The increase in the maximum ice thickness at the moment, To quantify the proportionality coefficient of ice reduction between ice ablation and shedding effect, After considering the effects of ice melting and shedding i The increase in the maximum ice thickness at the moment, express i The maximum ice coverage length of the airfoil at that moment; c represents the blade chord length, and n represents the set time; , are the collision coefficient and the freezing coefficient respectively, represents the capture rate, is the mass fraction of water droplets in the air, is the velocity of the water drop, is the unit time, is the density of ice accumulation, To solve for the proportionality factor when solving for the volume of ice accumulation; The key parameters of the icing model under set meteorological conditions are solved through icing simulation, specifically: Construct a physical model of the fan blades, use Fluent software to calculate the air flow field around the fan blades, import the flow field calculation results into the icing simulation software, preset the median diameter of the water droplets and the liquid water content in the air, and calculate the collision coefficient through the water droplet motion trajectory simulation and icing simulation. , freezing coefficient , solve for the proportionality coefficient when solving for the ice accumulation volume and the proportionality factor of ice reduction ; The discrete ice cover data are converted into continuous ice cover data corresponding to meteorological conditions at different times through data interpolation; Train the basic prediction model for wind power prediction and the power loss prediction model considering wind turbine icing scenarios respectively; Obtain meteorological data and icing data under the wind turbine icing scenario within a set time range in the test area, input the meteorological data into a trained basic prediction model to obtain a basic prediction result of wind power, and input the icing data and wind speed data into a trained power loss prediction model to obtain a power loss prediction result; Based on the basic wind power prediction results and power loss prediction results, the final short-term wind power prediction results for wind turbine icing scenarios are obtained.

2. A short-term wind power prediction method for wind turbine icing scenarios according to claim 1, characterized in that: Mass fraction of water droplets in air Specifically: ； Where Q and d represent the cloud water mixing ratio and air density, respectively.

3. The short-term wind power prediction method for wind turbine icing scenarios according to claim 1, characterized in that: The discrete ice cover data is converted into continuous ice cover data corresponding to meteorological conditions at different times through data interpolation, specifically: The values ​​of icing parameters α1 and α3 under any meteorological data are estimated by Kriging interpolation method, and then the icing data under meteorological data at different times are calculated.

4. The short-term wind power prediction method for wind turbine icing scenarios according to claim 1, characterized in that: The basic prediction model for wind power prediction and the power loss prediction model considering wind turbine icing scenarios are trained separately. Specifically: Constructing a dataset D1 containing conventional meteorological data and corresponding wind power data, and a dataset D2 and a dataset D3 containing meteorological data and corresponding wind power data under icing scenarios respectively; Use the data set D1 to train the basic prediction model to obtain a trained basic prediction model; The wind power of the data set D2 is predicted using the trained basic prediction model, and the difference between the predicted wind power value and the actual value is recorded as power loss; the power loss and the corresponding wind speed and icing data constitute the power loss data set D4; The power loss prediction model is trained using the data set D4 to obtain a trained power loss prediction model; The trained basic prediction model and power loss prediction model are tested using the dataset D3.

5. The short-term wind power prediction method for wind turbine icing scenarios according to claim 1, characterized in that: The basic prediction model uses the LightGBM model, and the power loss prediction model uses the Few-shot model.

6. A short-term wind power prediction system for wind turbine icing scenarios, characterized in that: include: The icing model simulation and solution module is used to consider the melting and shedding effects of ice, build an icing model suitable for wind turbine blades, and solve the key parameters of the icing model under set meteorological conditions through icing simulation; Combined with numerical weather forecast (NWP) data =15min resolution, It is defined as a unit time. It is assumed that the values ​​of meteorological factors and model parameters remain unchanged within a unit time. In the area of ​​, the increased ice mass per unit time is: The ice melting and shedding effects are considered to construct an ice covering model suitable for wind turbine blades, specifically: ； ； ； in, To consider the ice melting and shedding effect, the accumulated ice coverage of a single wind turbine is When the effects of ice melting and shedding are not considered i The increase in the maximum ice thickness at the moment, To quantify the proportionality coefficient of ice reduction between ice ablation and shedding effect, After considering the effects of ice melting and shedding i The increase in the maximum ice thickness at the moment, express i The maximum ice coverage length of the airfoil at that moment; c represents the blade chord length, and n represents the set time; , are the collision coefficient and the freezing coefficient respectively, represents the capture rate, is the mass fraction of water droplets in the air, is the velocity of the water drop, is the unit time, is the density of ice accumulation, To solve for the proportionality factor when solving for the volume of ice accumulation; The key parameters of the icing model under set meteorological conditions are solved through icing simulation, specifically: Construct a physical model of the fan blades, use Fluent software to calculate the air flow field around the fan blades, import the flow field calculation results into the icing simulation software, preset the median diameter of the water droplets and the liquid water content in the air, and calculate the collision coefficient through the water droplet motion trajectory simulation and icing simulation. , freezing coefficient , solve for the proportionality coefficient when solving for the ice accumulation volume and the proportionality factor of ice reduction ; An ice cover data interpolation module is used to convert discrete ice cover data into continuous ice cover data corresponding to meteorological conditions at different times through data interpolation; A model training module is used to train the basic prediction model for wind power prediction and the power loss prediction model considering wind turbine icing scenarios; The power prediction module is used to obtain meteorological data and icing data under the wind turbine icing scenario within a set time range in the test area. The meteorological data is input into a trained basic prediction model to obtain a basic wind power prediction result. The icing data and wind speed data are input into a trained power loss prediction model to obtain a power loss prediction result. Based on the basic wind power prediction result and the power loss prediction result, the final short-term wind power prediction result for the wind turbine icing scenario is obtained.

7. A terminal device, comprising a processor and a memory, wherein the processor is used to implement instructions; and the memory is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the short-term wind power prediction method for wind turbine icing scenarios as described in any one of claims 1-5.

8. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the short-term wind power prediction method for wind turbine icing scenarios as described in any one of claims 1-5.

Images