Multi-model weight optimization fusion wind power prediction method under different wind speed scenes

Abstract

The invention discloses a multi-model weight optimization fusion wind power prediction method under different wind speed scenes, and belongs to the technical field of wind power prediction, and the method comprises the steps: obtaining historical wind power data and historical meteorological data; performing wind speed scene recognition on the historical meteorological data, then performing clustering processing to obtain historical meteorological data under different wind speed scenes, and performing data set construction; constructing a multi-algorithm hybrid model based on three prediction models, namely a random forest prediction model, a bidirectional long-short-term memory neural network prediction model and a support vector regression prediction model; constructing a weight optimization function by using the target function and the constraint condition, and solving an optimal weight based on a KKT condition to obtain a trained multi-algorithm hybrid model under different wind speed scenes; and after wind speed scene recognition is carried out on the predicted meteorological data, a multi-algorithm hybrid model in a corresponding wind speed scene is input, and a wind power prediction result can be output. Therefore, wind power prediction can be realized.

Description

Technical Field

[0001] The present invention relates to a wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios, belonging to the technical field of wind power prediction. Background Art

[0002] As wind power technology matures and improves, the installed capacity and scale of wind turbines continue to increase. This rapid growth in wind power generation places higher demands on power forecasting accuracy. Wind speed, as a core driver, exhibits strong nonlinear and multimodal characteristics due to its dynamic characteristics. Traditional single prediction models struggle to adapt to the complex variations across the full wind speed range. Therefore, classifying wind speed scenarios into different scenarios and implementing dynamic fusion of multiple models has become a key path to improving forecasting accuracy.

[0003] Commonly used wind power prediction methods include but are not limited to the Random Forest (RF) algorithm, the Naive Bayes algorithm, the Support Vector Machine (SVR) algorithm, the Convolutional Neural Network (CNN) algorithm, the Transformer algorithm, and the Bidirectional Long Short-Term Memory (BiLSTM) algorithm.

[0004] The random forest algorithm is a supervised machine learning method based on ensemble learning. It builds multiple decision trees and aggregates the results to improve prediction accuracy and robustness. The support vector regression algorithm finds the optimal hyperplane to fit the relationship between input features and output power. It can effectively handle nonlinear relationships and has strong generalization capabilities for small samples.

[0005] The support vector machine algorithm implicitly implements high-dimensional spatial mapping through kernel functions to handle nonlinear relationships. Common kernel functions include linear kernels, polynomial kernels, and Gaussian kernels (RBFs). In wind power prediction, the Gaussian kernel is widely used because it can effectively model the complex nonlinear coupling between meteorological factors such as wind speed and temperature at different altitudes and power. Parameter adjustment is a key step in SVR: increasing the penalty coefficient can reduce training error but may overfit, while reducing the insensitive loss parameter improves regression accuracy but increases computational complexity. The value that determines the sample influence range requires a balance between model complexity and generalization ability. The advantage of SVR lies in its robustness to small sample data and high-dimensional features, and its use of kernel techniques to avoid explicit high-dimensional calculations. It is also insensitive to outliers and is suitable for scenarios where power fluctuations in wind speed data are caused by gusts or noise.

[0006] The long short-term memory (BiLSTM) neural network models multi-scale temporal and spatial correlations through a bidirectional gating mechanism. In unstable wind power output conditions, such as strong turbulent activity, BiLSTM has powerful time series modeling capabilities and can simultaneously consider past and future information, thereby better capturing the time dependence and dynamic characteristics of wind power. This makes it particularly suitable for predicting short-term fluctuations in scenarios with sudden weather changes.

[0007] While these algorithms have achieved some success in wind power forecasting, single algorithms often face challenges in handling the nonlinear and unsteady characteristics of data, potentially leading to insufficient generalization. One of the challenges facing wind power forecasting is the nonlinear and unsteady nature of wind speed, with significant differences in wind power output characteristics under different wind speed scenarios. Each model has its own advantages, but also limitations, making it difficult for a single algorithm to simultaneously address these complex situations. Summary of the Invention

[0008] The purpose of the present invention is to provide a wind power prediction method with multi-model weight optimization fusion under different wind speed scenarios. By integrating the prediction results of the random forest prediction model, the support vector regression prediction model and the bidirectional long short-term memory neural network prediction model, the accuracy and robustness of wind power prediction can be improved, providing strong support for the operation management and scheduling decision-making of wind farms.

[0009] In order to achieve the above technical objectives, the present invention will adopt the following technical solutions:

[0010] A wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios includes the following steps:

[0011] Obtain historical wind power data and historical meteorological data within a specified historical period; historical meteorological data includes multi-dimensional wind field observation data;

[0012] After performing wind speed scene recognition on historical meteorological data, clustering processing is performed to obtain historical meteorological data under different wind speed scenes, and the historical meteorological data under different wind speed scenes are combined with the historical wind power matched with their time and space to form a data set under the corresponding wind speed scene. The constructed data set under different wind speed scenes is then divided into a training set, a validation set, and a test set according to a preset ratio; in the data set, the feature data is the historical meteorological data, and the label data is the historical wind power;

[0013] A multi-algorithm hybrid model is constructed based on three heterogeneous prediction models: the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model. A weight optimization function is constructed for the three heterogeneous prediction models of the multi-algorithm hybrid model using the objective function and constraints. The optimal weight is solved based on the KKT condition to achieve the optimized fusion of the model weights between the prediction models under different wind speed scenarios, thereby obtaining a trained multi-algorithm hybrid model under different wind speed scenarios.

[0014] Based on the trained multi-algorithm hybrid model under different wind speed scenarios, the predicted meteorological data is subjected to wind speed scenario recognition and then input into the multi-algorithm hybrid model under the corresponding wind speed scenario to output the wind power prediction result.

[0015] Preferably, the wind speed scene recognition is performed on the historical meteorological data / forecast meteorological data, specifically including the following steps:

[0016] Based on the multi-dimensional wind field observation data of historical meteorological data / forecast meteorological data at any time, the horizontal wind energy capture efficiency at the corresponding time is calculated respectively. , vertical disturbance loss factor and wind direction yaw efficiency ;

[0017] Based on the calculated horizontal wind energy capture efficiency at any time , vertical disturbance loss factor and wind direction yaw efficiency , calculate the three-dimensional wind energy coupling efficiency index at the corresponding moment;

[0018] By comparing the calculated three-dimensional wind energy coupling efficiency index TWECI at any time with the preset three-dimensional wind energy coupling efficiency index thresholds under different wind speed scenarios, the wind speed scenario at the current moment can be identified.

[0019] Preferably, the different wind speed scenarios include: an ideal capture scenario, a stable wind scenario, a turbulent mixing scenario, and a wind energy inefficiency scenario;

[0020] The preset threshold of the 3D wind energy coupling efficiency index for the ideal capture scenario is (0.8~1.0]; the preset threshold of the 3D wind energy coupling efficiency index for the stable wind scenario is (0.5~0.8]; the preset threshold of the 3D wind energy coupling efficiency index for the turbulent mixing scenario is (0.3~0.5]; the preset threshold of the 3D wind energy coupling efficiency index for the low wind energy efficiency scenario is [0~0.3];

[0021] When identifying the wind speed scenario at any moment, if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.8~1.0], it indicates that the wind speed scenario at the current moment is an ideal capture scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.5~0.8], it indicates that the wind speed scenario at the current moment is a stable wind condition scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.3~0.5], it indicates that the wind speed scenario at the current moment is a turbulent mixing scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between [0~0.3], it indicates that the wind speed scenario at the current moment is an inefficient wind energy scenario.

[0022] Preferably, the horizontal wind energy capture efficiency at any time is , vertical disturbance loss factor and wind direction yaw efficiency Calculated by the following formula:

[0023] = ;

[0024] ;

[0025] = ;

[0026] In the above formula: is the horizontal wind speed; is the rated wind speed of the wind turbine; is the wind energy utilization coefficient; Indicates vertical wind speed; is the current yaw angle of the wind turbine nacelle; Indicates wind direction.

[0027] Preferably, the three-dimensional wind energy coupling efficiency index TWECI is calculated by the following formula:

[0028] ;

[0029] Where: represents the horizontal wind energy capture efficiency; represents the vertical disturbance loss factor; Indicates wind direction yaw efficiency.

[0030] Preferably, the multi-algorithm hybrid model is represented by the following formula:

[0031] ;

[0032] In the above formula, Represents the predicted value of the multi-algorithm hybrid model at time t; represents the predicted value of the k-th prediction model at time t, where k=1, 2, and 3 correspond to the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, respectively; Represents the model weight of the k-th prediction model in the multi-algorithm hybrid model.

[0033] Preferably, the multi-algorithm hybrid model is constructed by the following steps:

[0034] The constructed datasets are used to train, validate, and test each prediction model to optimize the hyperparameters of the corresponding prediction models;

[0035] Construct a weight optimization mathematical model and solve it based on the KKT condition to obtain the model weights of each prediction model in the multi-algorithm hybrid model, and use a sliding window mechanism to dynamically adjust the model weights;

[0036] The objective function of the weight optimization mathematical model is expressed as:

[0037] ;

[0038] Where: N represents the total number of time t, Indicates the actual power value at time t.

[0039] Preferably, when optimizing the hyperparameters of each prediction model, a single prediction parameter optimization model is defined as follows:

[0040] ;

[0041] ;

[0042] Where: is the prediction function of the k-th prediction model, ; Its hyperparameters; is its state variable at time t;

[0043] For the random forest prediction model, its hyperparameters is the number of trees K and the maximum depth; for the bidirectional long short-term memory neural network prediction model, its hyperparameters is the hidden layer dimension and learning rate; for the support vector regression prediction model, its hyperparameters is the penalty coefficient C and the Gaussian kernel function width σ.

[0044] Preferably, the KKT conditions specifically include:

[0045] Lagrangian function:

[0046] ;

[0047] Stability:

[0048] ;

[0049] Original feasibility:

[0050] Equality constraints: ; Inequality constraints: ;

[0051] Dual feasibility: ;

[0052] Complementary slackness: ;

[0053] in, represents the predicted value of the k-th prediction model at time t, where k=1, 2, and 3 correspond to the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, respectively; Represents the model weight of the k-th prediction model obtained by solving the KKT condition in the multi-algorithm hybrid model, represents the actual power value at time t, is the Lagrange multiplier corresponding to the equality constraint, is the KKT multiplier corresponding to the inequality constraint; j represents the traversal index used for internal summation, which, like k, represents each prediction model, j=1, 2, 3.

[0054] Another technical object of the present invention is to provide an electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, which, when running, executes the above-mentioned wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios.

[0055] Based on the above technical objectives, the present invention has the following advantages over the prior art:

[0056] The wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios described in the present invention first performs wind speed scenario recognition on the acquired historical meteorological data to construct a data set under different wind speed scenarios; then constructs a multi-algorithm hybrid model based on three heterogeneous prediction models: random forest prediction model, bidirectional long short-term memory neural network prediction model, and support vector regression prediction model, and uses the objective function and constraint conditions to construct a weight optimization function. The optimal weight is solved based on the KKT condition to achieve dynamic optimization and fusion of model weights between multiple prediction models under different wind speed scenarios, call on demand, complement each other's advantages, and achieve a double leap in overall accuracy and robustness. It was found in a certain wind farm application that the root mean square error of the multi-algorithm hybrid model under different wind speed scenarios is reduced by up to 22% compared with the single model prediction. It can be seen that the multi-algorithm hybrid model constructed by the present invention fully exploits the nonlinearity, temporal sequence and high-dimensional coupling characteristics in the wind power sequence, achieves complementary advantages under different wind speed scenarios, improves the wind power prediction accuracy, and provides strong support for the operation management and scheduling decision-making of the wind farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of the wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios described in the present invention. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The following description of at least one exemplary embodiment is actually only illustrative and in no way serves as any limitation on the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of the present invention. Unless otherwise specified, the relative arrangement of components and steps, expressions and numerical values ​​described in these embodiments do not limit the scope of the present invention. Technologies, methods and equipment known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the technologies, methods and equipment should be considered part of the specification. In all examples shown and discussed here, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of the exemplary embodiments may have different values.

[0059] Example 1

[0060] like Figure 1 As shown, the wind power prediction method of the present invention using multi-model weight optimization fusion under different wind speed scenarios specifically includes the following steps:

[0061] Step 1: Build a dataset for different wind speed scenarios:

[0062] The present invention takes into account the impact of nonlinearity and non-steadiness of wind speed on wind power prediction. Therefore, through cluster analysis, historical meteorological data are divided into different wind speed scenarios, thereby constructing data sets under different wind speed scenarios to train, test and verify the subsequently established multi-algorithm hybrid model.

[0063] The datasets for different wind speed scenarios are constructed through the following steps:

[0064] Step 1.1: Obtain historical wind power data and historical meteorological data:

[0065] Obtain historical wind power and historical meteorological data for a specified historical period. Historical meteorological data includes multi-dimensional wind field observation data and historical temperature. Multi-dimensional wind field observation data includes historical horizontal wind speed, historical vertical wind speed, and historical wind direction.

[0066] Historical meteorological data and historical wind power are sampled and recorded at set time intervals (e.g., every 15 minutes) by the Supervisory Control and Data Acquisition (SCADA) system of the actual wind farm, including wind speed and direction records detected by the wind tower.

[0067] Step 1.2: Identify wind speed scenarios based on historical meteorological data:

[0068] Based on multi-dimensional wind farm observation data and targeting wind power forecasting needs, this paper combines three-dimensional wind farm characteristics with wind turbine power and defines a three-dimensional Wind Energy Coupling Effectiveness Index (TWECI) to identify different wind speed scenarios based on historical meteorological data. The method specifically includes the following steps:

[0069] Step 1.2.1: Based on the multi-dimensional wind field observation data of the historical meteorological data at any time, calculate the horizontal wind energy capture efficiency at the corresponding time. , vertical disturbance loss factor and wind direction yaw efficiency .

[0070] Horizontal wind energy capture efficiency Calculated by the following formula:

[0071] = ;

[0072] In the above formula, is the rated wind speed of the wind turbine; is the horizontal wind speed; is the wind energy utilization coefficient, which can be found by querying The curve is obtained, represents the tip speed ratio, which is calculated by the following formula:

[0073] ;

[0074] In the above formula, is the horizontal wind speed; is the blade rotation speed; R is the blade rotation radius.

[0075] According to blade momentum theory, when the vertical wind speed exceeds 20% of the horizontal wind speed, the three-dimensional flow separation effect begins to dominate, and the vertical wind speed increases the loss of wind energy. Therefore, the present invention uses the following formula to calculate the vertical disturbance loss factor :

[0076] ;

[0077] In the above formula: Indicates vertical wind speed; is the horizontal wind speed.

[0078] Wind direction yaw efficiency Calculated by the following formula:

[0079] = ;

[0080] In the above formula: is the current yaw angle of the wind turbine nacelle from the SCADA encoder; Indicates wind direction.

[0081] Step 1.2.2: Based on the calculated horizontal wind energy capture efficiency at any time , vertical disturbance loss factor and wind direction yaw efficiency , calculate the three-dimensional wind energy coupling efficiency index at the corresponding moment. Among them: the three-dimensional wind energy coupling efficiency index TWECI is calculated by the following formula:

[0082] ;

[0083] Where: represents the horizontal wind energy capture efficiency; represents the vertical disturbance loss factor; Indicates wind direction yaw efficiency.

[0084] Step 1.2.3: Based on the calculated three-dimensional wind energy coupling efficiency index, the three-dimensional wind energy coupling efficiency index is compared with the preset three-dimensional wind energy coupling efficiency index thresholds under different wind speed scenarios to identify the wind speed scenario at the current moment.

[0085] Generally, the closer the three-dimensional wind energy coupling efficiency index is to 1, the better the wind energy coupling efficiency is, and the closer it is to the ideal power generation scenario. The closer it is to 0, the more unfavorable the power generation scenario is. Therefore, this step defines four wind speed scenarios, namely: ideal capture scenario, stable wind condition scenario, turbulent mixing scenario, and wind energy inefficiency scenario. The present invention also presets three-dimensional wind energy coupling efficiency index thresholds for each wind speed scenario, respectively: the three-dimensional wind energy coupling efficiency index threshold for the ideal capture scenario is (0.8~1.0]; the three-dimensional wind energy coupling efficiency index threshold for the stable wind condition scenario is (0.5~0.8]; the three-dimensional wind energy coupling efficiency index threshold for the turbulent mixing scenario is (0.3~0.5]; the three-dimensional wind energy coupling efficiency index threshold for the wind energy inefficiency scenario is [0~0.3].

[0086] Therefore, when identifying the wind speed scenario at any moment, if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.8~1.0], it indicates that the wind speed scenario at the current moment is an ideal capture scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.5~0.8], it indicates that the wind speed scenario at the current moment is a stable wind condition scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.3~0.5], it indicates that the wind speed scenario at the current moment is a turbulent mixing scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between [0~0.3], it indicates that the wind speed scenario at the current moment is an inefficient wind energy scenario.

[0087] Step 1.3: Based on the identified different wind speed scenarios, construct the corresponding wind speed scenario dataset. Specifically:

[0088] Clustering of historical meteorological data is performed based on the obtained wind speed scenarios to obtain historical meteorological data under different wind speed scenarios.

[0089] The historical meteorological data under different wind speed scenarios and their spatiotemporally matched historical wind power are combined to form a data set under the corresponding wind speed scenario; in the data set, the feature data is the historical meteorological data (including historical wind speed, etc.), and the label data is the historical wind power.

[0090] Step 2: Build a multi-algorithm hybrid model:

[0091] In order to give full play to the heterogeneity advantages of the three heterogeneous models, namely the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, for wind power prediction results, the present invention constructs a multi-algorithm hybrid model:

[0092] ;

[0093] In the above formula, Represents the predicted value of the multi-algorithm hybrid model at time t; represents the predicted value of the k-th prediction model at time t; Represents the model weight of the k-th prediction model in the multi-algorithm hybrid model; the value of k is k=1, 2, 3, corresponding to the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, respectively.

[0094] In this invention, based on the joint parameter adaptive optimization strategy, the objective function and constraints are used to construct a weight optimization function, and the optimal weight is solved based on the KKT condition to obtain a trained multi-algorithm hybrid model under different wind speed scenarios, realize the dynamic weight optimization fusion of multiple models under different wind speed scenarios, fully explore the nonlinearity, temporal sequence and high-dimensional coupling characteristics in the wind power sequence, realize multi-algorithm collaborative prediction under complex wind speed scenarios, improve the wind power prediction accuracy, and provide strong support for the operation management and scheduling decision-making of wind farms. Specifically, it includes the following steps:

[0095] Step 2.1: Hyperparameter optimization of single prediction model:

[0096] The datasets under different wind speed scenarios constructed in step 1 are divided into training set, validation set and test set in proportion.

[0097] The constructed training set is used to train each prediction model separately to optimize the hyperparameters of the corresponding prediction model, which can be expressed as:

[0098] ;

[0099] ;

[0100] Where, is the prediction function of the k-th prediction model, Its hyperparameters, is its state variable. For the random forest prediction model, its hyperparameter is the number of trees K and the maximum depth; for the bidirectional long short-term memory neural network prediction model, its hyperparameters is the hidden layer dimension and learning rate; for the support vector regression prediction model, its hyperparameters is the penalty coefficient C and the width of the Gaussian kernel function σ.

[0101] Step 2.2. Construction and solution of weight optimization mathematical model based on KKT conditions:

[0102] This step aims to dynamically assign optimal weights to the multi-algorithm hybrid model for different wind speed scenarios. Based on the three individual prediction models optimized in Step 2.1: the random forest prediction model (RF), the bidirectional long short-term memory neural network prediction model (BiLSTM), and the support vector regression prediction model (SVR), a weight optimization model is constructed. Its core is to solve a constrained quadratic programming problem.

[0103] First, define the weight optimization mathematical model. The goal is to minimize the difference between the weighted combination prediction value and the actual power value on a data set containing N time points. The mean square error (MSE) between them. Its objective function and constraints are as follows:

[0104] Objective function:

[0105] ;

[0106] in, is the variable to be optimized, which represents the weight of the k-th prediction model; is a known value, which represents the predicted power value of the k-th prediction model at time t; is a known value representing the predicted power of the kth prediction model at time t; k is the model index, k=1, 2, 3, corresponding to the random forest, BiLSTM, and SVR models, respectively; t is the index of the time point, t=1, 2, ..., N, where N is the total number of time points used for model training or weight optimization, which is determined by the size of the sliding window.

[0107] Constraints:

[0108] (equality constraints);

[0109] (inequality constraints);

[0110] in, is the predicted value of the kth model at time t, is the weight of the kth model to be solved. The equality constraint that the weights sum to 1 ensures the unbiased nature of the combination; the inequality constraint that the weights are non-negative ensures that each model contributes positively or zero to the final result.

[0111] Because this optimization problem involves both equality and inequality constraints, the traditional Lagrange multiplier method is no longer applicable. Therefore, the present invention employs the more universal Karush-Kuhn-Tucker (KKT) conditions to solve the problem. The KKT conditions provide a comprehensive theoretical framework for solving such constrained optimization problems.

[0112] To this end, we introduce the Lagrange multiplier Corresponding to the equality constraint, and introducing the KKT multiplier (k=1, 2, 3) corresponds to three inequality constraints, constructing the generalized Lagrangian function:

[0113] ;

[0114] in, is the constructed generalized Lagrangian function; is the Lagrange multiplier, is the dual variable associated with the equality constraint; is the KKT multiplier, which is related to the kth inequality constraint ≥0 associated dual variables.

[0115] According to KKT theory, the solution constraints of this optimization problem are as follows:

[0116] (a). L pair The gradient of is zero to maintain stationarity requirement:

[0117] (k=1, 2, 3);

[0118] Here, j is a traversal index used for the inner summation, and like k, it represents each prediction model (j=1, 2, 3). The traversal index j is used to distinguish it from the outer derivative variable k.

[0119] (b) The solution must satisfy all original constraints;

[0120] (c) The KKT multiplier corresponding to the inequality constraint must be non-negative;

[0121] (d) The product of the KKT multiplier and its corresponding inequality constraint is zero to maintain complementary slackness:

[0122] (k=1, 2, 3) ;

[0123] Complementary slackness is the core of the KKT condition, which reveals the structure of the optimal solution: if a certain weight In the optimal solution, it is strictly greater than zero, so the corresponding inequality constraint ≥ 0 is inactive, in which case the KKT multiplier Must be zero. Conversely, if a KKT multiplier >0, then the corresponding constraint must be tight (active), that is, =0.

[0124] The solution is as follows:

[0125] Based on the above KKT conditions, we design a solution algorithm. First, we assume that all inequality constraints are non-tight, that is, =0 (k=1, 2, 3), which means that we only consider the equality constraints for now. At this point, the stationarity condition is simplified to:

[0126] ;

[0127] This system of equations can be written in a more concise matrix form. Let , and define matrix A and vector B as follows:

[0128] ;

[0129] ;

[0130] in It is the element in the kth row and jth column of matrix A. It represents the inner product of the prediction sequence of model k and the prediction sequence of model j, reflecting the correlation between the prediction results of the two models. The kth element of vector B represents the inner product of the actual power series and the predicted series of model k, reflecting the accuracy of the prediction of model k.

[0131] Then the stationarity condition and the equality constraint in the original feasibility The following linear equations are formed:

[0132] ;

[0133] In the above formula, for the matrix , yes The 3x3 matrix formed; The weight to be sought , , The 3x1 column vector formed; is a 3x1 column vector with all elements set to 1, is a column vector The transpose of , that is, a 1x3 row vector [1, 1, 1]; for the matrix , B is composed of elements 3x1 column vector.

[0134] By solving the linear equations, a set of candidate weights can be obtained Then, check whether any of the solutions satisfies ≥0. If any solution All satisfied ≥0, the candidate weight is considered All the solutions included in the solution satisfy all KKT conditions at the same time, which is the optimal solution. <0, then the original feasibility is violated, indicating that the initial = 0 is not true. According to the complementary slackness, the weight The corresponding constraints must be tight, i.e. = 0. At this point, we set the weight The remaining weights are then restructured and a smaller set of linear equations is solved. This process is repeated until all the weights satisfy the non-negative condition.

[0135] Through the rigorous solution process based on the KKT conditions mentioned above, we can ensure that the optimal model weight combination that minimizes the prediction error under the current wind speed scenario is found. .

[0136] Step 2.3. Dynamic weight allocation mechanism:

[0137] To accommodate the non-stationary nature of wind power time series, a sliding window mechanism is used to dynamically adjust weights. The window length is set to 24 hours (96 15-minute sampling points), balancing computational efficiency and timeliness. The weights are updated every hour to ensure they match the current wind speed scenario.

[0138] Step 3: Output the multi-algorithm collaborative wind power prediction results under different wind speed scenarios.

[0139] Based on the multi-algorithm hybrid model for different wind speed scenarios constructed in step 2, the predicted meteorological data is subjected to corresponding wind speed scenario identification and then input into the multi-algorithm hybrid model for the corresponding wind speed scenario to output the wind power prediction result. It should be mentioned here that the specific method for identifying the corresponding wind speed scenario using the predicted meteorological data can refer to the specific method for identifying the corresponding wind speed scenario using the historical meteorological data described above, and will not be repeated here.

[0140] Example 2

[0141] The present invention also provides a storage medium, which stores a computer program that, when running, executes the wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios.

[0142] Example 3

[0143] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor runs the computer program, it executes the wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios.

[0144] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0145] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0146] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0147] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0148] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0149] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), a mobile hard drive, a magnetic disk, or an optical disk.

[0150] Application Examples

[0151] Taking different wind speed scenarios of a wind farm as an example, the weight distribution is shown in Table 1:

[0152] Table 1 Weight distribution under different wind speed scenarios

[0153]

[0154] In Table 1, ω1 represents the weight of the random forest prediction model in the multi-algorithm hybrid model; ω2 represents the weight of the bidirectional long short-term memory neural network prediction model in the multi-algorithm hybrid model; ω3 represents the weight of the support vector regression prediction model in the multi-algorithm hybrid model.

[0155] The random forest prediction model has the highest weight in the ideal capture scenario, while BiLSTM and SVR dominate in the turbulent mixing scenario and the wind energy inefficiency scenario, and can better capture the nonlinear process in the changing process.

[0156] Verified using data from a 52.8MW wind farm, the random forest prediction model achieved a single-model root mean square error (RMSE) of 5.21 kWh and a mean absolute error (MAE) of 1.64 kWh. This outperformed the SVR (RMSE 5.75 kWh, MAE 1.75), but was weaker than the BiLSTM (RMSE 4.93 kWh, MAE 1.54 kWh). However, after multi-model fusion (a multi-algorithm hybrid model), the RMSE dropped to 4.5 kWh and the MAE to 1.42 kWh, significantly improving prediction accuracy. This demonstrates the critical role of model fusion in predictive results and the effectiveness of dynamic weight optimization.

[0157] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios, characterized by: The steps include: Obtain historical wind power data and historical meteorological data within a specified historical period; historical meteorological data includes multi-dimensional wind field observation data; After performing wind speed scene recognition on historical meteorological data, clustering processing is performed to obtain historical meteorological data under different wind speed scenes, and the historical meteorological data under different wind speed scenes are combined with the historical wind power matched with their time and space to form a data set under the corresponding wind speed scene. The constructed data set under different wind speed scenes is then divided into a training set, a validation set, and a test set according to a preset ratio; in the data set, the feature data is the historical meteorological data, and the label data is the historical wind power; A multi-algorithm hybrid model is constructed based on three heterogeneous prediction models: the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model. A weight optimization function is constructed for the three heterogeneous prediction models of the multi-algorithm hybrid model using the objective function and constraints. The optimal weight is solved based on the KKT condition to achieve the optimized fusion of the model weights between the prediction models under different wind speed scenarios, thereby obtaining a trained multi-algorithm hybrid model under different wind speed scenarios. Based on the trained multi-algorithm hybrid model under different wind speed scenarios, the predicted meteorological data is subjected to wind speed scenario recognition and then input into the multi-algorithm hybrid model under the corresponding wind speed scenario to output the wind power prediction result.

2. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 1 is characterized in that: Using historical meteorological data / forecast meteorological data to identify wind speed scenarios includes the following steps: Based on the multi-dimensional wind field observation data of historical meteorological data / forecast meteorological data at any time, the horizontal wind energy capture efficiency at the corresponding time is calculated respectively. , vertical disturbance loss factor and wind direction yaw efficiency ; Based on the calculated horizontal wind energy capture efficiency at any time , vertical disturbance loss factor and wind direction yaw efficiency , calculate the three-dimensional wind energy coupling efficiency index at the corresponding moment; By comparing the calculated three-dimensional wind energy coupling efficiency index TWECI at any time with the preset three-dimensional wind energy coupling efficiency index thresholds under different wind speed scenarios, the wind speed scenario at the current moment can be identified.

3. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 2 is characterized in that: Different wind speed scenarios include: ideal capture scenario, stable wind scenario, turbulent mixing scenario, and wind energy inefficiency scenario; The preset threshold of the 3D wind energy coupling efficiency index for the ideal capture scenario is (0.8~1.0]; the preset threshold of the 3D wind energy coupling efficiency index for the stable wind scenario is (0.5~0.8]; the preset threshold of the 3D wind energy coupling efficiency index for the turbulent mixing scenario is (0.3~0.5]; the preset threshold of the 3D wind energy coupling efficiency index for the low wind energy efficiency scenario is [0~0.3]; When identifying the wind speed scenario at any moment, if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.8~1.0], it indicates that the wind speed scenario at the current moment is an ideal capture scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.5~0.8], it indicates that the wind speed scenario at the current moment is a stable wind condition scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between (0.3~0.5], it indicates that the wind speed scenario at the current moment is a turbulent mixing scenario; if the three-dimensional wind energy coupling efficiency index TWECI at the current moment is between [0~0.3], it indicates that the wind speed scenario at the current moment is an inefficient wind energy scenario.

4. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 2 is characterized in that: Horizontal wind energy capture efficiency at any time , vertical disturbance loss factor and wind direction yaw efficiency Calculated by the following formula: = ； ； = ； In the above formula: is the horizontal wind speed; is the rated wind speed of the wind turbine; is the wind energy utilization coefficient; Indicates vertical wind speed; is the current yaw angle of the wind turbine nacelle; Indicates wind direction.

5. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 4 is characterized in that: The three-dimensional wind energy coupling efficiency index TWECI is calculated by the following formula: ； Where: represents the horizontal wind energy capture efficiency; represents the vertical disturbance loss factor; Indicates wind direction yaw efficiency.

6. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 1 is characterized in that: The multi-algorithm hybrid model is expressed as follows: ； In the above formula, Represents the predicted value of the multi-algorithm hybrid model at time t; represents the predicted value of the k-th prediction model at time t, where k=1, 2, and 3 correspond to the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, respectively; Represents the model weight of the k-th prediction model in the multi-algorithm hybrid model.

7. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 6 is characterized in that: The multi-algorithm hybrid model is constructed through the following steps: The constructed datasets are used to train, validate, and test each prediction model to optimize the hyperparameters of the corresponding prediction models; Construct a weight optimization mathematical model and solve it based on the KKT condition to obtain the model weights of each prediction model in the multi-algorithm hybrid model, and use a sliding window mechanism to dynamically adjust the model weights; The objective function of the weight optimization mathematical model is expressed as: ； Where: N represents the total number of time t, Indicates the actual power value at time t.

8. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 7 is characterized in that: When optimizing the hyperparameters of each prediction model, the single prediction parameter optimization model is defined as follows: ； ； Where: is the prediction function of the k-th prediction model, ; Its hyperparameters; is its state variable at time t; For the random forest prediction model, its hyperparameters is the number of trees K and the maximum depth; for the bidirectional long short-term memory neural network prediction model, its hyperparameters is the hidden layer dimension and learning rate; for the support vector regression prediction model, its hyperparameters is the penalty coefficient C and the Gaussian kernel function width σ.

9. The wind power prediction method based on multi-model weight optimization fusion under different wind speed scenarios according to claim 7 is characterized in that: KKT conditions specifically include: Lagrangian function: ； Stability: ； Original feasibility: Equality constraints: ; Inequality constraints: ; Dual feasibility: ; Complementary slackness: ; in, represents the predicted value of the k-th prediction model at time t, where k=1, 2, and 3 correspond to the random forest prediction model, the bidirectional long short-term memory neural network prediction model, and the support vector regression prediction model, respectively; Represents the model weight of the k-th prediction model obtained by solving the KKT condition in the multi-algorithm hybrid model, represents the actual power value at time t, is the Lagrange multiplier corresponding to the equality constraint, is the KKT multiplier corresponding to the inequality constraint; j represents the traversal index used for internal summation, which, like k, represents each prediction model, j=1, 2, 3.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the computer program is running, the wind power prediction method of multi-model weight optimization fusion under different wind speed scenarios described in any one of claims 1 to 9 is executed.

Images