Multi-mode self-adaptive switching ultra-short-term wind power prediction method and multi-mode self-adaptive switching ultra-short-term wind power prediction system

Abstract

The invention discloses a multi-mode self-adaptive switching ultra-short-term wind power prediction method and system, and belongs to the technical field of industrial control, and the method comprises the steps: obtaining the meteorological data of a to-be-predicted time and the parameter data of a wind driven generator; marking timestamps based on the meteorological data and the parameter data of the wind driven generator, and performing data alignment to obtain an initial prediction data set; performing abnormal value detection on the data by using a clustering algorithm, performing sorting according to timestamps after detecting and eliminating abnormal data, performing normalization operation on an initial data set, and taking meteorological data after normalization operation and parameter data of the wind driven generator as input features; and according to the change of the wind speed measured in real time, performing multi-mode adaptive switching, inputting the input characteristics into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model, and predicting to obtain the output power of the fan. The method can improve the prediction accuracy.

Description

Technical Field

[0001] The invention belongs to the field of power system prediction, and relates to a multi-mode adaptive switching ultra-short-term wind power prediction method and system. Background Art

[0002] The output power of a wind turbine is an important indicator for evaluating the performance of a wind turbine, which directly determines the efficiency of the wind turbine in converting electrical energy into mechanical energy. The output power of a wind turbine refers to the power output of the wind turbine, that is, the power output of the rotor, which represents the ability of the wind turbine to convert electrical energy into mechanical energy. In the field of wind energy conversion, this indicator is a key parameter for evaluating the performance of a wind turbine. The calculation of the output power of a wind turbine usually involves multiple factors, including air density, rotor area, and wind speed. The output power of a wind turbine is one of the key indicators for evaluating the performance of a wind turbine, which directly determines the power generation efficiency and economic benefits of the wind turbine. By optimizing the blade design, controlling the speed of the wind turbine, and considering environmental conditions, the output power and power generation efficiency of the wind turbine can be effectively improved, promoting the development and application of wind power generation technology.

[0003] Ultra-short-term wind power forecasting refers to the prediction of the output power of wind turbines within the next 4 hours, providing technical support for dispatching and other related departments to adjust the spare capacity and unit control strategy to ensure the stable operation of the power system. Nowadays, common wind power forecasting methods can be divided into statistical learning methods and physical forecasting methods. However, both methods have certain defects. The physical forecasting method requires too much computing resources and requires very detailed meteorological data as support, which is not suitable for ultra-short-term wind power forecasting with high frequency refresh. Since the relationship between wind power and meteorological conditions and wind turbine parameters is not a simple linear relationship, traditional statistical learning methods may not be able to capture this nonlinear nature well. Secondly, when using statistical learning methods for wind power forecasting, the selection and extraction of appropriate features has a great impact on the accuracy of the prediction, and feature extraction is a task that relies heavily on personal experience and time. Summary of the invention

[0004] In order to improve the accuracy of ultra-short-term wind power prediction, the present invention proposes a multi-modal adaptive switching ultra-short-term wind power prediction method. The method of the present invention plays an important role in improving the predictability of the wind energy industry, reducing costs, and increasing the penetration rate of renewable energy.

[0005] To achieve the above purpose, the present invention adopts the following technical means:

[0006] The first aspect of the present invention is to provide a multi-mode adaptive switching ultra-short-term wind power prediction method, comprising:

[0007] Obtain meteorological data and wind turbine parameter data for the time to be predicted;

[0008] Mark the timestamp based on the meteorological data and wind turbine parameter data, and align the data to obtain the initial prediction data set;

[0009] The clustering algorithm is used to detect outliers in the data. After detecting and removing the abnormal data, the data is sorted according to the timestamp, and the initial data set is normalized. The normalized meteorological data and wind turbine parameter data are used as input features.

[0010] According to the changes in real-time measured wind speed, multi-modal adaptive switching is performed, and the input features are input into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model to predict the wind turbine output power.

[0011] As a further improvement of the present invention, the multi-mode adaptive switching includes:

[0012] Set the threshold wind speed and minimum cut-in wind speed, and implement the mode switching mechanism according to the changes in real-time measured wind speed;

[0013] The switching mechanism between multiple models has a switch. When the input wind speed is greater than the threshold wind speed, the control switch is in the on state, and the high-fluctuation wind speed-wind power model is used to predict the wind power. When the input wind speed is less than the threshold wind speed, the control switch is in the off state, and the low-fluctuation wind speed-wind power prediction model is used to predict the wind power.

[0014] As a further improvement of the present invention, after predicting the wind turbine output power, the method further includes:

[0015] The evaluation index score of the ultra-short-term wind power prediction model with multi-modal adaptive switching is calculated according to the predicted wind turbine output power, and a fitting diagram of the ultra-short-term wind power prediction with multi-modal adaptive switching and the actual wind power curve is drawn.

[0016] As a further improvement of the present invention, the training method of the pre-trained high-fluctuation wind speed-wind power prediction model or the pre-trained low-fluctuation wind speed-wind power prediction model comprises:

[0017] Obtain historical meteorological data, wind turbine output power, and wind turbine parameter data at the output power;

[0018] Mark the timestamps for historical meteorological data, wind turbine parameter data and output power, and perform data alignment;

[0019] After data alignment, the missing data in each column is filled, and then the output power statistics set is constructed. The abnormal data of the wind turbine is removed to obtain the initial training data set;

[0020] The clustering algorithm is used to detect outliers in the initial training data set. After detecting and removing the abnormal data, the data is sorted according to the timestamp and normalized to obtain a data matrix. The output power in the data matrix is ​​used as the output result, and the meteorological data and wind turbine parameter data are used as input features.

[0021] The input parameters are promoted to a high-dimensional space using a polynomial regression algorithm, and then a high-fluctuation wind speed-wind power prediction model is established using a multivariate linear regression algorithm. The high-fluctuation wind speed-wind power prediction model is trained in a training set, and parameter factors are calculated. Based on the parameter factors, a pre-trained high-fluctuation wind speed-wind power prediction model is obtained.

[0022] Taking the long short-term memory network as the basic model and combining it with the ultra-short-term wind power prediction scenario of multimodal adaptive switching, a low-fluctuation wind speed-wind power attention neural network is established. The attention neural network is iteratively trained in the training set to obtain the optimal model parameter values. Based on the optimal model parameter values, a pre-trained low-fluctuation wind speed-wind power prediction model is obtained.

[0023] As a further improvement of the present invention, the data completion method uses a nearest neighbor interpolation method to complete the missing data of each feature column.

[0024] As a further improvement of the present invention, the method of using a clustering algorithm to perform outlier detection on an initial training data set includes:

[0025] A hierarchical clustering algorithm is used to divide the data into specified wind speed intervals, and then clustering is performed by calculating the similarity between data points. Subsequently, similar classes are merged according to the distance between the center points of the clusters, and hierarchical division is performed according to the specified wind speed interval until the specified level is reached and the clustering process is stopped to detect abnormal data.

[0026] Eliminating abnormal data refers to eliminating data sets whose wind turbine output power is less than the set rated power and whose pitch angle is greater than the set pitch angle threshold.

[0027] As a further improvement of the present invention, in the high-fluctuation wind speed-wind power prediction model established by using a multiple linear regression algorithm, the multiple linear regression algorithm is:

[0028] y=β0+β1x1+β2x2+β3x3+β4x4+ε

[0029] Among them, β0 is a constant term, which represents the wind power level when all independent variables are zero; β1, β2, β3, and β4 are regression coefficients; x1, x2, x3, and x4 represent the input independent variables; and ε represents the random error constant.

[0030] The second aspect of the present invention is to provide a multi-mode adaptive switching ultra-short-term wind power prediction system, comprising:

[0031] An acquisition module is used to acquire meteorological data and wind turbine parameter data at the time to be predicted;

[0032] A marking module is used to mark timestamps based on meteorological data and wind turbine parameter data, and perform data alignment to obtain an initial prediction data set;

[0033] The normalization module is used to detect outliers on the data using a clustering algorithm, sort the data by timestamp after detecting and removing the abnormal data, and perform normalization on the initial data set, using the normalized meteorological data and wind turbine parameter data as input features;

[0034] The multimodal prediction module is used to perform multimodal adaptive switching according to the changes in real-time measured wind speed, input the input features into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model, and predict the wind turbine output power.

[0035] The third aspect of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the multi-modal adaptive switching ultra-short-term wind power prediction method when executing the computer program.

[0036] A fourth aspect of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the multi-modal adaptive switching ultra-short-term wind power prediction method is implemented.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention establishes an adaptive switching mechanism between the high-fluctuation wind speed-wind power model and the low-fluctuation wind speed-wind power attention neural network, analyzes the absolute error and explained variance score of the model, and finds that the multi-modal adaptive switching ultra-short-term wind power prediction method proposed in the present invention can improve the accuracy of the prediction, and plays an important role in improving the predictability of the wind energy industry, reducing costs, and increasing the penetration rate of renewable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flow chart of a multi-mode adaptive switching ultra-short-term wind power prediction method provided by the present invention;

[0040] Figure 2 A wind power prediction result fitting curve using the method of the present invention;

[0041] Figure 3 The present invention provides a multi-mode adaptive switching ultra-short-term wind power prediction system;

[0042] Figure 4 A schematic diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0044] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0045] In addition, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements explicitly listed, but may include other steps or elements not explicitly listed or inherent to such process, method, product, or apparatus.

[0046] Terminology explanation:

[0047] SCADA system, the full name of which is Supervisory Control And Data Acquisition, is a computer-based production process control and scheduling automation system.

[0048] like Figure 1 As shown, the first object of the present invention is to provide a multi-mode adaptive switching ultra-short-term wind power prediction method, comprising the following steps:

[0049] S1, obtaining meteorological data and wind turbine parameter data for the time to be predicted;

[0050] S2, marking timestamps based on meteorological data and wind turbine parameter data, and aligning the data to obtain an initial prediction data set;

[0051] S3, using clustering algorithm to detect outliers on the data, sorting them according to timestamps after detecting and removing abnormal data, and normalizing the initial data set, using the normalized meteorological data and wind turbine parameter data as input features;

[0052] S4, according to the change of real-time measured wind speed, multi-modal adaptive switching is performed, and the input features are input into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model to predict the wind turbine output power.

[0053] The principle of the method of the present invention is:

[0054] Steps S1-S4 are based on the demand for wind power prediction, combined with meteorological data and wind turbine parameter data, and use clustering algorithms to clean the data, and then predict the wind turbine output power through a pre-trained model.

[0055] Among them, S1-S2 is data acquisition and alignment, specifically:

[0056] S1: Obtain meteorological data (such as wind speed, wind direction, temperature, etc.) and wind turbine parameter data (such as pitch angle and rotation speed, etc.) at the time to be predicted.

[0057] S2: Mark the timestamp based on these data and perform data alignment to ensure the temporal consistency of meteorological data and wind turbine parameter data to form an initial prediction data set.

[0058] S3 is data cleaning and normalization. Clustering algorithms (such as K-Means, DBSCAN, etc.) are used to detect outliers in the initial data set, identify and remove abnormal data. The cleaned data is sorted and normalized according to certain rules (such as minimum-maximum normalization) to make the data on the same scale for subsequent model processing.

[0059] S4 is multi-mode adaptive switching and prediction, which performs multi-mode adaptive switching according to the changes in real-time measured wind speed. This means that when the wind speed fluctuates greatly, the high-fluctuation wind speed-wind power prediction model is used; when the wind speed fluctuates slightly, the low-fluctuation wind speed-wind power prediction model is used.

[0060] The normalized meteorological data and wind turbine parameter data are used as input features and input into the selected prediction model to predict the wind turbine output power.

[0061] Through data alignment and cleaning, the quality and accuracy of input data are ensured, providing a reliable basis for subsequent predictions. The multi-modal adaptive switching strategy can select the most appropriate prediction model according to different wind speed fluctuations, improving the accuracy and applicability of the prediction. By using the pre-trained model for prediction, combined with the comprehensive advantages of physical methods and statistical methods, the wind turbine output power can be estimated more accurately. Model switching is performed according to the changes in real-time measured wind speed to ensure the real-time and effectiveness of the prediction results.

[0062] Most importantly, the training method of the pre-trained high-fluctuation wind speed-wind power prediction model or the pre-trained low-fluctuation wind speed-wind power prediction model of the present invention includes:

[0063] Obtain historical meteorological data, wind turbine output power, and wind turbine parameter data at the output power;

[0064] Mark the timestamps for historical meteorological data, wind turbine parameter data and output power, and perform data alignment;

[0065] After data alignment, the missing data in each column is filled, and then the output power statistics set is constructed. The abnormal data of the wind turbine is removed to obtain the initial training data set;

[0066] The clustering algorithm is used to detect outliers in the initial training data set. After detecting and removing the abnormal data, the data is sorted according to the timestamp and normalized to obtain a data matrix. The output power in the data matrix is ​​used as the output result, and the meteorological data and wind turbine parameter data are used as input features.

[0067] The input parameters are promoted to a high-dimensional space using a polynomial regression algorithm, and then a high-fluctuation wind speed-wind power prediction model is established using a multivariate linear regression algorithm. The high-fluctuation wind speed-wind power prediction model is trained in a training set, and parameter factors are calculated. Based on the parameter factors, a pre-trained high-fluctuation wind speed-wind power prediction model is obtained.

[0068] Taking the long short-term memory network as the basic model and combining it with the ultra-short-term wind power prediction scenario of multimodal adaptive switching, a low-fluctuation wind speed-wind power attention neural network is established. The attention neural network is iteratively trained in the training set to obtain the optimal model parameter values. Based on the optimal model parameter values, a pre-trained low-fluctuation wind speed-wind power prediction model is obtained.

[0069] Accordingly, the multi-mode adaptive switching may be performed, specifically including:

[0070] Set the threshold wind speed and minimum cut-in wind speed, and implement the mode switching mechanism according to the changes in real-time measured wind speed;

[0071] The switching mechanism between multiple models has a switch. When the input wind speed is greater than the threshold wind speed, the control switch is in the on state, and the high-fluctuation wind speed-wind power model is used to predict the wind power. When the input wind speed is less than the threshold wind speed, the control switch is in the off state, and the low-fluctuation wind speed-wind power prediction model is used to predict the wind power.

[0072] Optionally, this threshold wind speed is generally determined based on the performance of the wind turbine, historical wind speed data, and predicted demand.

[0073] By setting the threshold wind speed and adopting a multi-model switching mechanism, the system can select the appropriate prediction model according to the data characteristics under different wind speed conditions, thereby improving the prediction accuracy. The high-fluctuation wind speed-wind power prediction model can handle situations with large wind speed changes, while the low-fluctuation wind speed-wind power prediction model can handle scenes with relatively stable wind speeds.

[0074] The multi-model switching mechanism enables the system to adapt to different wind speed conditions. Whether the wind speed fluctuates greatly or is relatively stable, the system can provide relatively accurate prediction results. This flexibility makes the system more adaptable and reliable in practical applications.

[0075] By selecting the appropriate prediction model, the system can reduce prediction errors and improve the accuracy of prediction results. This is of great significance to the operation management and scheduling of wind farms, and can help operators better formulate power generation plans and resource scheduling strategies.

[0076] As a preferred solution, after predicting the wind turbine output power, the method further includes:

[0077] The evaluation index score of the ultra-short-term wind power prediction model with multi-modal adaptive switching is calculated according to the predicted wind turbine output power, and a fitting diagram of the ultra-short-term wind power prediction with multi-modal adaptive switching and the actual wind power curve is drawn.

[0078] The method of the present invention is described in detail below with reference to specific embodiments.

[0079] An embodiment of the present invention provides an ultra-short-term wind power prediction method with multi-modal adaptive switching of multiple models dynamically switching. The embodiment of the present invention mainly involves the construction and correction of wind power time series data, high-fluctuation wind speed-wind power prediction model, low-fluctuation wind speed-wind power attention neural network training and the establishment of multi-model dynamic switching rules, specifically a ultra-short-term wind power prediction method with multi-modal adaptive switching.

[0080] The specific steps include:

[0081] 1. Training Phase

[0082] Step 1: Obtain historical meteorological data such as wind speed, wind direction, temperature, air density, air pressure, turbulence, precipitation, etc. for a certain place within one month. For example, the time interval between each piece of data is 15 minutes.

[0083] Step 2: Obtain the output power data of the wind turbines of a certain wind farm in the area within one month and the wind turbine parameter data such as the pitch angle and speed of the wind turbines under the output power, for example, the interval of each data is also 15 minutes;

[0084] Step 3: Timestamp the historical meteorological data, wind turbine parameter data and output power obtained in Steps 1 and 2; in order to improve the accuracy, for example, the time accuracy is accurate to seconds, the time interval between each data point is 15 minutes, and the aligned data is saved in ".csv" format;

[0085] Step 4: Complete the missing data in each column in step 3, and then use the wind turbine SCADA system to obtain the 15-minute periodic data of wind speed, pitch angle, speed, output power and other parameters of the wind turbine when it reaches rated power, build an output power statistical set, and remove abnormal wind turbine data to obtain the initial training data set;

[0086] Optionally, in step 4, the data completion method uses the nearest neighbor interpolation method to complete the missing data in each feature column; in the present invention, the elimination of abnormal wind turbine data refers to eliminating the data set whose wind turbine output power is less than the set rated power and whose pitch angle is greater than the set pitch angle threshold.

[0087] Step 5: Use a clustering algorithm to detect outliers on the data divided in step 4), sort them according to timestamps after removing the abnormal data, and normalize the initial data set to obtain the final data matrix, in which the output power is used as the output result, and the wind turbine parameters such as wind speed, wind direction, and rotation speed are input features;

[0088] Optionally, the clustering algorithm in step 5 adopts a hierarchical clustering algorithm, which first divides the data by specifying a wind speed interval, then performs clustering by calculating the similarity between data points, then merges similar classes based on the distance between cluster center points, and performs hierarchical division according to the specified wind speed interval, and stops the clustering process after division to the specified level, thereby realizing the detection of abnormal data.

[0089] Step 6: Divide the data matrix obtained in step 5 into a training set and a test set, use the polynomial regression algorithm to promote the input parameters such as wind speed, wind direction, temperature, and air pressure to a high-dimensional space, and then use the multivariate linear regression algorithm to establish a high-fluctuation wind speed-wind power prediction model, and train the model in the training set. After verification in the validation set, calculate the parameter factors of the model, and obtain the pre-trained high-fluctuation wind speed-wind power prediction model based on the parameter factors;

[0090] Optionally, in step 6, the data is divided according to time periods, and the principle of the multivariate linear regression algorithm is as follows:

[0091] y=β0+β1x1+β2x2+β3x3+β4x4+ε

[0092] Among them, β0 is a constant term, which represents the wind power level when all independent variables are zero; x1, x2, x3, and x4 represent the input independent variables; β1, β2, β3, and β4 are regression coefficients, that is, the model parameter factors that need to be calculated, representing the influence of each independent variable on the wind power calculation; ε represents the random error constant, which represents the noise and uncertainty that cannot be explained by the model.

[0093] Step 7: Based on the long short-term memory network as the basic model and combined with the ultra-short-term wind power prediction scenario of multimodal adaptive switching, a low-fluctuation wind speed-wind power attention neural network with 16 neurons and a network depth of 4 is established. The attention neural network is iteratively trained on the training set, and after verification based on the validation set, a pre-trained low-fluctuation wind speed-wind power prediction model is obtained based on the optimal model parameter values.

[0094] Optionally, the parameters for iterative training of the attention neural network in step 7 are as follows: the batch size is set to 256, the dropout rate is set to 0.15, the learning rate is set to 0.12, the actual iteration stage is 11, and the prediction time for each sample is 0.05 milliseconds.

[0095] 2. Prediction Phase

[0096] Step 8: Ultra-short-term wind power prediction with multi-mode adaptive switching in the present invention refers to the prediction of wind turbine output power based on the meteorological data and wind turbine input parameters for the next 4 hours. First, the meteorological data parameters for the next 4 hours are corrected using the model to obtain the corrected meteorological data;

[0097] Optionally, the model for correcting meteorological data in step 8 is corrected using a long short-term memory network.

[0098] Step 9: Set the threshold wind speed and the minimum cut-in wind speed. According to the changes in the real-time measured wind speed, establish a switching mechanism between multiple models to predict the ultra-short-term wind power of multi-modal adaptive switching within the next 4 hours. When the real-time wind speed is lower than the minimum cut-in wind speed, the wind power forecast is 0.

[0099] Optionally, the switching mechanism between the multiple models in step 9 is specifically a switch structure, wherein the threshold wind speed set in step 9 controls the start and stop of the switch. When the input wind speed is greater than the threshold wind speed, the control switch is in the on state, and a high-fluctuation wind speed-wind power model is used to perform ultra-short-term wind power prediction with multi-modal adaptive switching. When the input wind speed is less than the threshold wind speed, the control switch is in the off state, and a low-fluctuation wind speed-wind power attention neural network is used to perform ultra-short-term wind power prediction with multi-modal adaptive switching.

[0100] 3. Verification Phase

[0101] Step 10: Calculate the evaluation index score of the model based on the wind power data for the next 4 hours predicted in step 9, and draw a fitting graph of the ultra-short-term wind power prediction and the actual wind power curve of the multi-modal adaptive switching.

[0102] The model evaluation indicators in step 10 are evaluated using mean absolute error and explained variance score, where the lower the mean absolute error score and the higher the explained variance score, the higher the robustness and accuracy of the model.

[0103] like Figure 2 As shown, the method of the present invention is used to fit the real power curve and the predicted power curve. From the fitting results, it can be seen that the predicted result curve of the present invention has a high degree of fit with the actual power curve. Therefore, this method has high accuracy and meets conventional prediction requirements.

[0104] like Figure 3 As shown, the second object of the present invention is to provide a multi-mode adaptive switching ultra-short-term wind power prediction system, comprising:

[0105] An acquisition module is used to acquire meteorological data and wind turbine parameter data at the time to be predicted;

[0106] A marking module is used to mark timestamps based on meteorological data and wind turbine parameter data, and perform data alignment to obtain an initial prediction data set;

[0107] The normalization module is used to detect outliers on the data using a clustering algorithm, sort the data by timestamp after detecting and removing the abnormal data, and perform normalization on the initial data set, using the normalized meteorological data and wind turbine parameter data as input features;

[0108] The multimodal prediction module is used to perform multimodal adaptive switching according to the changes in real-time measured wind speed, input the input features into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model, and predict the wind turbine output power.

[0109] like Figure 4As shown, the third object of the present invention is to provide an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the multi-modal adaptive switching ultra-short-term wind power prediction method when executing the computer program.

[0110] A fourth object of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the ultra-short-term wind power prediction method with multi-mode adaptive switching is implemented.

[0111] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium).

[0116] As is well known to those of ordinary skill in the art, the term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storing information (such as computer readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A multi-mode adaptive switching ultra-short-term wind power prediction method, characterized in that: include: Obtain meteorological data and wind turbine parameter data for the time to be predicted; Mark the timestamp based on the meteorological data and wind turbine parameter data, and align the data to obtain the initial prediction data set; The clustering algorithm is used to detect outliers in the data. After detecting and removing the abnormal data, the data is sorted according to the timestamp, and the initial data set is normalized. The normalized meteorological data and wind turbine parameter data are used as input features. According to the changes in real-time measured wind speed, multi-modal adaptive switching is performed, and the input features are input into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model to predict the wind turbine output power.

2. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 1 is characterized in that: The multi-mode adaptive switching includes: Set the threshold wind speed and minimum cut-in wind speed, and implement the mode switching mechanism according to the changes in real-time measured wind speed; The switching mechanism between multiple models has a switch. When the input wind speed is greater than the threshold wind speed, the control switch is in the on state, and the high-fluctuation wind speed-wind power model is used to predict the wind power. When the input wind speed is less than the threshold wind speed, the control switch is in the off state, and the low-fluctuation wind speed-wind power prediction model is used to predict the wind power.

3. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 1 is characterized in that: After the wind turbine output power is predicted, the following steps are also included: The evaluation index score of the ultra-short-term wind power prediction model with multi-modal adaptive switching is calculated according to the predicted wind turbine output power, and a fitting diagram of the ultra-short-term wind power prediction with multi-modal adaptive switching and the actual wind power curve is drawn.

4. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 1 is characterized in that: The training method of the pre-trained high-fluctuation wind speed-wind power prediction model or the pre-trained low-fluctuation wind speed-wind power prediction model comprises: Obtain historical meteorological data, wind turbine output power, and wind turbine parameter data at the output power; Mark the timestamps for historical meteorological data, wind turbine parameter data and output power, and perform data alignment; After data alignment, the missing data in each column is filled, and then the output power statistics set is constructed. The abnormal data of the wind turbine is removed to obtain the initial training data set; The clustering algorithm is used to detect outliers in the initial training data set. After detecting and removing the abnormal data, the data is sorted according to the timestamp and normalized to obtain a data matrix. The output power in the data matrix is ​​used as the output result, and the meteorological data and wind turbine parameter data are used as input features. The input parameters are promoted to a high-dimensional space using a polynomial regression algorithm, and then a high-fluctuation wind speed-wind power prediction model is established using a multivariate linear regression algorithm. The high-fluctuation wind speed-wind power prediction model is trained in a training set, and parameter factors are calculated. Based on the parameter factors, a pre-trained high-fluctuation wind speed-wind power prediction model is obtained. Taking the long short-term memory network as the basic model and combining it with the ultra-short-term wind power prediction scenario of multimodal adaptive switching, a low-fluctuation wind speed-wind power attention neural network is established. The attention neural network is iteratively trained in the training set to obtain the optimal model parameter values. Based on the optimal model parameter values, a pre-trained low-fluctuation wind speed-wind power prediction model is obtained.

5. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 4 is characterized in that: The data completion method uses the nearest neighbor interpolation method to complete the missing data of each feature column.

6. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 4 is characterized in that: The method of using a clustering algorithm to perform outlier detection on an initial training data set includes: A hierarchical clustering algorithm is used to divide the data into specified wind speed intervals, and then clustering is performed by calculating the similarity between data points. Subsequently, similar classes are merged according to the distance between the center points of the clusters, and hierarchical division is performed according to the specified wind speed interval until the specified level is reached and the clustering process is stopped to detect abnormal data. Eliminating abnormal data refers to eliminating data sets whose wind turbine output power is less than the set rated power and whose pitch angle is greater than the set pitch angle threshold.

7. The multi-mode adaptive switching ultra-short-term wind power prediction method according to claim 4 is characterized in that: In the high-fluctuation wind speed-wind power prediction model established by using the multivariate linear regression algorithm, the multivariate linear regression algorithm is: y=β0+β1x1+β2x2+β3x3+β4x4+ε Among them, β0 is a constant term, which represents the wind power level when all independent variables are zero; β1, β2, β3, and β4 are regression coefficients; x1, x2, x3, and x4 represent the input independent variables; and ε represents the random error constant.

8. A multi-mode adaptive switching ultra-short-term wind power prediction system, characterized in that: include: An acquisition module is used to acquire meteorological data and wind turbine parameter data at the time to be predicted; A marking module is used to mark timestamps based on meteorological data and wind turbine parameter data, and perform data alignment to obtain an initial prediction data set; The normalization module is used to detect outliers on the data using a clustering algorithm, sort the data by timestamp after detecting and removing the abnormal data, and perform normalization on the initial data set, using the normalized meteorological data and wind turbine parameter data as input features; The multimodal prediction module is used to perform multimodal adaptive switching according to the changes in real-time measured wind speed, input the input features into a pre-trained high-fluctuation wind speed-wind power prediction model or a pre-trained low-fluctuation wind speed-wind power prediction model, and predict the wind turbine output power.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the ultra-short-term wind power prediction method with multi-mode adaptive switching as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the ultra-short-term wind power prediction method with multi-mode adaptive switching according to any one of claims 1 to 7 is implemented.

Images