Wind power plant power generation prediction and uncertainty analysis method and system based on hybrid intelligent algorithm

Abstract

The invention discloses a wind power plant power generation prediction and uncertainty analysis method and system based on a hybrid intelligent algorithm, and belongs to the technical field of wind power plant power generation prediction, and the method comprises the steps: employing a fuzzy K-means clustering algorithm to carry out the classification of wind speed data, and dividing a data set according to the wind speed grade; a plurality of support vector regression models are established, each model corresponds to different wind speed grades, and parameters of each support vector regression model are optimized by using an enhanced harmony search algorithm; and carrying out uncertainty analysis on a prediction result of the support vector regression model by adopting a quantile regression method, calculating prediction intervals under different confidence levels, and optimizing parameters of the quantile regression model by utilizing an enhanced harmony search algorithm. The method improves the organization and correlation of the data, enables the prediction model to be trained and optimized for different wind speed grades, effectively avoids the problems of local optimum and overfitting, improves the precision and generalization capability of wind power prediction, and achieves the efficient and precise prediction of the wind power.

Description

Technical Field

[0001] The present invention relates to the technical field of wind farm power generation prediction, and particularly relates to a method and system for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm. Background Art

[0002] The existing technologies have various deficiencies in wind farm power generation prediction and uncertainty analysis. First of all, traditional prediction methods such as the ARIMA model and support vector machines are difficult to fully capture the non-linearity, non-stationarity, and strong randomness of wind power data, resulting in limited prediction accuracy. When the ARIMA model deals with large fluctuations or sudden changes in wind power data, the prediction error increases significantly. Secondly, some methods focus on average power prediction and do not conduct in-depth uncertainty analysis, and cannot accurately evaluate the confidence interval and risk level of the prediction results. Some prediction methods based on physical models can establish power generation models according to the geographical environment of the wind farm, the characteristics of wind turbines, etc., but they do not adequately consider the uncertainty of meteorological conditions and random factors such as equipment failures, and cannot give early risk warnings.

[0003] In addition, the adaptability and generalization ability of existing models need to be improved. Differences in the geographical environment, wind turbine types, etc. of different wind farms cause the performance of the model to decline in different scenarios. For example, a prediction model that performs well in a coastal wind farm will have a significantly reduced prediction accuracy when directly applied to an inland wind farm due to differences in meteorological conditions and terrain. Insufficient data preprocessing and feature extraction capabilities are also shortcomings of the existing technologies. The original data of wind farms usually contains a large amount of noise, outliers, and missing values, and the data dimension is high. Existing methods are not perfect in data preprocessing and feature extraction, and cannot effectively remove noise, fill in missing values, and extract key features, resulting in low data quality input into the model and affecting the accuracy of prediction results.

[0004] Finally, although some advanced deep learning models have good prediction effects, they have high computational complexity and are difficult to meet real-time requirements. In power grid dispatching, rapid dispatching decisions need to be made based on the real-time power generation prediction data of wind farms, and some complex models may take a long time to complete a prediction calculation and cannot meet the needs of real-time power grid dispatching.

[0005] In summary, there are many deficiencies in the existing technologies in wind farm power generation prediction and uncertainty analysis, and they cannot meet the requirements of high-precision, high-adaptability, and real-time prediction in practical engineering. Therefore, there is an urgent need for a new analysis method that can fully consider the complex characteristics of wind power data, effectively improve prediction accuracy, and have good adaptability and generalization ability to meet the application requirements in different wind farm scenarios. Summary of the Invention

[0006] To solve the above technical problems, a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm is proposed, including obtaining historical wind power data and wind speed data of the wind farm, arranging the data in a time series, and performing data preprocessing to remove outliers and missing data;

[0007] Classify the wind speed data using the fuzzy K-means clustering algorithm, and divide the data set according to the wind speed levels;

[0008] Establish multiple support vector regression models, each model corresponding to a different wind speed level, and use the enhanced harmony search algorithm to optimize the parameters of each support vector regression model;

[0009] Use the quantile regression method to perform uncertainty analysis on the prediction results of the support vector regression model, calculate the prediction intervals at different confidence levels, and use the enhanced harmony search algorithm to optimize the parameters of the quantile regression model.

[0010] As a preferred solution of the method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm described in the present invention, wherein: classifying the time series data set of historical wind power output into different wind speed levels includes setting initial clustering centers based on the historical wind speed data of the wind farm, and calculating the Euclidean distance between the wind speed observation values and each clustering center;

[0011] Adopt the fuzzy K-means clustering algorithm, based on the calculated Euclidean distance, determine the membership degree of each wind speed observation value to different clustering centers, and perform initial clustering according to the principle of the minimum Euclidean distance;

[0012] Use the weighted mean calculation method to update the clustering centers, and adjust the classification of the data points according to the membership degree, and repeat the iterative calculation until the preset convergence condition is met;

[0013] After meeting the convergence condition, output the final classification result of the wind speed data, and form a data set of multiple wind speed levels.

[0014] As a preferred solution of the method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm described in the present invention, wherein: establishing multiple support vector regression models includes calculating the deviation from the hyperplane to the data points based on the wind speed level data set, defining the unit normal vector, slack variables, and the weight value of the penalty function, and establishing the objective optimization function;

[0015] Initialize the kernel function parameters, penalty coefficient, and slack variables of the support vector regression model, calculate the loss value using the training data, and solve the optimization objective function to obtain the initial model;

[0016] Based on the initialized SVR prediction model, the datasets of different wind speed levels are trained separately, the loss value of each model is calculated, and the preliminary prediction results are obtained based on the adjustment of the penalty coefficient and the slack variable.

[0017] As a preferred solution of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to the present invention, wherein: optimizing the parameters of each support vector regression model using the enhanced harmony search algorithm includes, after the SVR training is completed, initializing the enhanced harmony search algorithm, randomly generating a plurality of feasible parameter combinations, including the weight of the penalty function and the inflation parameter of the kernel function, and storing the parameter vector in the harmony matrix;

[0018] Based on the random generation method, within the preset parameter range, a plurality of candidate solutions are randomly generated according to the uniform distribution, the SVR prediction error corresponding to each candidate solution is calculated, and the optimal solution is selected according to the fitness value;

[0019] During the harmony search process, the loss value of the support vector regression model is calculated based on the current parameter combination, and the penalty coefficient and the kernel function parameters of the support vector regression model are dynamically adjusted according to the method combining local search and global search;

[0020] In each iteration process, the parameter vector with the optimal fitness value is selected, the harmony matrix is updated, and the loss value of the SVR prediction model is recalculated. After the optimization termination condition is satisfied, the optimal parameter combination is determined.

[0021] As a preferred solution of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to the present invention, wherein: performing uncertainty analysis includes, based on the trained SVR prediction model, obtaining the wind power prediction value, and constructing the corresponding independent variable sample set and random sample set;

[0022] Setting the confidence level range, defining the quantile factor, and calculating the regression parameters of the historical wind power data at different quantile levels;

[0023] Calculating the regression objective optimization function, taking the prediction error as the optimization objective, calculating the quantile regression parameters based on the loss function, and minimizing the regression error;

[0024] Calculating the quantile regression function, and calculating the upper confidence limit and the lower confidence limit of the predicted value based on the regression parameters of different confidence levels.

[0025] As a preferred solution of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to the present invention, wherein: optimizing the parameters of the quantile regression model includes, after calculating the confidence interval of the quantile regression, initializing the enhanced harmony search algorithm, generating a plurality of quantile regression parameter combinations, and storing them in the harmony matrix;

[0026] Based on the random generation method, within a preset parameter range, a plurality of candidate solutions are randomly generated according to a uniform distribution, and the quantile regression error corresponding to each candidate solution is calculated;

[0027] Calculate the fitness value of the current parameter combination, and adjust the regression parameters of the quantile regression model according to the method combining local search and global search;

[0028] In each iteration process, update the harmony matrix, recalculate the quantile regression error based on the adjusted quantile regression model parameters, and determine whether the optimization termination condition is satisfied;

[0029] After the optimization termination condition is satisfied, fix the finally optimized quantile regression parameters.

[0030] As a preferred solution of a wind farm power generation prediction and uncertainty analysis method based on a hybrid intelligent algorithm according to the present invention, wherein: the prediction result is evaluated for error, the root mean square error and the mean absolute error are calculated, and the model parameters are adjusted based on the error evaluation result.

[0031] Another object of the present invention is to provide a wind farm power generation prediction and uncertainty analysis system based on a hybrid intelligent algorithm. The present invention solves the problems that existing methods such as ARIMA models and support vector machines are difficult to effectively capture the non-linear, non-stationary and strong randomness characteristics of wind power data, especially when the wind speed fluctuates violently or the trend changes suddenly, the prediction error increases significantly. In addition, the existing technology lacks sufficient depth in uncertainty analysis and cannot accurately evaluate the confidence interval and risk level of the prediction result, resulting in a lack of reliable risk warning ability in practical applications.

[0032] Due to the differences in geographical environment, wind turbine types and meteorological conditions of different wind farms, the performance of existing models (such as physical models or single algorithm models) drops significantly when applied across scenarios. For example, when a model optimized for a coastal wind farm is directly applied to an inland wind farm, the prediction accuracy is significantly reduced due to terrain and meteorological differences. The existing technology fails to effectively solve the problems of insufficient data feature extraction and weak noise and outlier processing capabilities, further limiting the generality of the model.

[0033] Although some deep learning models have high prediction accuracy, their computational complexity is high and time-consuming, and they cannot meet the requirements of power grid scheduling for real-time prediction. Existing methods are prone to falling into local optimal solutions or overfitting during the parameter optimization process, lacking an efficient global optimization mechanism, resulting in low model training efficiency and difficulty in balancing prediction accuracy and computational resource consumption.

[0034] As a preferred solution of a wind farm power generation prediction and uncertainty analysis system based on a hybrid intelligent algorithm according to the present invention, it is characterized by including: a data acquisition module, configured to obtain historical wind power data and wind speed data of the wind farm, perform time series arrangement on the data, and perform data preprocessing to remove outliers and missing data;

[0035] A wind speed classification module, configured to classify the wind speed data by using a fuzzy K-means clustering algorithm and divide the data set according to wind speed levels;

[0036] A model construction module, configured to establish multiple support vector regression models, each model corresponding to a different wind speed level, and optimize the parameters of each support vector regression model by using an enhanced harmony search algorithm; and,

[0037] An uncertainty analysis module, configured to perform uncertainty analysis on the prediction results of the support vector regression models by using a quantile regression method, calculate prediction intervals at different confidence levels, and optimize the parameters of the quantile regression model by using an enhanced harmony search algorithm.

[0038] A computer device, including a memory and a processor, where the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm are implemented.

[0039] A computer-readable storage medium, on which a computer program is stored, and is characterized in that when the computer program is executed by a processor, the steps of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm are implemented.

[0040] The beneficial effects of the present invention: Using a fuzzy k-means clustering algorithm to classify the time series data set of historical wind power output can effectively extract the characteristics of similar wind levels, improve the organization and correlation of data. This classification method can divide the data into different wind speed levels, such as gentle breeze, moderate wind, strong wind, etc., according to the irregularity of wind speed, enabling subsequent prediction models to be trained and optimized separately for different wind speed levels, thereby improving the accuracy and adaptability of prediction.

[0041] Using an enhanced harmony search (EHS) algorithm to optimize the parameters of each support vector regression model can achieve high-precision prediction of wind power output. The support vector regression model can effectively handle the non-linear relationship of wind power data, while the enhanced harmony search algorithm can automatically optimize the parameters of the support vector regression model, avoiding convergence to local minima and overfitting in the case of too long training time, and improving the generalization ability and prediction accuracy of the model. Description of the Drawings

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0043] Figure 1 It is the overall flowchart of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm provided by an embodiment of the present invention.

[0044] Figure 2 It is the structural schematic diagram of a multiple support vector regression model based on EHS for a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm provided by an embodiment of the present invention.

[0045] Figure 3 It is the experimental result diagram of a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm provided by an embodiment of the present invention. Detailed implementation manners

[0046] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific implementation manners of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0047] Example 1, referring to Figures 1-3 , which is the first embodiment of the present invention. This embodiment provides a method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm, including:

[0048] Obtain the historical wind power data and wind speed data of the wind farm, perform time series sorting on the data, and perform data preprocessing to remove outliers and missing data;

[0049] Use the fuzzy K-means clustering algorithm to classify the wind speed data and divide the data set according to the wind speed level;

[0050] Establish multiple support vector regression models, each model corresponding to a different wind speed level, and use the enhanced harmony search algorithm to optimize the parameters of each support vector regression model;

[0051] The quantile regression method is used to perform uncertainty analysis on the prediction results of the support vector regression model, calculate the prediction intervals at different confidence levels, and optimize the parameters of the quantile regression model using the enhanced harmony search algorithm.

[0052] The classification of the time series dataset of historical wind power output into different wind speed levels includes setting initial clustering centers based on the historical wind speed data of the wind farm and calculating the Euclidean distance between the wind speed observations and each clustering center;

[0053] The fuzzy K-means clustering algorithm is adopted. Based on the calculated Euclidean distance, the membership degree of each wind speed observation to different clustering centers is determined, and initial clustering is performed according to the principle of the minimum Euclidean distance;

[0054] The clustering centers are updated using the weighted mean calculation method, and the classification of data points is adjusted according to the membership degree. The iterative calculation is repeated until the preset convergence condition is met;

[0055] After the convergence condition is met, the final classification result of the wind speed data is output, and a dataset of multiple wind speed levels is formed.

[0056] In a preferred embodiment of the present invention, the fuzzy k-means clustering algorithm is used to classify the time series dataset of historical wind power output into different wind speed levels, such as gentle breeze, moderate wind, strong wind, etc. Each dataset consists of multiple data points, and the classified data will be used for subsequent deterministic prediction and uncertainty analysis. Using the k-means clustering algorithm to divide n observations into k clusters is as follows:

[0057]

[0058] Among them, Z j is the j-th observation, w ji is the synaptic weight, ||Z j -C i || is the Euclidean distance, C i is the i-th clustering center, and the formula is as follows:

[0059]

[0060] w ji is expressed as:

[0061]

[0062] Among them, L m is a weight index, usually set to 2.

[0063] Said establishing a plurality of support vector regression models includes calculating the deviation from the hyperplane to the data points based on the wind speed level data set, defining the unit normal vector, the slack variable, and the weight value of the penalty function, and establishing an objective optimization function;

[0064] Initialize the kernel function parameters, penalty coefficient, and slack variable of the support vector regression model, calculate the loss value using the training data, and solve the optimization objective function to obtain the initial model;

[0065] Based on the initialized SVR prediction model, train the data sets of different wind speed levels respectively, calculate the loss value of each model, and obtain the preliminary prediction results based on the adjustment of the penalty coefficient and the slack variable.

[0066] In a preferred embodiment of the present invention, the quantile regression (quantile regression) method based on EHS is used to perform uncertainty analysis on the predicted values and provide the confidence interval of the predicted values. Set the upper and lower confidence levels, estimate the unknown parameters using the enhanced harmony search algorithm, and generate the upper and lower confidence intervals to evaluate the uncertainty of the prediction results.

[0067] Let {yi: i = 1, 2,..., n} be a set of random samples, and {xi: i = 1, 2,..., n} be the corresponding independent variables. Among them, λ ∈ [0, 1] is the quantile factor, and β(λ) is the regression parameter, and its estimation is as follows:

[0068]

[0069] For any λ ∈ [0, 1], only one solution of β(λ) can be obtained as the λ quantile factor.

[0070] The quantile function formula is:

[0071]

[0072] Among them, β0 is a constant, β i (λ)(i = 1, 2,..., K) is the parameter of the quantile regression, K is the number of quantile variables, and Z i is the i-th quantile variable.

[0073] The upper and lower confidence intervals of the wind power prediction are expressed as:

[0074] P t,upper = P t,fore ×(1 + U(λ))

[0075] P t,lower = P t,fore ×(1 + U(1 - λ))

[0076] Among them, P t,fore is the predicted value at time t.

[0077] As Figure 2 shown, optimizing the parameters of each support vector regression model using the enhanced harmony search algorithm includes, after the completion of SVR training, initializing the enhanced harmony search algorithm, randomly generating multiple feasible parameter combinations, including the weight of the penalty function and the inflation parameter of the kernel function, and storing the parameter vectors in the harmony matrix;

[0078] Based on the random generation method, within the preset parameter range, randomly generate multiple candidate solutions according to the uniform distribution, calculate the SVR prediction error corresponding to each candidate solution, and screen the optimal solution according to the fitness value;

[0079] During the harmony search process, calculate the loss value of the support vector regression model based on the current parameter combination, and dynamically adjust the penalty coefficient and kernel function parameters of the support vector regression model according to the method combining local search and global search;

[0080] In each iteration process, select the parameter vector with the optimal fitness value, update the harmony matrix, and recalculate the loss value of the SVR prediction model. After meeting the optimization termination condition, determine the optimal parameter combination.

[0081] Performing uncertainty analysis includes, based on the trained SVR prediction model, obtaining the wind power prediction value, and constructing the corresponding independent variable sample set and random sample set;

[0082] Set the confidence level range, define the quantile factor, and calculate the regression parameters of the historical wind power data at different quantile levels;

[0083] Calculate the regression objective optimization function, with the prediction error as the optimization objective, calculate the quantile regression parameters based on the loss function, and minimize the regression error;

[0084] Calculate the quantile regression function, and calculate the upper confidence limit and lower confidence limit of the predicted value based on the regression parameters at different confidence levels.

[0085] In a preferred embodiment of the present invention, multiple support vector regression models are established, and each model corresponds to a different wind speed level. The enhanced harmony search algorithm is used to optimize the parameters of each support vector regression model to improve the prediction accuracy.

[0086] (1) Establish multiple support vector regression models

[0087]

[0088] where u is the unit normal vector to the hyperplane, ψ is the distance from the origin to the hyperplane, n is the number of training data, ψ k is the k-th slack variable, and σ is the weight of the penalty function.

[0089] (2) Use the enhanced harmony search algorithm for parameter optimization

[0090] Initially generate some feasible harmonics and store them in the harmony matrix (HM):

[0091]

[0092] Wherein, is the j-th weight of the penalty function, j = 1, 2,..., S, and S is the size of the harmony matrix, is the dilation parameter of the j-th kernel function.

[0093] Each column vector represents a feasible solution, which is randomly generated as follows:

[0094]

[0095] In the formula, h = 1, 2, and are the lower bound and the upper bound, and rand is a random number uniformly distributed between 0 and 1.

[0096] For each vector solution obtained using the above formula, the solution with the largest HM value is regarded as the optimal solution of SVR.

[0097] The parameters of the optimized quantile regression model include initializing the enhanced harmony search algorithm after calculating the confidence interval of the quantile regression, generating multiple combinations of quantile regression parameters, and storing them in the harmony matrix;

[0098] Based on the random generation method, within the preset parameter range, randomly generate multiple candidate solutions according to the uniform distribution, and calculate the quantile regression error corresponding to each candidate solution;

[0099] Calculate the fitness value of the current parameter combination, and adjust the regression parameters of the quantile regression model according to the method combining local search and global search;

[0100] In each iteration process, update the harmony matrix, and re-calculate the quantile regression error based on the adjusted quantile regression model parameters, and judge whether the optimization termination condition is satisfied;

[0101] After the optimization termination condition is satisfied, fix the finally optimized quantile regression parameters.

[0102] Evaluate the error of the prediction result, calculate the root mean square error and the mean absolute error, and adjust the model parameters based on the error evaluation result.

[0103] In a preferred embodiment of the present invention, the accuracy of the prediction result is evaluated by indicators such as the root mean square error (RMSE) and the mean absolute error (MAE).

[0104] (1) Root Mean Square Error (RMSE), the calculation formula is as follows:

[0105]

[0106] This index measures the square root of the model prediction error and reflects the accuracy of the prediction result.

[0107] (2) Mean Absolute Error (MAE), the calculation formula is as follows:

[0108]

[0109] This index measures the average absolute value of the model prediction error and reflects the stability of the model.

[0110] Refer to Figure 3 , the rated capacity of the wind turbine is 2 kw, and the corresponding wind speed is 12 m / s. Five SVR models of gentle breeze, gentle breeze, moderate wind, cool breeze and strong wind are adopted in this paper. In order to verify the performance of the proposed method, it is compared with the ANN algorithm.

[0111] Example 2 is the second example of the present invention, which is different from the previous example in that:

[0112] If the said function is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that makes a contribution to the prior art, or a 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 may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks or optical discs that can store program codes.

[0113] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definitional sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0114] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0115] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0116] Embodiment 3 is the fourth embodiment of the present invention, and this embodiment provides a wind farm power generation prediction and uncertainty analysis system based on a hybrid intelligent algorithm, including.

[0117] A data acquisition module, configured to obtain historical wind power data and wind speed data of a wind farm, perform time series sorting on the data, and perform data preprocessing to remove outliers and missing data;

[0118] A wind speed classification module, configured to classify the wind speed data using the fuzzy K-means clustering algorithm and divide the data set according to wind speed levels;

[0119] A model construction module, which is used to establish multiple support vector regression models, each model corresponding to a different wind speed level, and use an enhanced harmony search algorithm to optimize the parameters of each support vector regression model; and,

[0120] An uncertainty analysis module, which is used to perform uncertainty analysis on the prediction results of the support vector regression model by using the quantile regression method, calculate the prediction intervals at different confidence levels, and use the enhanced harmony search algorithm to optimize the parameters of the quantile regression model.

[0121] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm, characterized in that: including Obtain the historical wind power data and wind speed data of a wind farm, organize the data in a time series, and perform data preprocessing to remove outliers and missing data; Use the fuzzy K-means clustering algorithm to classify the wind speed data and divide the data set according to the wind speed levels; Establish multiple support vector regression models, each model corresponding to a different wind speed level, and use the enhanced harmony search algorithm to optimize the parameters of each support vector regression model; Adopt the quantile regression method to conduct uncertainty analysis on the prediction results of the support vector regression models, calculate the prediction intervals at different confidence levels, and use the enhanced harmony search algorithm to optimize the parameters of the quantile regression model.

2. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 1, wherein: The classification of the time series data set of the historical wind power output into different wind speed levels includes setting initial clustering centers based on the historical wind speed data of the wind farm and calculating the Euclidean distance between the wind speed observations and each clustering center; Adopt the fuzzy K-means clustering algorithm, based on the calculated Euclidean distance, determine the membership degrees of each wind speed observation to different clustering centers, and perform initial clustering according to the principle of the minimum Euclidean distance; Update the clustering centers using the weighted mean calculation method, and adjust the classification of the data points according to the membership degrees, and repeat the iterative calculation until the preset convergence condition is satisfied; After the convergence condition is satisfied, output the final classification result of the wind speed data and form data sets of multiple wind speed levels.

3. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 2, characterized in that: The establishment of multiple support vector regression models includes calculating the deviation from the hyperplane to the data points based on the wind speed level data set, defining the unit normal vector, slack variables, and the weight value of the penalty function, and establishing the objective optimization function; Initialize the kernel function parameters, penalty coefficient, and slack variables of the support vector regression model, calculate the loss value using the training data, and solve the optimization objective function to obtain the initial model; Based on the initialized SVR prediction model, train the data sets of different wind speed levels respectively, calculate the loss value of each model, and obtain the preliminary prediction results based on the adjustment of the penalty coefficient and slack variables.

4. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 3, characterized in that: The use of the enhanced harmony search algorithm to optimize the parameters of each support vector regression model includes, after the SVR training is completed, initializing the enhanced harmony search algorithm, randomly generating multiple feasible parameter combinations, including the weight value of the penalty function and the inflation parameter of the kernel function, and storing the parameter vectors in the harmony matrix; Based on the random generation method, within the preset parameter range, randomly generate multiple candidate solutions according to the uniform distribution, calculate the SVR prediction error corresponding to each candidate solution, and screen the optimal solution according to the fitness value; During the harmony search process, calculate the loss value of the support vector regression model based on the current parameter combination, and dynamically adjust the penalty coefficient and kernel function parameters of the support vector regression model according to the method combining local search and global search; In each iteration process, select the parameter vector with the optimal fitness value, update the harmony matrix, and recalculate the loss value of the SVR prediction model. After the optimization termination condition is satisfied, determine the optimal parameter combination.

5. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 4, wherein: The uncertainty analysis includes obtaining the wind power prediction value based on the trained SVR prediction model, and constructing the corresponding independent variable sample set and random sample set; Set the confidence level range, define the quantile factor, and calculate the regression parameters of the historical wind power data at different quantile levels; Calculate the regression objective optimization function, take the prediction error as the optimization objective, calculate the quantile regression parameters based on the loss function, and minimize the regression error; Calculate the quantile regression function, and calculate the upper confidence limit and lower confidence limit of the prediction value based on the regression parameters at different confidence levels.

6. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 4, wherein: The parameters for optimizing the quantile regression model include initializing the enhanced harmony search algorithm after calculating the confidence interval of the quantile regression, generating multiple combinations of quantile regression parameters, and storing them in the harmony matrix; Based on the random generation method, within the preset parameter range, randomly generate multiple candidate solutions according to the uniform distribution, and calculate the quantile regression error corresponding to each candidate solution; Calculate the fitness value of the current parameter combination, and adjust the regression parameters of the quantile regression model according to the method combining local search and global search; In each iteration process, update the harmony matrix, and recalculate the quantile regression error based on the adjusted quantile regression model parameters to determine whether the optimization termination condition is satisfied; After the optimization termination condition is satisfied, fix the finally optimized quantile regression parameters.

7. The method for wind farm power generation prediction and uncertainty analysis based on a hybrid intelligent algorithm according to claim 4, wherein: Evaluate the error of the prediction result, calculate the root mean square error and mean absolute error, and adjust the model parameters based on the error evaluation result.

8. A wind farm power generation prediction and uncertainty analysis system based on a hybrid intelligent algorithm, which applies a wind farm power generation prediction and uncertainty analysis method according to any one of claims 1 to 7, characterized in that, Including: A data acquisition module for obtaining the historical wind power data and wind speed data of the wind farm, arranging the data in time series, and performing data preprocessing to remove outliers and missing data; A wind speed classification module for classifying the wind speed data by using the fuzzy K-means clustering algorithm and dividing the data set according to the wind speed level; A model construction module for establishing multiple support vector regression models, each model corresponding to a different wind speed level, and optimizing the parameters of each support vector regression model by using the enhanced harmony search algorithm; And, An uncertainty analysis module for performing uncertainty analysis on the prediction results of the support vector regression model by using the quantile regression method, calculating the prediction intervals at different confidence levels, and optimizing the parameters of the quantile regression model by using the enhanced harmony search algorithm.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of a wind farm power generation prediction and uncertainty analysis method based on a hybrid intelligent algorithm according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of a wind farm power generation prediction and uncertainty analysis method based on a hybrid intelligent algorithm according to any one of claims 1 to 7.

Images