Wind turbine power recursive prediction method, device, computer equipment and storage medium
Through the segmented linear regression model, the missing value and BiLSTM model prediction are filled, combined with single-layer and double-layer BiLSTM models, the problem of high data scarcity and stability requirements of fan power prediction is solved, and the accuracy and interpretability prediction of fan power is achieved .
Patent Information
- Application Number
- CN202311565839.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-22
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-11-22
AI Technical Summary
The existing fan power prediction methods have poor prediction accuracy and lack interpretability in the case of high data stability requirements and scarce data. Physical methods and artificial intelligence methods rely on a large amount of data and computing resources, and statistical methods ignore the laws and physical characteristics of wind turbines.
The segmented linear regression model is used to fill the missing value data in the fan power time series, and the initial fan power recursive prediction model is constructed. The overall characteristics and laws are captured through the BiLSTM model, and the prediction is combined with the single-layer and double-layer BiLSTM models to improve the prediction results. accuracy.
It improves the accuracy and interpretability of fan power prediction, can accurately predict under a small amount of data, and improves the diversity of data and the reliability of prediction models.
Smart Images

Figure CN117578428B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind turbine power prediction, and in particular to a wind turbine power recursive prediction method, device, computer equipment and storage medium. Background Art
[0002] Wind energy is a clean energy, but wind power generation faces some challenges, such as unstable wind speed, low wind turbine efficiency, difficult grid dispatching and poor safety. In order to overcome the above problems, the accuracy of wind turbine power prediction is crucial. The existing wind turbine power prediction methods are divided into three types: physical, statistical and artificial intelligence.
[0003] However, the above three prediction methods all have certain defects. The physical method and artificial intelligence method require a large amount of training data and computing resources, and the statistical method ignores the laws and physical characteristics of wind turbines. In addition, in the case of a small amount of available data, the prediction results of the above three prediction methods lack interpretability. Therefore, the existing wind turbine power prediction methods have high requirements for data stability and are difficult to collect data, resulting in poor accuracy in wind turbine power prediction. Summary of the invention
[0004] Based on this, it is necessary to provide a wind turbine power recursive prediction method, device, computer equipment, computer-readable storage medium and computer program product that can improve prediction accuracy and interpretability in response to the above technical problems.
[0005] In a first aspect, the present application provides a wind turbine power recursive prediction method, comprising:
[0006] Construct a wind turbine power time series based on wind turbine power data in historical periods;
[0007] Obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture local features and regularities in the target wind turbine power time series;
[0008] Training an initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0009] According to the wind turbine power recursive prediction, the wind turbine power of the wind farm within the test time is predicted
[0010] In one embodiment, the obtaining missing value data in the wind turbine power time series includes:
[0011] The wind turbine power data in the wind turbine power time series that meets the missing value preset condition is determined as the missing value data; the missing value preset condition includes that the value of the wind turbine power data is zero, and the change value of the wind turbine power data relative to the fan power data of the previous time series is greater than a preset threshold.
[0012] In one embodiment, before filling the missing data by the piecewise linear regression model, the method further includes:
[0013] Interpolating and filling the missing data to obtain an intermediate wind turbine power time series;
[0014] An initial piecewise linear regression model is trained according to the intermediate wind turbine power time series to optimize the parameters of the initial piecewise linear regression model until the trained piecewise linear regression model is obtained.
[0015] In one embodiment, the training of the initial piecewise linear regression model according to the intermediate wind turbine power time series includes:
[0016] Obtaining exponentially weighted moving average characteristics and moving median characteristics of the intermediate wind turbine power time series;
[0017] Processing the intermediate wind turbine power time series according to the BiLSTM model to obtain meta-features of the intermediate wind turbine power time series;
[0018] The initial piecewise linear regression model is trained according to the meta-feature, the exponentially weighted moving average feature and the moving median feature to optimize the parameters of the initial piecewise linear regression model until the trained piecewise linear regression model is obtained.
[0019] In one embodiment, the training of the initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained includes:
[0020] Perform preprocessing to obtain a moving average sequence;
[0021] The moving average sequence is divided into a training set and a validation set, and the initial wind turbine power recursive prediction is trained to obtain the wind turbine power recursive prediction model; including a single-layer BiLSTM model for processing lightweight wind turbine power prediction tasks, and a double-layer BiLSTM model for processing heavyweight wind turbine power prediction tasks.
[0022] In one embodiment, predicting the wind turbine power of the wind farm within the test time according to the recursive prediction of the wind turbine power includes:
[0023] Obtain a first prediction result and a second prediction result predicted by the single-layer BiLSTM model and the double-layer BiLSTM model respectively;
[0024] The first prediction result and the second prediction result are integrated to obtain the wind turbine power of the wind farm within the test time.
[0025] In a second aspect, the present application further provides a wind turbine power recursive prediction device, the device comprising:
[0026] A sequence construction module is used to construct a wind turbine power time series based on wind turbine power data in historical periods;
[0027] A sequence updating module, used for obtaining missing data in the wind turbine power time series, and filling the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used for capturing local features and regularities in the target wind turbine power time series;
[0028] A model training module, used for recursively predicting the wind turbine power according to the target wind turbine power time series to optimize the parameters of the wind turbine power recursive prediction until a trained wind turbine power recursive prediction is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0029] The data prediction module is used to predict the wind turbine power of the wind farm within the test time according to the recursive prediction of the wind turbine power.
[0030] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0031] Construct a wind turbine power time series based on wind turbine power data in historical periods;
[0032] Obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture local features and regularities in the target wind turbine power time series;
[0033] Training an initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0034] The wind turbine power of the wind farm within the test time is predicted according to the recursive prediction of the wind turbine power.
[0035] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:
[0036] Construct a wind turbine power time series based on wind turbine power data in historical periods;
[0037] Obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture local features and regularities in the target wind turbine power time series;
[0038] Training an initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0039] The wind turbine power of the wind farm within the test time is predicted according to the recursive prediction of the wind turbine power.
[0040] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:
[0041] Construct a wind turbine power time series based on wind turbine power data in historical periods;
[0042] Obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture local features and regularities in the target wind turbine power time series;
[0043] Training an initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0044] The wind turbine power of the wind farm within the test time is predicted according to the recursive prediction of the wind turbine power.
[0045] The above-mentioned wind turbine power recursive prediction method, device, computer equipment, storage medium and computer program product construct a wind turbine power time series based on wind turbine power data of historical time periods; obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture the local characteristics and laws in the target wind turbine power time series, so that the missing data can be accurately filled according to the laws of the wind turbine power historical data, thereby improving the diversity of data; the initial wind turbine power recursive prediction model is trained according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series; according to the wind turbine power recursive prediction, the wind turbine power of the wind farm within the test time is predicted, so that according to the captured overall characteristics and laws, the wind turbine power can be accurately predicted based on a small amount of data through the wind turbine power recursive prediction model, thereby improving the accuracy and interpretability of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1 A diagram of an application environment of a wind turbine power recursive prediction method in one embodiment;
[0048] Figure 2 1 is a flow chart of a recursive prediction method for wind turbine power in one embodiment;
[0049] Figure 3 A schematic diagram of a flow chart of steps for generating a page data packet in one embodiment;
[0050] Figure 4 Another schematic diagram of a flow chart of a wind turbine power recursive prediction method in one embodiment;
[0051] Figure 5 is a schematic diagram of a search result of missing value data of wind farm A in one embodiment;
[0052] Figure 6 is a schematic diagram of a recursive prediction result of wind turbine power of wind farm A in one embodiment;
[0053] Figure 7 is a schematic diagram of a search result of missing value data of wind farm B in one embodiment;
[0054] Figure 8 is a schematic diagram of a recursive prediction result of wind turbine power of wind farm B in one embodiment;
[0055] Fig. 9 is a structural block diagram of a wind turbine power recursive prediction device in one embodiment;
[0056] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0058] In traditional technology, wind turbine power prediction mainly adopts three major methods: physics, statistics and artificial intelligence. These prediction methods each have some defects. Physical methods and artificial intelligence methods rely on a large amount of training data and high computing resources, while statistical methods tend to ignore the laws and physical characteristics of wind turbines. In addition, when data is scarce, existing methods often find it difficult to provide highly interpretable prediction results. Therefore, the current wind turbine power prediction method has high requirements for data stability, and data acquisition also faces certain difficulties, which leads to insufficient accuracy in wind turbine power prediction. The wind turbine power recursive prediction method of the present application can effectively solve the above problems caused by traditional technology.
[0059] The wind turbine power recursive prediction method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.
[0060] In the application scenario of the present application, the terminal 102 constructs a wind turbine power time series based on the wind turbine power data of the historical period; obtains the missing data in the wind turbine power time series, and fills the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture the local characteristics and laws in the target wind turbine power time series; the initial wind turbine power recursive prediction model is trained according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series; based on the wind turbine power recursive prediction, the wind turbine power of the wind farm during the test time is predicted.
[0061] In an exemplary embodiment, Figure 2 As shown in FIG. 1 , a recursive prediction method for wind turbine power is provided. The method is applied to Figure 1 The terminal 102 in FIG. 1 is used as an example to illustrate the method, which includes the following steps S202 to S208, wherein:
[0062] Step S202: construct a wind turbine power time series based on wind turbine power data in a historical period.
[0063] The wind turbine power data is the power generated by the wind turbines in the wind farm, and can also be referred to as wind power output data. In some embodiments, the wind turbine power data is collected at preset time intervals during the operation of the wind farm, for example, the wind turbine power data is collected every 15 minutes to construct a wind turbine power time series.
[0064] In some embodiments, the constructed wind turbine power time series can be expressed as {P1, P2, ..., P n}, P is the power value, n represents the number of elements in the sequence, for example, a wind turbine power time series can be {100, 102, 98, 100, 98, ...}. Since the power of each wind turbine in the wind turbine power time series is collected at a preset time interval, taking the collection every 15 minutes as an example, the wind turbine power time series can be a corresponding time feature sequence {T1, T2, ..., T n}, P n With T n Corresponding to, for example, {2021-01-01 00:00, 2021-01-01 00:15, 2021-01-01 00:30, …}.
[0065] Step S204, obtaining missing data in the wind turbine power time series, and filling the missing data through a piecewise linear regression model to obtain a target wind turbine power time series;
[0066] Among them, missing data refers to data that was not collected at the time of collection. It is understandable that the data may be interfered by external factors during collection, or abnormal situations such as sudden power outages or shutdowns may occur. Therefore, there may be some missing data in the wind turbine power time series, which need to be automatically found and completed. Since the missing data is generally saved as 0, it may be confused with the normal 0 value of wind turbine power data during the actual search process, and it needs to be accurately distinguished.
[0067] In some embodiments, the piecewise linear regression model can be used to capture local features and regularities in the target wind turbine power time series. It can be understood that the piecewise linear regression model is not just a simple linear regression of the entire wind turbine power time series, but can also divide the time series into different segments according to the local characteristics of the data, and perform linear regression in each segment. This processing method makes the model more adaptable to changes and fluctuations in the data, so that the target wind turbine power can be predicted more accurately.
[0068] Through the above method, the piecewise linear regression model is used to fill the missing data and obtain the complete target wind turbine power time series. The data filling process helps to retain the local characteristics in the data, thereby improving the accuracy of filling the missing values and making the final time series more complete and reliable.
[0069] Step S206: training the initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained.
[0070] In some embodiments, the target wind turbine power time series can be processed to extract training data for training the initial wind turbine power recursive prediction model, such as meta-features, weighted moving average features, and moving median features, and a trained wind turbine power recursive prediction model can be obtained through training. The wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series.
[0071] It can be understood that through repeated training and parameter adjustment, the initial wind turbine power recursive prediction model is gradually optimized, the fitting ability of the target data is improved, and the prediction results output by the trained wind turbine power recursive prediction model are more accurate and reliable.
[0072] Step S208: predicting the wind turbine power of the wind farm within the test time according to the recursive prediction of the wind turbine power.
[0073] In some embodiments, multiple data can be used to predict a single data to enhance the prediction effect. Since there is a certain relationship between wind turbine power and time, it is considered to add time features to assist prediction. When selecting a calculation model, a combination of a double-layer BiLSTM and a single-layer BiLSTM model can be used to improve the prediction ability.
[0074] Among them, the BiLSTM model is improved based on the LSTM model. The BiLSTM model introduces bidirectional information flow, which is achieved by using two LSTM units at each time step. Therefore, the hidden state of each time step is composed of information from two directions. This improvement in bidirectional information flow enables BiLS TM to capture the contextual information of the input sequence more comprehensively, thereby performing better in many sequence modeling tasks. The expression of the BiLSTM model can be:
[0075]
[0076] In summary, by using the trained wind turbine power recursive prediction model to predict the wind farm during the test period, the power performance of the wind turbine in the future period can be estimated. The entire prediction process is based on the analysis of historical data and model parameters, aiming to provide a reasonable estimate of the future wind turbine power of the wind farm. This prediction method can be used to guide the operation plan of the power system, improve the stability of the power grid, and reduce dependence on non-renewable energy.
[0077] The above-mentioned wind turbine power recursive prediction method, device, computer equipment, storage medium and computer program product construct a wind turbine power time series based on wind turbine power data of historical time periods; obtain missing data in the wind turbine power time series, and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture the local characteristics and laws in the target wind turbine power time series, so that the missing data can be accurately filled according to the laws of the wind turbine power historical data, thereby improving the diversity of data; the initial wind turbine power recursive prediction model is trained according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series; according to the wind turbine power recursive prediction, the wind turbine power of the wind farm within the test time is predicted, so that according to the captured overall characteristics and laws, the wind turbine power can be accurately predicted based on a small amount of data through the wind turbine power recursive prediction model, thereby improving the accuracy and interpretability of the prediction.
[0078] In an exemplary embodiment, Figure 3 FIG. 2 is a flow chart of the steps of generating a page data packet in one embodiment. Before step S204, the following steps are also included:
[0079] Step S302, interpolating and filling missing data to obtain the intermediate wind turbine power time series.
[0080] See also Figure 4 , Figure 4 FIG. 2 is another flow chart of a recursive prediction method for wind turbine power in one embodiment. Figure 4 As shown, before training the piecewise linear regression model, the missing data can be simply linearly filled to obtain the intermediate wind turbine power time series, so that the training features for training the initial piecewise linear regression model can be extracted from the intermediate wind turbine power time series.
[0081] It can be understood that first obtaining the intermediate wind turbine power time series by interpolation can maintain the continuity and integrity of the wind turbine power time series. Interpolation can fill in the missing power values in the wind turbine power time series, making the data in the entire series more complete, laying the foundation for accurately extracting training features for model training from the intermediate wind turbine power time series. It also helps to ensure the continuity and availability of the wind turbine power time series over the entire time range.
[0082] Step S304: training an initial piecewise linear regression model according to the intermediate wind turbine power time series to optimize the parameters of the initial piecewise linear regression model until a trained piecewise linear regression model is obtained.
[0083] It can be understood that by optimizing the parameters of the initial piecewise linear regression model with the features extracted from the intermediate wind turbine power time series, it can better fit the local characteristics and laws of the intermediate wind turbine power time series. By continuously adjusting and optimizing the parameters of the model, the piecewise linear regression model is more accurate and reliable over the entire time series. The piecewise linear regression model trained in this way will be able to better capture the local characteristics in the wind turbine power time series, providing a more reliable basis for subsequent prediction and analysis.
[0084] In an exemplary embodiment, step S304 includes:
[0085] Obtain the exponentially weighted moving average and moving median characteristics of the intermediate wind turbine power time series;
[0086] The intermediate wind turbine power time series is processed according to the BiLSTM model to obtain the meta-features of the intermediate wind turbine power time series;
[0087] The initial piecewise linear regression model is trained according to the meta-features, the exponentially weighted moving average features and the moving median features to optimize the parameters of the initial piecewise linear regression model until a trained piecewise linear regression model is obtained.
[0088] like Figure 4 As shown, in some embodiments, meta-features such as the overall trend, periodicity, and abnormal pattern of wind turbine power data can be extracted from the intermediate wind turbine power time series through a BiLSTM model. Meta-features are prediction results that can make up for missing values, but there will be local distortion or local oscillation. The weighted moving average feature and the moving median feature are features of the data itself. The former can adapt to changes in data more quickly, is more sensitive to recent changes, and can extract the main trends of the data; the latter can better retain local features in the data and is not affected by extreme values. However, the two cannot directly make up for a large number of missing values. Combining the three features and training a piecewise elastic network regression model can complement each other, more accurately predict data, and make up for data.
[0089] It can be understood that by obtaining the exponentially weighted moving average feature and moving median feature of the intermediate wind turbine power time series, the overall trend and local features in the intermediate wind turbine power time series can be captured. Then, the intermediate wind turbine power time series is processed using the BiLSTM model to obtain meta-features, which can be used to represent higher-level abstract characteristics of the sequence. The obtained meta-features, exponentially weighted moving average features, and moving median features are then combined to train the initial piecewise linear regression model. By continuously optimizing the parameters of the model, it can better fit the various characteristics and laws of the intermediate wind turbine power time series. This process continues until the trained piecewise linear regression model is obtained. Finally, this trained model can more accurately predict the missing values in the wind turbine power time series, providing a more reliable foundation for subsequent data analysis and application.
[0090] In an exemplary embodiment, step S204 includes:
[0091] The wind turbine power data that meets the missing value preset condition in the wind turbine power time series is determined as missing value data; the missing value preset condition includes that the value of the wind turbine power data is zero, and the change value of the wind turbine power data relative to the fan power data of the previous time series is greater than a preset threshold.
[0092] As an example, assume that the preset threshold of the drop span is 10, and the wind turbine power time series is {20, 0, 25, 30, 22, 12, 8, 0, 18}. Since the value of P2 drops from 20 to 0 and the drop span is greater than the preset threshold, P2 is missing data. Since the value of P8 drops from 8 to 0 and the drop span is less than the preset threshold, P8 is not missing data.
[0093] In an exemplary embodiment, step S206 includes:
[0094] Perform preprocessing to obtain a moving average sequence;
[0095] The moving average sequence is divided into a training set and a validation set, and the initial wind turbine power recursive prediction is trained to obtain a wind turbine power recursive prediction model; including a single-layer BiLSTM model for processing lightweight wind turbine power prediction tasks, and a double-layer BiLSTM model for processing heavyweight wind turbine power prediction tasks.
[0096] In some embodiments, there are a lot of fluctuations and noises in the target wind turbine power time series after missing values are supplemented, which has a great impact on subsequent predictions. By calculating the moving average, random noise or fluctuations in the data can be effectively removed, making the data more stable, which is very helpful for identifying trends and patterns, and can highlight long-term trends and suppress short-term fluctuations.
[0097] In some embodiments, the moving average sequence and performance can also be preprocessed and standardized into a sequence with zero mean and unit variance, and then the number of multi-step prediction steps is specified, and time features are prepared for training single-layer BiLSTM models and double-layer BiLSTM models. The adaptability to lightweight and heavyweight wind turbine power prediction tasks can be improved by training the two models.
[0098] In an exemplary embodiment, step S208 includes:
[0099] Obtain the first prediction result and the second prediction result predicted by the single-layer BiLSTM model and the double-layer BiLSTM model respectively;
[0100] The first prediction result and the second prediction result are integrated to obtain the wind turbine power of the wind farm within the test time.
[0101] In some embodiments, when the first prediction results and the second prediction results of the single-layer BiLSTM model and the double-layer BiLSTM model are fused, a weighted method is used to gradually transition from the single-layer BiLSTM results to the double-layer BiLSTM results. This weighted method can be by setting a series of weight values so that more reliance is placed on the results of the single-layer BiLSTM in the initial stage, and then gradually increasing reliance on the results of the double-layer BiLSTM until a comprehensive and balanced prediction result is finally formed. This processing method can comprehensively utilize the advantages of the single-layer BiLSTM model and the double-layer BiLSTM model at different stages to improve the overall prediction performance.
[0102] In the specific implementation process, please refer to Figure 5 and Figure 6As shown in the figure, the wind turbine power data of wind farm A from January to July 2021 is selected, with a data sampling interval of 15 minutes, and the power data for the next three hours is predicted. The threshold for the model to search for missing value points is set to 10, the number of hidden layers of the BiLSTM model is 200, and the Adam adaptive learning rate optimization algorithm is used. The maximum number of iterations is 200, and recursive prediction is performed according to the method of predicting 1 data with 30 data. In terms of combining the results of the single-layer and double-layer BiLSTM models, based on the characteristics of both, the final prediction results will gradually transition from the single-layer BiLSTM results to the double-layer BiLSTM results in a weighted manner. The results of searching for missing values are shown in the figure. Figure 5 As shown in the figure, we can see that the program automatically searches for data anomalies according to the threshold and successfully distinguishes normal zero values. The final model prediction evaluation index values are shown in Table 1. The set prediction interval is predicted based on the trained model, and the results are as follows Figure 6 As shown, from Figure 6 It can be seen that the current model has a certain degree of credibility for short-term predictions in the future and is also relatively accurate in terms of trend predictions.
[0103] MAE RMSE 2.92 3.02
[0104] Table 1
[0105] Among them, the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE) values are very close, indicating that the prediction results are highly accurate.
[0106] like Figure 7 and Figure 8 As shown in Figure 2, the wind turbine power data of wind farm B from January to July 2021 is selected, with a data sampling interval of 15 minutes, and the power data for the next three hours is predicted. The model settings are the same as before, and the calculation results are as follows: Figure 7 , Figure 8 And as shown in Table 2.
[0107] MAE RMSE 0.43 0.55
[0108] Table 2
[0109] from Figure 5 , Figure 6 It can be seen from Table 2 that the program has strong performance in both missing data identification and data recursive prediction. Combining the results of the two examples, it can be concluded that the program can be used to perform short-term wind turbine power recursive prediction when the data resources are small and there are continuous missing values.
[0110] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0111] Based on the same inventive concept, the embodiment of the present application also provides a wind turbine power recursive prediction device for implementing the wind turbine power recursive prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more wind turbine power recursive prediction device embodiments provided below can refer to the limitations of the wind turbine power recursive prediction method above, and will not be repeated here.
[0112] In an exemplary embodiment, Fig. 9 As shown, a wind turbine power recursive prediction device is provided, comprising: a sequence construction module 410, a sequence update module 420, a model training module 430 and a data prediction module 440, wherein:
[0113] A sequence construction module 410 is used to construct a wind turbine power time series based on wind turbine power data in a historical period;
[0114] The sequence updating module 420 is used to obtain missing data in the wind turbine power time series and fill the missing data through a piecewise linear regression model to obtain a target wind turbine power time series; the piecewise linear regression model is used to capture local features and regularities in the target wind turbine power time series;
[0115] The model training module 430 is used to train the initial wind turbine power recursive prediction model according to the target wind turbine power time series to optimize the parameters of the initial wind turbine power recursive prediction until a trained wind turbine power recursive prediction model is obtained; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series;
[0116] The data prediction module 440 is used to predict the wind turbine power of the wind farm within the test time according to the recursive prediction of the wind turbine power.
[0117] In one embodiment, the sequence updating module 420 is further specifically configured to:
[0118] The wind turbine power data that meets the missing value preset condition in the wind turbine power time series is determined as missing value data; the missing value preset condition includes that the value of the wind turbine power data is zero, and the change value of the wind turbine power data relative to the fan power data of the previous time series is greater than a preset threshold.
[0119] In one embodiment, the device further comprises:
[0120] The intermediate sequence generation module is used to interpolate and fill in the missing data to obtain the intermediate wind turbine power time series;
[0121] The piecewise linear regression model generation module is used to train the initial piecewise linear regression model according to the intermediate wind turbine power time series to optimize the parameters of the initial piecewise linear regression model until a trained piecewise linear regression model is obtained.
[0122] In one embodiment, the piecewise linear regression model generation module includes:
[0123] A feature acquisition submodule is used to obtain the exponentially weighted moving average feature and moving median feature of the intermediate wind turbine power time series;
[0124] The meta-feature acquisition submodule is used to process the intermediate wind turbine power time series according to the BiLSTM model to obtain the meta-features of the intermediate wind turbine power time series;
[0125] The initial piecewise linear regression model is trained according to the meta-features, the exponentially weighted moving average features and the moving median features to optimize the parameters of the initial piecewise linear regression model until a trained piecewise linear regression model is obtained.
[0126] In one embodiment, the model training module 430 is further specifically used for:
[0127] Perform preprocessing to obtain a moving average sequence;
[0128] The moving average sequence is divided into a training set and a validation set, and the initial wind turbine power recursive prediction is trained to obtain a wind turbine power recursive prediction model; including a single-layer BiLSTM model for processing lightweight wind turbine power prediction tasks, and a double-layer BiLSTM model for processing heavyweight wind turbine power prediction tasks.
[0129] In one embodiment, the data prediction module 440 is further specifically configured to:
[0130] Obtain the first prediction result and the second prediction result predicted by the single-layer BiLSTM model and the double-layer BiLSTM model respectively;
[0131] The first prediction result and the second prediction result are integrated to obtain the wind turbine power of the wind farm within the test time.
[0132] The wind turbine power recursive prediction method provided in the embodiment of the present application can accurately fill in missing data according to the rules of historical wind turbine power data and improve data diversity. It can also accurately predict wind turbine power based on a small amount of data through a wind turbine power recursive prediction model according to the captured overall characteristics and rules, thereby improving the accuracy and interpretability of the prediction.
[0133] Each module in the above wind turbine power recursive prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.
[0134] Fig.10 In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a recursive prediction method for wind turbine power is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.
[0135] Those skilled in the art will understand that Fig.10The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0136] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0137] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0138] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0139] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0140] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0141] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0142] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A recursive prediction method for wind turbine power, characterized in that: The method comprises: Construct a wind turbine power time series based on wind turbine power data in historical periods; Obtain missing data in the wind turbine power time series, perform interpolation to fill the missing data, and obtain an intermediate wind turbine power time series; obtain the exponentially weighted moving average feature and the moving median feature of the intermediate wind turbine power time series; process the intermediate wind turbine power time series according to the BiLSTM model to obtain the meta-features of the intermediate wind turbine power time series; train an initial piecewise linear regression model according to the meta-features, the exponentially weighted moving average feature and the moving median feature to optimize the parameters of the initial piecewise linear regression model until the trained piecewise linear regression model is obtained, and fill the missing data through the piecewise linear regression model to obtain the target wind turbine power time series; the piecewise linear regression model is used to capture the local features and laws in the target wind turbine power time series; Preprocessing is performed to obtain a moving average sequence; the moving average sequence is divided into a training set and a validation set, and the initial wind turbine power recursive prediction is trained to obtain the wind turbine power recursive prediction model; including a single-layer BiLSTM model for processing lightweight wind turbine power prediction tasks, and a double-layer BiLSTM model for processing heavyweight wind turbine power prediction tasks; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series; The wind turbine power of the wind farm within the test time is predicted according to the recursive prediction of the wind turbine power.
2. The method according to claim 1, characterized in that The obtaining of missing value data in the wind turbine power time series includes: The wind turbine power data in the wind turbine power time series that meets the missing value preset condition is determined as the missing value data; the missing value preset condition includes that the value of the wind turbine power data is zero, and the change value of the wind turbine power data relative to the fan power data of the previous time series is greater than a preset threshold.
3. The method according to claim 1, characterized in that The step of predicting the wind turbine power of the wind farm within the test time according to the wind turbine power recursive prediction includes: Obtain a first prediction result and a second prediction result predicted by the single-layer BiLSTM model and the double-layer BiLSTM model respectively; The first prediction result and the second prediction result are integrated to obtain the wind turbine power of the wind farm within the test time.
4. The method according to claim 1, characterized in that: The missing data refers to data that has not been collected at the time of collection.
5. A wind turbine power recursive prediction device, characterized in that: The device comprises: A sequence construction module is used to construct a wind turbine power time series based on wind turbine power data in historical periods; A sequence updating module is used to obtain missing data in the wind turbine power time series, and an intermediate sequence generating module is used to interpolate and fill the missing data to obtain an intermediate wind turbine power time series; a feature acquiring submodule is used to obtain the exponentially weighted moving average feature and the moving median feature of the intermediate wind turbine power time series; a meta-feature acquiring submodule is used to process the intermediate wind turbine power time series according to the BiLSTM model to obtain the meta-feature of the intermediate wind turbine power time series; a piecewise linear regression model generating module is used to train an initial piecewise linear regression model according to the meta-feature, the exponentially weighted moving average feature and the moving median feature to optimize the parameters of the initial piecewise linear regression model until the trained piecewise linear regression model is obtained, and the missing data is filled by the piecewise linear regression model to obtain the target wind turbine power time series; the piecewise linear regression model is used to capture the local features and laws in the target wind turbine power time series; A model training module is used for preprocessing to obtain a moving average sequence; dividing the moving average sequence into a training set and a validation set, training the initial wind turbine power recursive prediction, and obtaining the wind turbine power recursive prediction model; including a single-layer BiLSTM model for processing lightweight wind turbine power prediction tasks, and a double-layer BiLSTM model for processing heavyweight wind turbine power prediction tasks; the wind turbine power recursive prediction model is used to capture the overall characteristics and laws in the target wind turbine power time series; The data prediction module is used to predict the wind turbine power of the wind farm within the test time according to the recursive prediction of the wind turbine power.
6. The device according to claim 5, characterized in that The sequence updating module is also used for: The wind turbine power data in the wind turbine power time series that meets the missing value preset condition is determined as the missing value data; the missing value preset condition includes that the value of the wind turbine power data is zero, and the change value of the wind turbine power data relative to the fan power data of the previous time series is greater than a preset threshold.
7. The device according to claim 5, characterized in that The data prediction module is also used for: The first prediction result and the second prediction result predicted by the single-layer BiLSTM model and the double-layer BiLSTM model are respectively obtained; the first prediction result and the second prediction result are merged to obtain the wind turbine power of the wind farm within the test time.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.
