Wind power prediction system integrating ensemble learning and time sequence analysis

Abstract

The invention discloses a wind power prediction system fusing ensemble learning and time sequence analysis, and relates to the technical field of new energy, and the method comprises the steps: extracting a characteristic variable matrix of wind power, determining a weight identifier, carrying out the mining of a time-phase linear law based on the characteristic variable matrix, and constructing a multivariate time sequence curve. And constructing a power prediction model integrating a plurality of base predictors by combining a multivariate time sequence curve and adopting a data-driven training method. And on the basis of the model, monitoring and returning multiple paths of feature data, analyzing a prediction task type, and determining final power prediction data. And displaying a power prediction result to a terminal user through visual display, and reversely guiding operation and management of the wind turbine generator through data, so that efficient prediction and real-time optimization are realized. The technical problems of inaccurate wind power prediction, insufficient timeliness and poor response to complex environment change in the prior art are solved, and the technical effect of improving the accuracy and timeliness of wind power prediction is achieved.

Description

Technical Field

[0001] The present application relates to the field of new energy technology, and in particular to a wind power prediction system integrating integrated learning and time series analysis. Background Art

[0002] Wind power forecasting plays an important role in the operation management of wind farms and grid dispatching. Since the output of wind power is highly random and volatile and is greatly affected by meteorological conditions, accurate wind power forecasting is crucial to ensuring the stable operation of the power system, improving the efficiency of grid dispatching, and promoting the consumption of new energy. However, traditional power forecasting methods often find it difficult to achieve high-precision forecasts when dealing with complex time series data and multi-dimensional feature variables, and have limited performance in dealing with various forecasting tasks (such as short-term and long-term forecasts). Therefore, there is an urgent need for a new wind power forecasting system that can integrate multi-dimensional data, combine ensemble learning and time series analysis, so as to improve the forecasting accuracy, optimize the grid dispatching plan, and achieve efficient utilization of new energy power.

[0003] At the current stage, relevant technologies have technical problems such as inaccurate wind power prediction, insufficient timeliness and poor response to complex environmental changes. Summary of the invention

[0004] The present application provides a wind power prediction system that integrates integrated learning and time series analysis. It extracts the characteristic variable matrix of wind power, first determines the weight identifier, and then mines the linear law of the phase based on the characteristic variable matrix to construct a multivariate time series curve. On this basis, combined with the multivariate time series curve, a data-driven training method is used to construct a power prediction model that integrates multiple base predictors. Then, based on the model, multi-channel feature data is monitored and transmitted back, and the prediction task type is analyzed to determine the final power prediction data. Finally, the power prediction results are presented to the end user through a visual display. At the same time, these data are used to reversely guide the operation and management of the wind turbine, achieve efficient prediction and real-time optimization, and achieve the technical effect of improving the accuracy and timeliness of wind power prediction.

[0005] This application provides a wind power prediction system integrating integrated learning and time series analysis, including:

[0006] A characteristic variable matrix mining module, the characteristic variable matrix mining module is used to mine the characteristic variable matrix based on wind power, wherein the characteristic variable matrix has a weight identifier; a time series curve construction module, the time series curve construction module is used to mine the phase linear law for the characteristic variable matrix and construct a multivariate time series curve; a power prediction model construction module, the power prediction model construction module is used to combine the multivariate time series curve, perform data-driven training, and construct a power prediction model, wherein the power prediction model integrates multiple base predictors; a power prediction data determination module, the power prediction data determination module is used to monitor multiple characteristic data and transmit them back based on the characteristic variable matrix, and perform prediction analysis based on the prediction task type in combination with the power prediction model to determine the power prediction data; a visualization display module, the visualization display module is used to perform terminal visualization of the power prediction data and reversely guide unit management.

[0007] The wind power prediction system proposed in this application, which integrates integrated learning and time series analysis, first determines the weight identifier by extracting the characteristic variable matrix of wind power, and then mines the linear laws of the phases based on the characteristic variable matrix to construct a multivariate time series curve. On this basis, combined with the multivariate time series curve, a data-driven training method is adopted to construct a power prediction model integrating multiple base predictors. Then, based on this model, multi-channel feature data is monitored and transmitted back, and the prediction task type is analyzed to determine the final power prediction data. Finally, the power prediction results are presented to the end user through a visual display. At the same time, these data are used to reversely guide the operation and management of the wind turbines, realize efficient prediction and real-time optimization, and achieve the technical effect of improving the accuracy and timeliness of wind power prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0009] Figure 1 A schematic diagram of the structure of a wind power prediction system integrating integrated learning and time series analysis provided in an embodiment of the present application;

[0010] Figure 2 A schematic diagram of the timing curve construction module structure of a wind power prediction system that integrates integrated learning and timing analysis provided in an embodiment of the present application.

[0011] Explanation of the reference numerals: characteristic variable matrix mining module 10 , timing curve construction module 20 , power prediction model construction module 30 , power prediction data determination module 40 , visualization display module 50 . DETAILED DESCRIPTION

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0013] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0014] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0015] The embodiment of the present application provides a wind power prediction system integrating integrated learning and time series analysis, such as Figure 1 As shown, the system comprises:

[0016] The characteristic variable matrix mining module 10 is used to mine the characteristic variable matrix based on wind power, wherein the characteristic variable matrix has a weight identifier. Specifically, in the wind power prediction system integrating integrated learning and time series analysis, the characteristic variable matrix mining module 10 mines the characteristic variable matrix based on wind power, which is constructed by considering various factors affecting wind power such as wind speed, wind direction, meteorological conditions and equipment parameters, and has a weight identifier. The weight determination unit traverses the matrix, determines the first weight based on the autocorrelation of the variable, and determines the second weight based on the mutual correlation. The mean calculation unit calculates the mean of the first weight and the second weight to obtain a weight distribution, which corresponds to the characteristic variable matrix one by one, reflecting the importance of each characteristic variable in the wind power prediction. The subsequent system modules will pay more attention to the characteristic variables with larger weights to improve the prediction accuracy.

[0017] In a possible implementation, the characteristic variable matrix mining module 10 further includes: a weight determination unit, the weight determination unit is used to traverse the characteristic variable matrix, determine the first weight based on the variable autocorrelation, and determine the second weight based on the mutual correlation. Specifically, the weight determination unit first starts to traverse the characteristic variable matrix, which is mined based on wind power and contains various characteristic variables that may affect wind power. During the traversal process, the first weight is determined based on the variable autocorrelation. The variable autocorrelation measures the degree of similarity of a variable at different time points. If a characteristic variable has a high autocorrelation over time, it means that the change of this variable has a certain regularity. For example, if the wind speed changes relatively stably over a continuous time period, then the autocorrelation of the characteristic variable of wind speed is high. According to the degree of autocorrelation, a suitable first weight is determined. The higher the correlation, the larger the first weight may be, indicating that this characteristic variable is relatively important in wind power prediction. The second weight is determined based on the mutual correlation. The mutual correlation measures the degree of association between different characteristic variables. If there is a strong positive or negative correlation between two characteristic variables, it means that their mutual correlation is high. For example, there may be a certain correlation between wind speed and wind turbine speed. When the wind speed increases, the wind turbine speed may also increase accordingly. The second weight is determined according to the degree of mutual correlation. The stronger the mutual correlation, the larger the second weight may be, which means that the two characteristic variables play a more important role in jointly affecting wind power.

[0018] A mean calculation unit, the mean calculation unit is used to perform mean calculation on the first weight and the second weight as a weight distribution, wherein the weight distribution corresponds to the characteristic variable matrix one by one. Specifically, the mean calculation unit receives the first weight and the second weight determined by the weight determination unit, and then performs mean calculation on the first weight and the second weight. The result of the mean calculation will be used as the weight distribution. The weight distribution reflects the comprehensive importance of each characteristic variable in the wind power prediction. By taking the average of the first weight (determined based on the autocorrelation of the variable) and the second weight (determined based on the mutual correlation), the change law of a characteristic variable itself and its correlation with other characteristic variables can be comprehensively considered, so as to obtain a more comprehensive weight evaluation. The weight distribution corresponds to the characteristic variable matrix one by one, which means that for each characteristic variable in the characteristic variable matrix, there is a specific weight value corresponding to it. In the subsequent wind power prediction process, the weight value will be used to adjust the influence of different characteristic variables to improve the accuracy of the prediction.

[0019] The timing curve construction module 20 is used to mine the time-phase linear law for the characteristic variable matrix and construct a multivariate timing curve. Specifically, in the wind power prediction system integrating integrated learning and timing analysis, the timing curve construction module 20 works on the characteristic variable matrix. First, the characteristic variable matrix traversal unit traverses the matrix and calls the historical variable records. Then, the first clustering result determination unit clusters the historical variable records once based on the seasonality. Then, the second clustering result determination unit clusters the first clustering results twice based on the periodicity. Finally, the multivariate coordinate system construction subunit in the multivariate timing curve construction unit constructs a multivariate coordinate system based on the characteristic variable matrix, and the characteristic linear trend mining subunit mines the characteristic linear trend. The first multivariate timing curve construction subunit constructs a multivariate timing curve based on the multivariate coordinate system and the characteristic linear trend, providing a basis for the subsequent power prediction model construction.

[0020] In one possible implementation, Figure 2 As shown, the timing curve construction module 20 further includes: a characteristic variable matrix traversal unit, which is used to traverse the characteristic variable matrix, retrieve and call historical variable records. Specifically, the characteristic variable matrix traversal unit aims to traverse the entire characteristic variable matrix. The characteristic variable matrix is ​​mined based on wind power and contains multiple characteristic variables related to wind power, such as meteorological factors such as wind speed, wind direction, temperature, and equipment parameters such as wind turbine speed. During the traversal process, the unit will check each element in the matrix one by one, retrieve and call historical variable records, and the historical variable records contain the values ​​of each characteristic variable in the past period of time, providing a data basis for subsequent analysis and clustering.

[0021] The first clustering result determination unit is used to traverse the historical variable records, perform clustering based on seasonality, and determine the first clustering result. Specifically, the first clustering result determination unit then processes the historical variable records retrieved by the characteristic variable matrix traversal unit. The first clustering result determination unit will traverse these historical variable records and perform clustering based on seasonality. Seasonality means that with the change of seasons, the values ​​of characteristic variables will show a certain regularity. For example, in different seasons, meteorological conditions such as wind speed and temperature will be different, and the changes may have an impact on wind power. By analyzing the data in the historical variable records, data points with similar characteristics are classified into one category according to seasonality to determine the first clustering result. The first clustering result preliminarily groups the data according to seasonal characteristics, providing a basis for further analysis.

[0022] The second clustering result determination unit is used to traverse the first clustering result, perform secondary clustering based on periodicity, and determine the second clustering result. Specifically, the second clustering result determination unit processes the first clustering result again. The second clustering result determination unit will traverse the first clustering result and perform secondary clustering based on periodicity. Periodicity means that similar value patterns will be repeated in certain characteristic variables within a certain time period. For example, certain meteorological conditions may show certain regularities within a daily, weekly or monthly time period. By analyzing the data in the first clustering result, the data is further subdivided according to the periodicity to determine the second clustering result. The second clustering result more finely groups the data according to the characteristics of the time period, providing more accurate information for constructing a multivariate time series curve.

[0023] A multivariate time-series curve construction unit, the multivariate time-series curve construction unit is used to construct the multivariate time-series curve based on the first clustering result and the second clustering result. Specifically, the multivariate time-series curve construction unit constructs the multivariate time-series curve based on the first clustering result and the second clustering result. First, the multivariate time-series curve construction unit will comprehensively consider the information contained in the first clustering result and the second clustering result, determine the basic data for constructing the multivariate time-series curve, and use the data to construct the curve. By constructing the multivariate time-series curve, the changing trend of each characteristic variable over time and the relationship between them can be more intuitively displayed. The multivariate time-series curve will provide an important basis for the subsequent construction of the power prediction model and help improve the accuracy of wind power prediction.

[0024] In a possible implementation, the multivariate time series curve construction unit further includes: a multivariate coordinate system construction subunit, the multivariate coordinate system construction subunit is used to construct a multivariate coordinate system based on the characteristic variable matrix, wherein the multivariate coordinate system takes the characteristic variable and time as the axis. Specifically, the multivariate coordinate system construction subunit first operates based on the characteristic variable matrix, the characteristic variable matrix is ​​formed by mining various characteristic variables based on wind power, and contains multiple characteristic variables related to wind power, such as meteorological factors such as wind speed, wind direction, temperature, and equipment parameters such as the speed of the wind turbine. The multivariate coordinate system construction subunit constructs a multivariate coordinate system. In the multivariate coordinate system, the characteristic variable and time are used as the axis, each characteristic variable is used as a dimension, and time is used as another dimension. Through construction, the values ​​of each characteristic variable at different time points can be represented in this multivariate coordinate system, which provides an intuitive visualization basis for the subsequent analysis of the changing trend of the characteristic variable over time.

[0025] A characteristic linear trend mining subunit is used to identify a first clustering cluster and mine characteristic linear trends, wherein the first clustering cluster is any one of the first clustering result and the second clustering result, and the characteristic linear trend corresponds to the characteristic variable matrix one by one. Specifically, the characteristic linear trend mining subunit first identifies the first clustering cluster, which can be any one of the first clustering result and the second clustering result. The first clustering result is obtained by clustering once based on seasonality, and the second clustering result is obtained by clustering twice based on periodicity. The characteristic linear trend mining subunit mines characteristic linear trends. For the data in the identified first clustering cluster, the linear trend of each characteristic variable changing over time is found through data analysis methods. The characteristic linear trend corresponds to the characteristic variable matrix one by one, which means that for each characteristic variable, its linear change trend in a specific clustering cluster can be determined. Mining linear trends can help us better understand the relationship between characteristic variables and time, as well as the mutual influence between different characteristic variables.

[0026] The first multivariate time series curve construction subunit is used to construct a first multivariate time series curve based on the multivariate coordinate system and the characteristic linear trend. Specifically, when constructing the first multivariate time series curve, the first multivariate time series curve construction subunit first constructs the multivariate coordinate system based on the multivariate coordinate system construction subunit. The coordinate system provides a spatial framework for curve construction, in which characteristic variables and time are used as coordinate axes, so that the values ​​of each characteristic variable at different time points can be located in this coordinate system. Combined with the characteristic linear trend mined by the characteristic linear trend mining subunit, the first multivariate time series curve is constructed in the multivariate coordinate system using linear trend information. The curve can intuitively show the changes of each characteristic variable over time, as well as the relationship between them. By constructing the first multivariate time series curve, it can provide an important reference basis for subsequent wind power forecasting and help improve the accuracy of the forecast.

[0027] The power prediction model construction module 30 is used to combine the multivariate time series curve, perform data-driven training, and construct a power prediction model, wherein the power prediction model integrates multiple base predictors. Specifically, in the wind power prediction system that integrates integrated learning and time series analysis, the power prediction model construction module 30 works in combination with the multivariate time series curve. First, the sample data determination unit calls the historical variable record to determine the sample data, combines the multivariate time series curve data with the sample data during the training process, and uses a data-driven method to allow the model to automatically learn the relationship between the characteristic variables and wind power. The model integrates multiple base predictors, and gradually trains multiple base predictors through the first sample part extraction subunit, the supervised training subunit, and the sample data traversal subunit. Finally, the first base predictor integration subunit integrates these base predictors, and the output mean of N base predictors is used as the output of the power prediction model to cope with the uncertainty and complexity in wind power prediction and provide support for accurate prediction of wind power.

[0028] In a possible implementation, the power prediction model construction module 30 further includes: a sample data determination unit, the sample data determination unit is used to call the historical variable record, integrate and determine the sample data, wherein the sample data includes the mapped characteristic variable sample-power prediction sample. Specifically, the sample data determination unit aims to call the historical variable record, which contains various data related to wind power in the past period of time, including the value of the characteristic variable and the corresponding actual power value, etc. The sample data determination unit integrates the called historical variable record, and in the integration process, screens out the data related to the characteristic variable and power prediction, and organizes it into a specific format, and finally determines the sample data, wherein the sample data includes the mapped characteristic variable sample and the power prediction sample, the characteristic variable sample refers to the value of various characteristic variables extracted from the historical data that may affect the wind power, such as wind speed, wind direction, temperature, etc., and the power prediction sample is the corresponding actual power value. In this way, basic data is provided for the generation of the subsequent power prediction model.

[0029] A power prediction model generation unit, the power prediction model generation unit is used to generate the power prediction model based on the sample data and data-driven training until convergence. Specifically, the power prediction model generation unit starts to build a power prediction model based on the sample data, takes the sample data as input, and adopts a data-driven training method, so that the model can automatically learn the relationship between the characteristic variables and the power prediction from the sample data, without the need to manually design a complex mathematical model, and train until convergence. During the training process, the parameters of the model are continuously adjusted so that the output of the model gradually approaches the actual power prediction sample. Through continuous iteration and optimization, until the performance of the model reaches a certain standard or convergence condition, such as the prediction error of the model is within an acceptable range, a power prediction model is generated, and the model can accurately predict the value of wind power based on the input characteristic variable sample. Through data-driven training based on sample data, the generated power prediction model can adapt to different wind power scenarios and data distributions, and improve the accuracy and reliability of wind power prediction.

[0030] In a possible implementation, the power prediction model generation unit further includes: a first sample part extraction subunit, the first sample part extraction subunit is used to divide the sample data and extract the first sample part. Specifically, the first sample part extraction subunit first operates on the sample data, the sample data is a data set containing the mapped feature variable sample-power prediction sample, the data is integrated and determined from the historical variable records, and is used to build the power prediction model. The first sample part extraction subunit divides the sample data and extracts a part from the complete sample data as the first sample part. The first sample part can be randomly selected according to a certain proportion, or selected according to a specific rule, such as selecting the first several records in the data set, etc. The extracted first sample part will be used for supervised training of the first base predictor.

[0031] A supervised training subunit, the supervised training subunit is used to supervise the training of the first base predictor based on the first sample part. Specifically, the supervised training subunit adopts a supervised training method to train the first base predictor. During the supervised training process, the characteristic variable samples in the first sample part are used as input, and the corresponding power prediction samples are used as the output target value. The first base predictor adjusts its own parameters by learning the mapping relationship between the input and the target value. For example, if the first base predictor is a neural network model, it will continuously adjust the connection weights and biases between neurons according to the input characteristic variables (such as wind speed, wind direction, etc.) and the corresponding actual power values, so that the output of the model is as close to the target power value as possible. Through supervised training, the first base predictor learns a preliminary prediction relationship between the characteristic variables and the power.

[0032] The sample data traversal subunit is used to traverse the sample data and supervise the training of the Nth basis predictor. Specifically, the sample data traversal subunit traverses the entire sample data, which is similar to the supervised training of the first basis predictor, but operates on the entire sample data. During the traversal process, each data point or data subset in the sample data is taken as input in turn to train the Nth basis predictor. By comprehensively traversing the sample data, the Nth basis predictor can learn various patterns and rules in the sample data, and different basis predictors can be exposed to different data distributions during the training process, thereby improving their generalization ability and diversity.

[0033] The first base predictor integration subunit is used to integrate the first base predictor up to the Nth base predictor to generate the power prediction model, wherein the output mean of the N base predictors is used as the output of the power prediction model. Specifically, the first base predictor integration subunit is used to integrate multiple base predictors to generate a final power prediction model, firstly integrate the first base predictor, and then gradually include other base predictors (up to the Nth base predictor) into the integration range. During the integration process, the output mean of the N base predictors is used as the output of the power prediction model. When power prediction is required, the input feature variables are provided to the N base predictors at the same time, each base predictor will give a prediction result, and the N prediction results are averaged. The average value obtained is the final power prediction value. Through the integration method, the advantages of each base predictor are fully utilized, the error and uncertainty of a single base predictor are reduced, and the overall performance and stability of the power prediction model are improved.

[0034] The power prediction data determination module 40 is used to monitor and transmit back multiple characteristic data based on the characteristic variable matrix, and to perform prediction analysis based on the prediction task type in combination with the power prediction model to determine the power prediction data. Specifically, in the wind power prediction system integrating integrated learning and time series analysis, the power prediction data determination module 40 monitors and transmits back multiple characteristic data such as meteorological data and wind turbine operating status data from multiple channels based on the characteristic variable matrix, and then performs prediction analysis based on the prediction task type including short-term prediction and long-term prediction in combination with the power prediction model. For short-term prediction, the real-time relationship between recent characteristic data changes and power is emphasized, and for long-term prediction, the impact of factors such as meteorological trends and seasonal changes on power is considered. The transmitted characteristic data is input into the model, and the power prediction data is obtained through model calculation, which provides a decision-making basis for power grid dispatching, power trading and wind farm operation management.

[0035] In a possible implementation, the power prediction data determination module 40 further includes: a prediction task type component unit, the prediction task type component unit is used to predict that the prediction task type includes short-term prediction and long-term prediction. Specifically, in the wind power prediction system integrating integrated learning and time series analysis, the prediction task type component unit is responsible for determining that the prediction task type includes short-term prediction and long-term prediction. For short-term prediction, the system first collects real-time data updated at a high frequency, such as meteorological data and wind turbine operating status data, performs feature analysis and processing, pays attention to the trend and fluctuation of feature changes in a short period of time, and inputs the data into the power prediction model after preprocessing. The model uses a method suitable for short-term fast calculation to obtain the wind power prediction value for the next few minutes to hours. Long-term prediction requires the integration of historical data and macro data over a longer time range, in-depth analysis and mining of long-term trend laws, extraction of relevant features to establish complex models, such as combining time series analysis and machine learning algorithms to predict wind power in the next few days to weeks or even months, and provide decision support for power grid dispatching, power trading and wind farm equipment maintenance. By clarifying different task types and corresponding processing methods, the power system operation and management needs are met.

[0036] In a possible implementation, the power prediction data determination module 40 further includes: a timestamp constraint synchronization unit, which is used to synchronize the timestamp constraints of the multiple characteristic data. Specifically, in the wind power prediction system integrating integrated learning and time series analysis, the timestamp constraint synchronization unit first records accurate timestamps for each characteristic data (including meteorological data, wind turbine operation data, etc.) during data collection. Then, these timestamped data are received and analyzed, the difference in timestamps of different data is checked, and a reasonable threshold is set. If the difference exceeds the threshold, it is determined that synchronization processing is required. Then, for data with timestamp differences exceeding the threshold, synchronization adjustment is performed by interpolation or extrapolation algorithms (such as estimating the missing time point value based on adjacent time point data) or time delay compensation. Finally, the data after synchronization processing is consistent in time, and can be more accurately used for subsequent power prediction model construction and other module work, ensuring that different characteristic variable data are matched at the same time point, improving model accuracy and system reliability, and providing strong guarantees for wind power prediction and system operation.

[0037] A visualization display module 50 is used to visualize the power forecast data on the terminal and to provide reverse guidance for unit management. Specifically, in a wind power forecasting system that integrates integrated learning and time series analysis, the visualization display module 50 first receives the power forecast data from the power forecast data determination module and performs preliminary processing, such as sorting in chronological order to make it meet the visualization requirements. Then, a variety of visualization methods are designed according to different data and user needs, including a time series graph with time as the horizontal axis and power forecast value as the vertical axis to show the trend of power changes over time, a bar chart or pie chart to show the contribution or proportion of different factors to power forecasting, and when applicable to distributed wind farms, a map is used to mark the location of wind turbines and power forecasts. At the same time, an interactive interface is designed to facilitate users to operate and view data details at different time points. Then, based on the power forecast data displayed visually, reverse guidance is provided for unit management, such as adjusting the power generation plan according to the power forecast value, increasing the number of wind turbines or improving efficiency when the forecast value is high, and reducing the number of wind turbines when the forecast value is low; judging the operating status of wind turbines by observing the data change trend and comparing historical data to schedule equipment maintenance; helping to optimize the resource allocation of wind farms, analyzing and adjusting areas with low power forecasts; and providing support for long-term development and strategic decisions, such as deciding whether to expand, upgrade equipment or improve technology. Through these processes, the visualization display module not only improves the readability and usability of the forecast results, but also strongly supports the efficient operation and scientific management of wind farms.

[0038] The embodiment of the present application adopts the method of extracting the characteristic variable matrix of wind power, first determining the weight identifier, and then mining the phase linear law based on the characteristic variable matrix to construct a multivariate time series curve. On this basis, combined with the multivariate time series curve, a data-driven training method is adopted to construct a power prediction model integrating multiple base predictors. Then, based on the model, multi-channel characteristic data is monitored and transmitted back, the prediction task type is analyzed, and the final power prediction data is determined. Finally, the power prediction results are presented to the end user through a visual display. At the same time, these data are used to reversely guide the operation and management of the wind turbine, realize efficient prediction and real-time optimization, and achieve the technical effect of improving the accuracy and timeliness of wind power prediction.

[0039] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0040] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A wind power prediction system integrating ensemble learning and time series analysis, characterized in that: The system comprises: A characteristic variable matrix mining module, wherein the characteristic variable matrix mining module is used to mine a characteristic variable matrix based on wind power, wherein the characteristic variable matrix has a weight identifier; A time series curve construction module, wherein the time series curve construction module is used to mine the time phase linear law for the characteristic variable matrix and construct a multivariate time series curve; A power prediction model construction module, wherein the power prediction model construction module is used to perform data-driven training in combination with the multivariate timing curve to construct a power prediction model, wherein the power prediction model integrates a plurality of base predictors; A power prediction data determination module, the power prediction data determination module is used to monitor and return multiple characteristic data based on the characteristic variable matrix, and perform prediction analysis based on the prediction task type in combination with the power prediction model to determine the power prediction data; A visualization display module is used to perform terminal visualization of the power prediction data and reversely guide unit management.

2. The wind power prediction system integrating integrated learning and time series analysis according to claim 1, characterized in that: The characteristic variable matrix mining module includes: A weight determination unit, the weight determination unit is used to traverse the characteristic variable matrix, determine a first weight based on the variable autocorrelation, and determine a second weight based on the mutual correlation; A mean calculation unit is used to calculate the mean of the first weight and the second weight as a weight distribution, wherein the weight distribution corresponds one-to-one to the feature variable matrix.

3. The wind power prediction system integrating integrated learning and time series analysis according to claim 1, characterized in that: The timing curve construction module includes: A characteristic variable matrix traversal unit, the characteristic variable matrix traversal unit is used to traverse the characteristic variable matrix, retrieve and call historical variable records; A first clustering result determination unit, the first clustering result determination unit is used to traverse the historical variable records, perform a clustering based on seasonality, and determine a first clustering result; A second clustering result determining unit, the second clustering result determining unit is used to traverse the first clustering result, perform secondary clustering based on periodicity, and determine a second clustering result; A multivariate time series curve construction unit is used to construct the multivariate time series curve based on the first clustering result and the second clustering result.

4. The wind power prediction system integrating integrated learning and time series analysis as claimed in claim 3, characterized in that: The multivariate time series curve construction unit comprises: A multivariate coordinate system construction subunit, the multivariate coordinate system construction subunit is used to construct a multivariate coordinate system based on the characteristic variable matrix, wherein the multivariate coordinate system takes the characteristic variable and time as axes; A characteristic linear trend mining subunit, the characteristic linear trend mining subunit is used to identify a first clustering cluster and mine a characteristic linear trend, wherein the first clustering cluster is any one of the first clustering result and the second clustering result, and the characteristic linear trend corresponds to the characteristic variable matrix one by one; The first multivariate time series curve construction subunit is used to construct a first multivariate time series curve based on the multivariate coordinate system and the characteristic linear trend.

5. The wind power prediction system integrating integrated learning and time series analysis as claimed in claim 3, characterized in that: The power prediction model building module includes: A sample data determination unit, the sample data determination unit is used to call the historical variable record, integrate and determine sample data, wherein the sample data includes a mapped characteristic variable sample-power prediction sample; A power prediction model generation unit is used to generate the power prediction model based on the sample data and data-driven training until convergence.

6. The wind power prediction system integrating integrated learning and time series analysis as claimed in claim 5, characterized in that: The power prediction model generating unit comprises: A first sample part extraction subunit, the first sample part extraction subunit is used to divide the sample data and extract a first sample part; A supervised training subunit, the supervised training subunit is used for supervised training of a first base predictor based on the first sample portion; A sample data traversal subunit, the sample data traversal subunit is used to traverse the sample data and supervise the training of the Nth basis predictor; The first base predictor integration subunit is used for integrating the first base predictor to the Nth base predictor to generate the power prediction model, wherein the output mean of the N base predictors is taken as the output of the power prediction model.

7. The wind power prediction system integrating integrated learning and time series analysis according to claim 1, characterized in that: A prediction task type composition unit is used for predicting that the task type includes short-term prediction and long-term prediction.

8. The wind power prediction system integrating integrated learning and time series analysis as claimed in claim 1, characterized in that: A timestamp constraint synchronization unit is used to synchronize timestamp constraints on the multiple channels of feature data.

Images