A long-term wind power prediction method and device, storage medium, and computer equipment

Abstract

The present application discloses a long-term wind power prediction method and device, storage medium, and computer equipment. The method includes: acquiring historical wind power data, and performing signal decomposition on the historical wind power data to obtain multiple target decomposition signals; determining a target input matrix corresponding to each target decomposition signal based on the target decomposition signal and the historical climate data in the first historical time period; inputting the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and updating the target input matrix corresponding to each target decomposition signal; inputting each updated target input matrix into the preset wind power prediction model to obtain second wind power prediction data, until the wind power prediction data at the last target prediction moment is obtained; and determining the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data.

Description

Technical Field

[0001] The present application relates to the technical field of wind power generation, and in particular to a long-term wind power prediction method and device, a storage medium, and a computer device. Background Art

[0002] With the global shortage of fossil energy and the increasing awareness of environmental protection, the advantages of new energy are becoming increasingly prominent. However, although new energy is safe and environmentally friendly, it also has the disadvantage of poor predictability. Taking wind power generation as an example, due to the natural intermittent and random nature of wind energy, it has poor predictability. In order to ensure the stability of the power grid when using wind power generation, it is very important to predict wind power before wind power generation.

[0003] At present, all wind power prediction methods are ultra-short-term prediction or short-term prediction, which are generally predicted 15 minutes, 30 minutes and 1 hour in advance. Few people study long-term wind power prediction. However, the trading mode of my country's electricity spot market is generally day-ahead trading. If wind power generation is introduced into market transactions, it is necessary to predict the wind power on the trading day in the day before, so as to ensure the wind power output. That is, wind power generation usually requires day-ahead wind power prediction to ensure power stability. Therefore, how to accurately and conveniently perform long-term wind power prediction has become an urgent problem to be solved in this field. Summary of the invention

[0004] In view of this, the present application provides a long-term wind power prediction method and apparatus, a storage medium, and a computer device, which can accurately and conveniently perform long-term wind power prediction.

[0005] According to one aspect of the present application, a long-term wind power prediction method is provided, comprising:

[0006] Acquire historical wind power data in a first historical time period of the same length as the target prediction time period, and perform signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain a plurality of target decomposition signals;

[0007] Determine a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data and historical wind speed data;

[0008] Input the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and update the target input matrix corresponding to each target decomposition signal based on the first wind power prediction data, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model;

[0009] Input each updated target input matrix into the preset wind power prediction model to obtain the second wind power prediction data at the second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, and obtain the wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and the preset prediction time interval;

[0010] Based on the wind power prediction data at each of the target prediction moments corresponding to each of the target decomposition signals, the predicted total wind power corresponding to the target prediction time period is determined.

[0011] Optionally, determining a target input matrix corresponding to each target decomposition signal based on the target decomposition signal and the historical climate data in the first historical time period specifically includes:

[0012] The historical climate data within the first historical time period is obtained, and based on any of the target decomposition signals and the historical climate data, values ​​are taken according to the preset prediction time interval to obtain the target input matrix corresponding to any of the target decomposition signals.

[0013] Optionally, inputting the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and updating the target input matrix corresponding to each target decomposition signal based on the first wind power prediction data specifically includes:

[0014] Inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals;

[0015] Determine target update data of each target decomposition signal at the first target prediction moment based on the first wind power prediction data corresponding to each target decomposition signal and the predicted climate data corresponding to the first target prediction moment;

[0016] Based on the target update data, the data of the historical moment with the longest distance from the current time in the target input matrix corresponding to each target decomposition signal is updated to obtain an updated target input matrix corresponding to each target decomposition signal, wherein the historical moment is determined based on the first historical time period and the preset prediction time interval.

[0017] Optionally, determining the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal specifically includes:

[0018] Determine the first wind power corresponding to any of the target prediction moments based on the wind power prediction data corresponding to each of the target decomposition signals at any of the target prediction moments, and determine the predicted total wind power corresponding to the target prediction time period based on the first wind power corresponding to each of the target prediction moments; or

[0019] Based on the wind power prediction data corresponding to each target prediction moment of any target decomposition signal, the second wind power corresponding to any target decomposition signal is determined, and based on the second wind power corresponding to each target decomposition signal, the predicted total wind power corresponding to the target prediction time period is determined.

[0020] Optionally, inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals specifically includes:

[0021] Inputting the target input matrix corresponding to each of the target decomposition signals into the convolutional neural network model to obtain a target space feature vector corresponding to each of the target decomposition signals;

[0022] The target space feature vector is input into the long short-term memory model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals.

[0023] Optionally, after determining the predicted total wind power corresponding to the target prediction time period, the method further includes:

[0024] When any moment within the target prediction time period is before the current time, the actual total wind power corresponding to the predicted total wind power is obtained, and when the error between the actual total wind power and the predicted total wind power is greater than a preset error threshold, the preset wind power prediction model is trained based on the wind power prediction data corresponding to each target prediction moment within the target prediction time period, the actual wind power data and the actual climate data to obtain an updated preset wind power prediction model.

[0025] Optionally, before inputting the target input matrix into a preset wind power prediction model, the method further includes:

[0026] Acquire sample data in a second historical time period, and divide the sample data into training sample data and update sample data according to time, wherein the sample data includes sample wind power real data and sample climate data;

[0027] Determine a first sample input matrix based on the training sample data, input the first sample input matrix into an initial wind power prediction model to obtain sample wind power prediction data corresponding to a first historical moment, and adjust model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model;

[0028] The training sample data is updated by the updated sample data, a second sample input matrix is ​​determined, the second sample input matrix is ​​input into the updated initial wind power prediction model, the sample wind power prediction data corresponding to the second historical moment is obtained, and based on the sample wind power real data at the second historical moment and the sample wind power prediction data, the model parameters of the updated initial wind power prediction model are adjusted until the model loss value is less than the preset loss threshold, thereby obtaining the preset wind power prediction model.

[0029] According to another aspect of the present application, a long-term wind power prediction device is provided, comprising:

[0030] A signal decomposition module is used to obtain historical wind power data in a first historical time period of the same length as the target prediction time period, and perform signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain multiple target decomposition signals;

[0031] a matrix determination module, configured to determine a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data and historical wind speed data;

[0032] A first input module is used to input the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and based on the first wind power prediction data, update the target input matrix corresponding to each target decomposition signal, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model;

[0033] A second input module is used to input each updated target input matrix into the preset wind power prediction model to obtain second wind power prediction data at a second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, to obtain wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and a preset prediction time interval;

[0034] The wind power determination module is used to determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal.

[0035] Optionally, the matrix determination module is specifically used to:

[0036] The historical climate data within the first historical time period is obtained, and based on any of the target decomposition signals and the historical climate data, values ​​are taken according to the preset prediction time interval to obtain the target input matrix corresponding to any of the target decomposition signals.

[0037] Optionally, the first input module specifically includes:

[0038] A matrix input unit, used for inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals;

[0039] a data determination unit, configured to determine target update data of each target decomposition signal at the first target prediction moment based on the first wind power prediction data corresponding to each target decomposition signal and the predicted climate data corresponding to the first target prediction moment;

[0040] A data updating unit is used to update the data of the historical moment with the longest distance from the current time in the target input matrix corresponding to each of the target decomposition signals according to the target update data, so as to obtain an updated target input matrix corresponding to each of the target decomposition signals, wherein the historical moment is determined based on the first historical time period and the preset prediction time interval.

[0041] Optionally, the wind power determination module is specifically used to:

[0042] Based on the wind power prediction data corresponding to each of the target decomposition signals at any of the target prediction moments, determine the first wind power corresponding to any of the target prediction moments, and based on the first wind power corresponding to each of the target prediction moments, determine the predicted total wind power corresponding to the target prediction time period; or, based on the wind power prediction data corresponding to each of the target prediction moments of any of the target decomposition signals, determine the second wind power corresponding to any of the target decomposition signals, and based on the second wind power corresponding to each of the target decomposition signals, determine the predicted total wind power corresponding to the target prediction time period.

[0043] Optionally, the matrix input unit is specifically used to:

[0044] The target input matrix corresponding to each of the target decomposition signals is input into the convolutional neural network model to obtain the target space feature vector corresponding to each of the target decomposition signals; the target space feature vector is input into the long short-term memory model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals.

[0045] Optionally, the device further comprises:

[0046] A model updating module is used for, after determining the predicted total wind power corresponding to the target prediction time period, obtaining the actual total wind power corresponding to the predicted total wind power when any moment within the target prediction time period is before the current time, and when the error between the actual total wind power and the predicted total wind power is greater than a preset error threshold, training the preset wind power prediction model based on the wind power prediction data corresponding to each of the target prediction moments within the target prediction time period, the actual wind power data and the actual climate data, to obtain an updated preset wind power prediction model.

[0047] Optionally, the device further comprises:

[0048] A data partitioning module, used for obtaining sample data in a second historical time period before inputting the target input matrix into a preset wind power prediction model, and dividing the sample data into training sample data and update sample data according to time, wherein the sample data includes sample wind power real data and sample climate data;

[0049] A model training module is used to determine a first sample input matrix based on the training sample data, input the first sample input matrix into an initial wind power prediction model, obtain sample wind power prediction data corresponding to a first historical moment, and adjust the model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model; update the training sample data through the updated sample data, determine a second sample input matrix, input the second sample input matrix into the updated initial wind power prediction model, obtain sample wind power prediction data corresponding to a second historical moment, and adjust the model parameters of the updated initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the second historical moment until the model loss value is less than a preset loss threshold, thereby obtaining the preset wind power prediction model.

[0050] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned long-term wind power prediction method is implemented.

[0051] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned long-term wind power prediction method when executing the program.

[0052] By means of the above technical scheme, the present application provides a long-term wind power prediction method and device, storage medium, and computer equipment. Before performing wind power prediction, the time period to be predicted can be determined first, that is, the target prediction time period can be determined first. After determining the target prediction time period, the first historical time period of the same length as the target prediction time period can be determined, and then the historical wind power data in the first historical time period can be obtained. After that, the complementary set empirical mode decomposition method can be used to perform signal decomposition on the historical wind power data, and then multiple target decomposition signals can be obtained. In addition, the historical climate data in the first historical time period can be obtained, and the target input matrix corresponding to each target decomposition signal can be determined based on the target decomposition signal and the historical climate data. The historical climate data may include one or more of historical temperature data, historical humidity data, and historical wind speed data. After determining the target input matrix, the target input matrix can be input into the preset wind power prediction model, and the first wind power prediction data corresponding to the first target prediction moment of each target decomposition signal can be obtained. Then, the data in the target input matrix corresponding to each target decomposition signal can be updated based on the first wind power prediction data corresponding to the first target prediction moment. After the target input matrix corresponding to each target decomposition signal is updated, the updated target input matrix can be input into the preset wind power prediction model again, and the second wind power prediction data corresponding to each target decomposition signal at the second target prediction moment is output. Then, the data in the target input matrix is ​​updated using the second wind power prediction data corresponding to the second target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the second target prediction moment, and the target input matrix after the second update is input into the preset wind power prediction model to obtain the third wind power prediction data corresponding to each target decomposition signal at the third target prediction moment... When the number of wind power prediction data of each target decomposition signal is consistent with the number of target prediction moments corresponding to the target prediction time period, it means that the wind power prediction data corresponding to the last target prediction moment in the target prediction time period has been obtained. After obtaining the wind power prediction data corresponding to each target decomposition signal at each target prediction moment, the predicted total wind power in the target prediction time period can be further determined based on these wind power prediction data. The embodiment of the present application can continuously update the target input matrix through wind power prediction data, thereby obtaining the wind power prediction data corresponding to each target prediction moment, and can accurately and conveniently perform long-term wind power prediction.

[0053] 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. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0055] Figure 1 A schematic diagram of a process flow of a long-term wind power prediction method provided by an embodiment of the present application is shown;

[0056] Figure 2 A schematic diagram of a process flow of another long-term wind power prediction method provided by an embodiment of the present application is shown;

[0057] Figure 3 A schematic structural diagram of a long-term wind power prediction device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0058] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0059] In this embodiment, a long-term wind power prediction method is provided. Figure 1 As shown, the method includes:

[0060] Step 101, obtaining historical wind power data in a first historical time period of the same length as a target prediction time period, and performing signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain a plurality of target decomposition signals;

[0061] The long-term wind power prediction method provided by the embodiment of the present application can be used to predict long-term wind power such as day-ahead wind power. Before predicting wind power, you can first determine the time period you want to predict, that is, first determine the target prediction time period. Since the current trading mode of my country's electricity spot market is generally day-ahead trading, the target prediction time period can be one day, or several hours. After determining the target prediction time period, you can first determine the first historical time period of the same length as the target prediction time period. Here, since the climate data is less mutated, the first historical time period can be a time period that is adjacent to the target prediction time period and has the same length as the target prediction time period. For example, if the target prediction time period is 0:00 tomorrow to 24:00 tomorrow, then the first historical time period can be determined as 0:00 today to 24:00 today, and then the historical wind power data in the first historical time period can be obtained. After that, the complementary set empirical mode decomposition method, i.e., CEEMD method, can be used to perform signal decomposition on the historical wind power data, and then multiple target decomposition signals can be obtained. The traditional ensemble empirical mode decomposition (EEMD) method has information loss when restoring the original signal and cannot perfectly restore the original signal, which leads to reduced accuracy of wind power forecast data. The CEEMD method can perfectly solve this problem. The principle of the CEEMD method is to introduce positive and negative white noises as auxiliary noises into the original signal, in order to make the redundant auxiliary white noise signals disappear after decomposition and reduce the number of iterations.

[0062] Step 102, determining a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data, and historical wind speed data;

[0063] In this embodiment, the historical climate data in the first historical time period can also be obtained, and then the target input matrix corresponding to each target decomposition signal can be determined based on the target decomposition signal and the historical climate data. The historical climate data may include one or more of the historical temperature data, the historical humidity data, and the historical wind speed data. The main factor affecting wind power generation is the change in wind speed, and the main factor affecting the wind speed is the factors of meteorology and terrain. However, after the wind power plant is built, the influencing factors of the terrain have been fixed, so it is necessary to evaluate the meteorological data for its impact on wind power. Meteorology and wind speed maintain a nonlinear correlation, and there are regional climate and terrain factors that limit the fluctuations they produce. In addition, the weather data of NWP (Numerical Weather Prediction) is also the forecast of the future day by China's National Meteorological Bureau, which can more accurately reflect the meteorological information of the forecast day, so temperature, humidity, and wind speed can be selected as historical climate data.

[0064] Step 103, inputting the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each of the target decomposition signals, and updating the target input matrix corresponding to each of the target decomposition signals based on the first wind power prediction data, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model;

[0065] In this embodiment, after determining the target input matrix, the target input matrix can be input into the preset wind power prediction model, and the first wind power prediction data of the first target prediction moment corresponding to each target decomposition signal can be obtained. Then, the data in the target input matrix corresponding to each target decomposition signal can be updated with the first wind power prediction data corresponding to the first target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the first target prediction moment. For example, the target prediction time period is tomorrow 0:00 to tomorrow 24:00, the historical wind power data and historical climate data obtained are the data corresponding to today 0:00 to today 24:00, and the preset prediction time interval is 15min. Then the first historical moment corresponding to the first historical time period is today 0:15, and so on. The last historical moment is today 24:00, and the first target prediction moment is tomorrow 0:15. Then the data corresponding to today 0:15 can be replaced by the data corresponding to the first target prediction moment, and finally there are still 96 moments of data left. Here, the preset wind power prediction model can include a trained convolutional neural network model and a long short-term memory model. Convolutional neural networks can automatically extract features between input data. Simple convolutional neural networks include convolutional layers, pooling layers, fully connected layers, and normalization layers. The traditional neural network operation process does not extract the associated features between the input data. For the prediction of wind power, the correlation between the power at this moment and the next moment is relatively weak for traditional load forecasting technology. Therefore, for traditional forecasting technology, only time series data is used for forecasting, and the accuracy of the forecast results is low. However, the correlation between wind power and climate is strong, so it is possible to choose to extract the spatial features between wind power and climate data. Specifically, the decomposed IMF signal can be combined with the wind speed data, temperature data, and humidity data in the climate data to generate 4D data, and the input of the convolutional neural network can be 2*2*1*96. The convolutional layer of the convolutional neural network can be 6 layers, the convolution kernel size can be 2*2*1, the pooling layer selects the maximum pooling layer, and the activation function is relu. The long short-term memory model can be a model built based on the long short-term memory network (LSTM) algorithm. LSTM is a special type of RNN network. It is a recursive neural network that uses back-propagation time training and overcomes the vanishing gradient problem. Different from previous deep neural network methods that can only train each sample, LSTM can model sequence data and "remember" user behavior at different times. The output data of each layer will be used as the input data of the next layer for training. While making full use of the object's sequence attribute information, it can better "remember" long-term object behavior information.Traditional neural networks do not consider what information to use for the data at the next moment, but only process the data at the current moment. However, the LSTM operation process considers the input data at the previous moment and uses it as the input content at this moment. If the correlation at the previous moment is not considered during the wind power prediction process, the accuracy of wind power prediction will be reduced. Therefore, the LSTM network that can extract time features is selected as the model for the prediction stage.

[0066] Step 104, input each updated target input matrix into the preset wind power prediction model, obtain the second wind power prediction data at the second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, obtain the wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and the preset prediction time interval;

[0067] In this embodiment, after the target input matrix corresponding to each target decomposition signal is updated, the updated target input matrix can be input into the preset wind power prediction model again, and the second wind power prediction data corresponding to each target decomposition signal at the second target prediction moment can be output. For example, the first target prediction moment is 0:15, and the preset prediction time interval is 15min, then the second target prediction moment is 0:30. At this time, if the target prediction time period is 30 minutes, the wind power prediction ends, otherwise, the target input matrix update step is repeated, and the data in the target input matrix is ​​updated using the second wind power prediction data corresponding to the second target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the second target prediction moment, and the target input matrix after the second update is input into the preset wind power prediction model to obtain the third wind power prediction data corresponding to each target decomposition signal at the third target prediction moment... When the number of wind power prediction data of each target decomposition signal is consistent with the number of target prediction moments corresponding to the target prediction time period, it means that the wind power prediction data corresponding to the last target prediction moment in the target prediction time period has been obtained. For example, if the target prediction time period is 24 hours and the preset prediction time interval is 15 minutes, the target prediction time period can be divided into 96 target prediction moments. When the number of wind power prediction data of each target decomposition signal is 96, it means that the wind power prediction data of the last target prediction moment has been obtained.

[0068] The above process can be further expressed by the formula: Among them, P t-95 ~P tis the wind power decomposition value corresponding to the target decomposition signal A from the first historical moment to the last historical moment in the first historical time period, P' t+1 According to P t-95 ~P t The wind power prediction data at the first target prediction moment corresponding to the target decomposition signal A is obtained, and then P' t+1 Replace the original P t-95 , according to P t-94 ~P t and P' t+1 Determine the wind power forecast data P' corresponding to the second target forecast time t+2 , and then the wind power prediction data obtained by prediction replaces the wind power decomposition value of the corresponding historical moment in the first historical time period, and is used to predict the wind power prediction data corresponding to the next target prediction moment, until the wind power prediction data corresponding to the last target prediction moment is obtained; α represents the mapping relationship between input data and output data, and the data in brackets are used to predict the wind power at the next target prediction moment. The wind power prediction data and the wind power decomposition value corresponding to the target decomposition signal A in the first historical time period are put into the trained network as input using a roller method, and repeated multiple times until the prediction length is met.

[0069] Step 105: Determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal.

[0070] In this embodiment, after obtaining the wind power prediction data corresponding to each target decomposition signal at each target prediction moment, the predicted total wind power within the target prediction time period can be further determined based on these wind power prediction data. For example, the wind power prediction data corresponding to each target decomposition signal at each target prediction moment within the target prediction time period can be added together to determine the predicted total wind power. For example, the historical wind power data is decomposed into three target decomposition signals, namely signal 1, signal 2, and signal 3, and each target decomposition signal corresponds to four wind power prediction data. Then, the wind power prediction data corresponding to each target decomposition signal at each target prediction moment is these four wind power prediction data, and the total wind power prediction value corresponding to each target decomposition signal is the sum of the four wind power prediction data corresponding to the target decomposition signal. When the predicted total wind power is determined later, the total wind power prediction values ​​corresponding to each target decomposition signal can be added together to obtain the result.

[0071] In an embodiment of the present application, optionally, before step 103, the method further includes: obtaining sample data within a second historical time period, dividing the sample data according to time into training sample data and update sample data, wherein the sample data includes sample wind power real data and sample climate data; determining a first sample input matrix based on the training sample data, inputting the first sample input matrix into an initial wind power prediction model, obtaining sample wind power prediction data corresponding to a first historical moment, and adjusting model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, Update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model; update the training sample data through the updated sample data, determine a second sample input matrix, input the second sample input matrix into the updated initial wind power prediction model, obtain the sample wind power prediction data corresponding to the second historical moment, and adjust the model parameters of the updated initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the second historical moment, until the model loss value is less than the preset loss threshold, and obtain the preset wind power prediction model.

[0072] In this embodiment, before the target input matrix is ​​input into the preset wind power prediction model, the initial wind power prediction model can be trained to obtain the preset wind power prediction model. Here, the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model. Specifically, the second historical time period can be determined first, and then the sample data corresponding to the second historical time period can be obtained. Here, the sample data can also include sample wind power real data and sample climate data, and the sample climate data can include one or more of sample temperature data, sample humidity data and sample wind speed data. When the data specifically included in the sample climate data is determined, when the preset wind power prediction model is used to predict wind power later, the historical climate data should also include the same data. For example, the sample climate data includes sample temperature data, sample humidity data and sample wind speed data, then the historical climate data should also include historical temperature data, historical humidity data and historical wind speed data. When the sample climate data includes sample temperature data and sample wind speed data, then the historical climate data should also include historical temperature data and historical wind speed data. After determining the sample data, the sample data can be divided according to time, and the sample data can be divided into two parts, one part is the training sample data, and the other part is the update sample data, wherein the data generation time corresponding to the training sample data is earlier than the data generation time corresponding to the update sample data, and the training sample data is used to initially input into the initial wind power prediction model, and the update sample data is used to update the training sample data after each wind power prediction. For example, the training sample data includes data at four historical moments. After a wind power prediction is performed, the data at the earliest historical moment in the update sample data can be used to replace the data at the earliest historical moment in the training sample data to form new training sample data. After the sample data is divided to obtain the training sample data, the first sample input matrix can be determined according to the training sample data. The method for determining the first sample input matrix is ​​the same as the method for determining the target input matrix, which will not be repeated here. The first sample input matrix is ​​input into the initial wind power prediction model to obtain the sample wind power prediction data corresponding to the first historical moment, and then the model parameters in the initial wind power prediction model can be adjusted based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and the initial wind power prediction model can be updated. Afterwards, the data at the earliest historical moment in the training sample data can be updated by updating the sample data, determining the second sample input matrix, inputting the second sample input matrix into the updated initial wind power prediction model, and obtaining the sample wind power prediction data corresponding to the second historical moment. Then, based on the sample wind power real data at the second historical moment and the sample wind power prediction data, the model parameters of the updated initial wind power prediction model can be adjusted.Each time a sample wind power prediction data is obtained, the model loss value of the model can be calculated. Specifically, the sample wind power prediction data and the sample wind power real data can be used for calculation until the model loss value is less than the preset loss threshold, and the preset wind power prediction model is obtained. Here, the model loss value can be determined by the root mean square error and the mean absolute percentage error. The smaller the value of the root mean square error and the mean absolute percentage error, the higher the accuracy of the model prediction. Among them, the root mean square error can be expressed by the formula:. The mean absolute percentage error can be expressed as Where: y predict Represents sample wind power prediction data; y real represents the sample wind power true data, and m represents the number of corresponding target decomposition signals when the sample wind power true data is divided into target decomposition signals.

[0073] By applying the technical solution of this embodiment, before performing wind power prediction, the time period to be predicted can be determined first, that is, the target prediction time period can be determined first. After determining the target prediction time period, the first historical time period of the same length as the target prediction time period can be determined, and then the historical wind power data in the first historical time period can be obtained. After that, the complementary set empirical mode decomposition method can be used to perform signal decomposition on the historical wind power data, and then multiple target decomposition signals can be obtained. In addition, the historical climate data in the first historical time period can be obtained, and the target input matrix corresponding to each target decomposition signal can be determined based on the target decomposition signal and the historical climate data. The historical climate data may include one or more of historical temperature data, historical humidity data, and historical wind speed data. After determining the target input matrix, the target input matrix can be input into the preset wind power prediction model, and the first wind power prediction data corresponding to the first target prediction moment of each target decomposition signal can be obtained. Then, the data in the target input matrix corresponding to each target decomposition signal can be updated based on the first wind power prediction data corresponding to the first target prediction moment. After the target input matrix corresponding to each target decomposition signal is updated, the updated target input matrix can be input into the preset wind power prediction model again, and the second wind power prediction data corresponding to each target decomposition signal at the second target prediction moment is output. Then, the data in the target input matrix is ​​updated using the second wind power prediction data corresponding to the second target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the second target prediction moment, and the target input matrix after the second update is input into the preset wind power prediction model to obtain the third wind power prediction data corresponding to each target decomposition signal at the third target prediction moment... When the number of wind power prediction data of each target decomposition signal is consistent with the number of target prediction moments corresponding to the target prediction time period, it means that the wind power prediction data corresponding to the last target prediction moment in the target prediction time period has been obtained. After obtaining the wind power prediction data corresponding to each target decomposition signal at each target prediction moment, the predicted total wind power in the target prediction time period can be further determined based on these wind power prediction data. The embodiment of the present application can continuously update the target input matrix through wind power prediction data, thereby obtaining the wind power prediction data corresponding to each target prediction moment, and can accurately and conveniently perform long-term wind power prediction.

[0074] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another long-term wind power prediction method is provided, such as Figure 2 As shown, the method includes:

[0075] Step 201, obtaining historical wind power data in a first historical time period of the same length as the target prediction time period, and performing signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain a plurality of target decomposition signals;

[0076] In this embodiment, after determining the target prediction time period, a first historical time period of the same length as the target prediction time period can be determined first, and then the historical wind power data in the first historical time period can be obtained. Then, the complementary set empirical mode decomposition method, i.e., the CEEMD method, can be used to perform signal decomposition on the historical wind power data, and then multiple target decomposition signals can be obtained. When the traditional set empirical mode decomposition EEMD method restores the original signal, there is a lack of information and the original signal cannot be perfectly restored, which leads to a decrease in the accuracy of the wind power prediction data, and the CEEMD method can perfectly solve this problem. The principle of the CEEMD method is to introduce positive and negative white noises that are opposite to each other as auxiliary noises in the original signal, in order to make the redundant auxiliary white noise signals participating in the decomposition disappear and reduce the number of iterations.

[0077] Step 202, obtaining historical climate data within the first historical time period, taking values ​​according to the preset prediction time interval based on any of the target decomposition signals and the historical climate data, and obtaining the target input matrix corresponding to any of the target decomposition signals, wherein the historical climate data includes at least one of historical temperature data, historical humidity data, and historical wind speed data;

[0078] In this embodiment, after obtaining multiple target decomposition signals, historical climate data within the first historical time period can be obtained. Then, based on each target decomposition signal and historical climate data, values ​​can be taken according to a preset prediction time interval, and then the target input matrix corresponding to the target decomposition signal can be obtained. For example, after the historical wind power prediction data is decomposed by the complementary set empirical mode decomposition method, it is decomposed into i IMF signals, and then based on each IMF signal and historical climate data, values ​​are taken according to a preset prediction time interval of 15 minutes, and the target input matrices corresponding to the first IMF signal to the i-th IMF signal can be obtained respectively. For example, for the target input matrix corresponding to the i-th IMF signal, the target input matrix can include the historical wind power data and historical climate data indicated by the IMF signal corresponding to each target prediction moment. For the current wind power prediction method, the prediction time interval is usually 15 minutes, so the preset prediction time interval can also be 15 minutes.

[0079] Step 203, inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals;

[0080] In this embodiment, the target input matrix corresponding to each target decomposition signal can be respectively input into the preset wind power prediction model, and the preset wind power prediction model can output the first wind power prediction data at the first target prediction moment corresponding to each target decomposition signal. For example, if the historical wind power prediction data is decomposed into i IMF signals, then the preset wind power prediction model can output the first wind power prediction data corresponding to the i first target prediction moments.

[0081] Step 204, determining target update data of each target decomposition signal at the first target prediction moment based on the first wind power prediction data corresponding to each target decomposition signal and the predicted climate data corresponding to the first target prediction moment;

[0082] In this embodiment, after determining the first wind power prediction data corresponding to each target decomposition signal, the predicted climate data corresponding to the first target prediction moment can also be obtained. Here, the predicted climate data and the data included in the historical climate data are the same. For example, the historical climate data includes historical temperature data, historical humidity data, and historical wind speed data, and the predicted climate data also includes predicted temperature data, predicted humidity data, and predicted wind speed data. Then, based on the first wind power prediction data and predicted climate data corresponding to each target decomposition signal, the target update data at the first target prediction moment corresponding to each target decomposition signal can be determined.

[0083] Step 205, based on the target update data, updating the data of the historical moment with the longest distance from the current time in the target input matrix corresponding to each of the target decomposition signals, to obtain an updated target input matrix corresponding to each of the target decomposition signals, wherein the historical moment is determined based on the first historical time period and the preset prediction time interval;

[0084] In this embodiment, the data in the target input matrix corresponding to each target decomposition signal can be further updated based on the target update data. Specifically, the data corresponding to the historical moment with the longest time distance from the current time can be found from the target input matrix, and the data can be replaced with the target update data to achieve the update of the target input matrix. The historical moment here can be determined by the first historical time period and the preset prediction time interval.

[0085] Step 206, input each updated target input matrix into the preset wind power prediction model, obtain the second wind power prediction data at the second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, obtain the wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and the preset prediction time interval;

[0086] In this embodiment, after the target input matrix corresponding to each target decomposition signal is updated, the updated target input matrix can be input into the preset wind power prediction model again, and the second wind power prediction data corresponding to each target decomposition signal at the second target prediction moment can be output. For example, the first target prediction moment is 0:15, and the preset prediction time interval is 15min, then the second target prediction moment is 0:30. At this time, if the target prediction time period is 30 minutes, the wind power prediction ends, otherwise, the target input matrix update step is repeated, and the data in the target input matrix is ​​updated using the second wind power prediction data corresponding to the second target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the second target prediction moment, and the target input matrix after the second update is input into the preset wind power prediction model to obtain the third wind power prediction data corresponding to each target decomposition signal at the third target prediction moment... When the number of wind power prediction data of each target decomposition signal is consistent with the number of target prediction moments corresponding to the target prediction time period, it means that the wind power prediction data corresponding to the last target prediction moment in the target prediction time period has been obtained. For example, if the target prediction time period is 24 hours and the preset prediction time interval is 15 minutes, the target prediction time period can be divided into 96 target prediction moments. When the number of wind power prediction data of each target decomposition signal is 96, it means that the wind power prediction data of the last target prediction moment has been obtained.

[0087] Step 207: Determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal.

[0088] In this embodiment, after obtaining the wind power prediction data corresponding to each target decomposition signal at each target prediction moment, the predicted total wind power within the target prediction time period can be further determined based on these wind power prediction data. For example, after the historical wind power prediction data is decomposed into i IMF signals by complementary set empirical mode decomposition method, i first wind power prediction data, second wind power prediction data can be obtained, etc. until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, all wind power prediction data corresponding to all target decomposition signals are added together to obtain the predicted total wind power.

[0089] In an embodiment of the present application, optionally, step 203 specifically includes: inputting the target input matrix corresponding to each of the target decomposition signals into the convolutional neural network model to obtain a target space feature vector corresponding to each of the target decomposition signals; inputting the target space feature vector into the long short-term memory model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals.

[0090] In this embodiment, after obtaining the target input matrix corresponding to each target decomposition signal, the target input matrix corresponding to each target decomposition signal can be input into the convolutional neural network model of the preset wind power prediction model, and the target space feature vector corresponding to each target decomposition signal can be extracted through the convolutional neural network model. Then, the extracted target space feature vector can be input into the long short-term memory model of the preset wind power prediction model, and the first wind power prediction data at the first target prediction moment corresponding to each target decomposition signal can be obtained.

[0091] In an embodiment of the present application, optionally, after step 207, the method further includes: when any moment within the target prediction time period is before the current time, obtaining the actual total wind power corresponding to the predicted total wind power, and when the error between the actual total wind power and the predicted total wind power is greater than a preset error threshold, based on the wind power prediction data corresponding to each of the target prediction moments within the target prediction time period, the wind power actual data and the actual climate data, training the preset wind power prediction model to obtain an updated preset wind power prediction model.

[0092] In this embodiment, when any moment within the target prediction time period is in the past, that is, the target prediction time period has passed, then the actual total wind power corresponding to the predicted total wind power of the target prediction time period can be obtained, and the predicted total wind power and the actual total wind power can be further compared. When the difference between the predicted total wind power and the actual total wind power is greater than the preset error threshold, it means that the prediction accuracy of the preset wind power prediction model is not high, and then the preset wind power prediction model can be trained based on the wind power prediction data, wind power actual data, and actual climate data at each target prediction moment within the target prediction time period, and then an updated preset wind power prediction model is obtained. When the difference between the predicted total wind power and the actual total wind power obtained by predicting through the preset wind power prediction model is large, the preset wind power prediction model is re-trained in this application, which is conducive to improving the prediction accuracy of the preset wind power prediction model.

[0093] In the embodiment of the present application, optionally, step 207 may specifically include step 207-1 or step 207-2:

[0094] Step 207-1, based on the wind power prediction data corresponding to each of the target decomposition signals at any of the target prediction moments, determine the first wind power corresponding to any of the target prediction moments, and based on the first wind power corresponding to each of the target prediction moments, determine the predicted total wind power corresponding to the target prediction time period.

[0095] In this embodiment, after obtaining the wind power prediction data corresponding to each target prediction moment and each target decomposition signal, the first wind power corresponding to each target prediction moment can be determined first, that is, for each target prediction moment, the wind power prediction values ​​corresponding to different target decomposition signals corresponding to the target prediction moment are added together to obtain the first wind power corresponding to the target prediction moment. Then, the first wind powers corresponding to different target prediction moments can be added together to determine the predicted total wind power corresponding to the target prediction time period, that is, the total power output of wind power generation within the target prediction time period.

[0096] Step 207-2, based on the wind power prediction data corresponding to each target prediction moment of any of the target decomposition signals, determine the second wind power corresponding to any of the target decomposition signals, and based on the second wind power corresponding to each of the target decomposition signals, determine the predicted total wind power corresponding to the target prediction time period.

[0097] In this embodiment, after obtaining the wind power prediction data corresponding to each target prediction moment and each target decomposition signal, the wind power prediction data corresponding to each target decomposition signal at all target prediction moments can be added together to obtain the second wind power of each target decomposition signal within the target prediction time period. Then, the second wind powers corresponding to different target decomposition signals can be added together to determine the predicted total wind power corresponding to the target prediction time period, that is, the total power output of wind power generation within the target prediction time period.

[0098] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a long-term wind power prediction device, such as Figure 3 As shown, the device comprises:

[0099] A signal decomposition module is used to obtain historical wind power data in a first historical time period of the same length as the target prediction time period, and perform signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain multiple target decomposition signals;

[0100] a matrix determination module, configured to determine a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data and historical wind speed data;

[0101] A first input module is used to input the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and based on the first wind power prediction data, update the target input matrix corresponding to each target decomposition signal, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model;

[0102] A second input module is used to input each updated target input matrix into the preset wind power prediction model to obtain second wind power prediction data at a second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, to obtain wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and a preset prediction time interval;

[0103] The wind power determination module is used to determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal.

[0104] Optionally, the matrix determination module is specifically used to:

[0105] The historical climate data within the first historical time period is obtained, and based on any of the target decomposition signals and the historical climate data, values ​​are taken according to the preset prediction time interval to obtain the target input matrix corresponding to any of the target decomposition signals.

[0106] Optionally, the first input module specifically includes:

[0107] A matrix input unit, used for inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals;

[0108] a data determination unit, configured to determine target update data of each target decomposition signal at the first target prediction moment based on the first wind power prediction data corresponding to each target decomposition signal and the predicted climate data corresponding to the first target prediction moment;

[0109] A data updating unit is used to update the data of the historical moment with the longest distance from the current time in the target input matrix corresponding to each of the target decomposition signals according to the target update data, so as to obtain an updated target input matrix corresponding to each of the target decomposition signals, wherein the historical moment is determined based on the first historical time period and the preset prediction time interval.

[0110] Optionally, the wind power determination module is specifically used to:

[0111] Based on the wind power prediction data corresponding to each of the target decomposition signals at any of the target prediction moments, determine the first wind power corresponding to any of the target prediction moments, and based on the first wind power corresponding to each of the target prediction moments, determine the predicted total wind power corresponding to the target prediction time period; or, based on the wind power prediction data corresponding to each of the target prediction moments of any of the target decomposition signals, determine the second wind power corresponding to any of the target decomposition signals, and based on the second wind power corresponding to each of the target decomposition signals, determine the predicted total wind power corresponding to the target prediction time period.

[0112] Optionally, the matrix input unit is specifically used to:

[0113] The target input matrix corresponding to each of the target decomposition signals is input into the convolutional neural network model to obtain the target space feature vector corresponding to each of the target decomposition signals; the target space feature vector is input into the long short-term memory model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals.

[0114] Optionally, the device further comprises:

[0115] A model updating module is used for, after determining the predicted total wind power corresponding to the target prediction time period, obtaining the actual total wind power corresponding to the predicted total wind power when any moment within the target prediction time period is before the current time, and when the error between the actual total wind power and the predicted total wind power is greater than a preset error threshold, training the preset wind power prediction model based on the wind power prediction data corresponding to each of the target prediction moments within the target prediction time period, the actual wind power data and the actual climate data, to obtain an updated preset wind power prediction model.

[0116] Optionally, the device further comprises:

[0117] A data partitioning module, used for obtaining sample data in a second historical time period before inputting the target input matrix into a preset wind power prediction model, and dividing the sample data into training sample data and update sample data according to time, wherein the sample data includes sample wind power real data and sample climate data;

[0118] A model training module is used to determine a first sample input matrix based on the training sample data, input the first sample input matrix into an initial wind power prediction model, obtain sample wind power prediction data corresponding to a first historical moment, and adjust the model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model; update the training sample data through the updated sample data, determine a second sample input matrix, input the second sample input matrix into the updated initial wind power prediction model, obtain sample wind power prediction data corresponding to a second historical moment, and adjust the model parameters of the updated initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the second historical moment until the model loss value is less than a preset loss threshold, thereby obtaining the preset wind power prediction model.

[0119] It should be noted that for other corresponding descriptions of the functional units involved in the long-term wind power prediction device provided in the embodiment of the present application, reference can be made to Figure 1 to Figure 2 The corresponding description in the method will not be repeated here.

[0120] Based on the above Figure 1 to Figure 2 The method shown in the embodiment of the present application is accordingly provided with a storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned Figure 1 to Figure 2 The long-term wind power forecasting method shown.

[0121] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0122] Based on the above Figure 1 to Figure 2 The method shown, and Figure 3 In order to achieve the above-mentioned purpose, the embodiment of the present application further provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figure 1 to Figure 2 The long-term wind power forecasting method shown.

[0123] Optionally, the computer device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), etc.

[0124] Those skilled in the art will appreciate that the computer device structure provided in this embodiment does not limit the computer device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0125] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages and saves the hardware and software resources of the computer device, and supports the operation of information processing programs and other software and / or programs. The network communication module is used to realize communication between the components inside the storage medium, and communication with other hardware and software in the physical device.

[0126] Through the description of the above implementation mode, the technicians in this field can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. Before performing wind power prediction, the time period to be predicted can be determined first, that is, the target prediction time period can be determined first. After determining the target prediction time period, the first historical time period of the same length as the target prediction time period can be determined, and then the historical wind power data in the first historical time period can be obtained. After that, the complementary set empirical mode decomposition method can be used to perform signal decomposition on the historical wind power data, and then multiple target decomposition signals can be obtained. In addition, the historical climate data in the first historical time period can be obtained, and the target input matrix corresponding to each target decomposition signal can be determined based on the target decomposition signal and the historical climate data. The historical climate data can include one or more of historical temperature data, historical humidity data, and historical wind speed data. After determining the target input matrix, the target input matrix can be input into the preset wind power prediction model, and the first wind power prediction data corresponding to the first target prediction moment of each target decomposition signal can be obtained. Then, the data in the target input matrix corresponding to each target decomposition signal can be updated based on the first wind power prediction data corresponding to the first target prediction moment. After the target input matrix corresponding to each target decomposition signal is updated, the updated target input matrix can be input into the preset wind power prediction model again, and the second wind power prediction data corresponding to each target decomposition signal at the second target prediction moment is output. Then, the data in the target input matrix is ​​updated using the second wind power prediction data corresponding to the second target prediction moment, and one or more of the temperature data, humidity data and wind speed data predicted by the meteorological department corresponding to the second target prediction moment, and the target input matrix after the second update is input into the preset wind power prediction model to obtain the third wind power prediction data corresponding to each target decomposition signal at the third target prediction moment... When the number of wind power prediction data of each target decomposition signal is consistent with the number of target prediction moments corresponding to the target prediction time period, it means that the wind power prediction data corresponding to the last target prediction moment in the target prediction time period has been obtained. After obtaining the wind power prediction data corresponding to each target decomposition signal at each target prediction moment, the predicted total wind power in the target prediction time period can be further determined based on these wind power prediction data. The embodiment of the present application can continuously update the target input matrix through wind power prediction data, thereby obtaining the wind power prediction data corresponding to each target prediction moment, and can accurately and conveniently perform long-term wind power prediction.

[0127] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0128] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.

Claims

1. A long-term wind power prediction method, characterized in that: include: Acquire historical wind power data in a first historical time period of the same length as the target prediction time period, and perform signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain a plurality of target decomposition signals; Determine a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data and historical wind speed data; Input the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and update the target input matrix corresponding to each target decomposition signal based on the first wind power prediction data, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model; Input each updated target input matrix into the preset wind power prediction model to obtain the second wind power prediction data at the second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, and obtain the wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and the preset prediction time interval; Determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal; Before inputting the target input matrix into a preset wind power prediction model, the method further includes: Acquire sample data in a second historical time period, and divide the sample data into training sample data and update sample data according to time, wherein the sample data includes sample wind power real data and sample climate data; Determine a first sample input matrix based on the training sample data, input the first sample input matrix into an initial wind power prediction model to obtain sample wind power prediction data corresponding to a first historical moment, and adjust model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model; The training sample data is updated by the updated sample data, a second sample input matrix is ​​determined, the second sample input matrix is ​​input into the updated initial wind power prediction model, the sample wind power prediction data corresponding to the second historical moment is obtained, and based on the sample wind power real data at the second historical moment and the sample wind power prediction data, the model parameters of the updated initial wind power prediction model are adjusted until the model loss value is less than the preset loss threshold, thereby obtaining the preset wind power prediction model.

2. The method according to claim 1, characterized in that The step of determining a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period specifically includes: The historical climate data within the first historical time period is obtained, and based on any of the target decomposition signals and the historical climate data, values ​​are taken according to the preset prediction time interval to obtain the target input matrix corresponding to any of the target decomposition signals.

3. The method according to claim 2, characterized in that The step of inputting the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and updating the target input matrix corresponding to each target decomposition signal based on the first wind power prediction data, specifically includes: Inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals; Determine target update data of each target decomposition signal at the first target prediction moment based on the first wind power prediction data corresponding to each target decomposition signal and the predicted climate data corresponding to the first target prediction moment; Based on the target update data, the data of the historical moment with the longest distance from the current time in the target input matrix corresponding to each target decomposition signal is updated to obtain an updated target input matrix corresponding to each target decomposition signal, wherein the historical moment is determined based on the first historical time period and the preset prediction time interval.

4. The method according to claim 2 or 3, characterized in that: The step of determining the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal specifically includes: Determine the first wind power corresponding to any of the target prediction moments based on the wind power prediction data corresponding to each of the target decomposition signals at any of the target prediction moments, and determine the predicted total wind power corresponding to the target prediction time period based on the first wind power corresponding to each of the target prediction moments; or Based on the wind power prediction data corresponding to each target prediction moment of any of the target decomposition signals, the second wind power corresponding to any of the target decomposition signals is determined, and based on the second wind power corresponding to each of the target decomposition signals, the predicted total wind power corresponding to the target prediction time period is determined.

5. The method according to claim 3, characterized in that: The step of inputting the target input matrix corresponding to each of the target decomposition signals into the preset wind power prediction model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals specifically includes: Inputting the target input matrix corresponding to each of the target decomposition signals into the convolutional neural network model to obtain a target space feature vector corresponding to each of the target decomposition signals; The target space feature vector is input into the long short-term memory model to obtain the first wind power prediction data at the first target prediction moment corresponding to each of the target decomposition signals.

6. The method according to claim 1, characterized in that After determining the predicted total wind power corresponding to the target prediction time period, the method further includes: When any moment within the target prediction time period is before the current time, the actual total wind power corresponding to the predicted total wind power is obtained, and when the error between the actual total wind power and the predicted total wind power is greater than a preset error threshold, the preset wind power prediction model is trained based on the wind power prediction data corresponding to each of the target prediction moments within the target prediction time period, the actual wind power data and the actual climate data, to obtain an updated preset wind power prediction model.

7. A long-term wind power prediction device, characterized in that: include: A signal decomposition module is used to obtain historical wind power data in a first historical time period of the same length as the target prediction time period, and perform signal decomposition on the historical wind power data based on a complementary set empirical mode decomposition method to obtain multiple target decomposition signals; a matrix determination module, configured to determine a target input matrix corresponding to each of the target decomposition signals based on the target decomposition signals and the historical climate data in the first historical time period, wherein the historical climate data includes at least one of historical temperature data, historical humidity data and historical wind speed data; A first input module is used to input the target input matrix into a preset wind power prediction model to obtain first wind power prediction data at a first target prediction moment corresponding to each target decomposition signal, and based on the first wind power prediction data, update the target input matrix corresponding to each target decomposition signal, wherein the preset wind power prediction model includes a trained convolutional neural network model and a long short-term memory model; A second input module is used to input each updated target input matrix into the preset wind power prediction model to obtain second wind power prediction data at a second target prediction moment corresponding to each target decomposition signal, until the number of wind power prediction data corresponding to each target decomposition signal is the same as the number of target prediction moments corresponding to the target prediction time period, to obtain wind power prediction data at the last target prediction moment corresponding to each target decomposition signal, wherein the target prediction moment is determined based on the target prediction time period and a preset prediction time interval; A wind power determination module, configured to determine the predicted total wind power corresponding to the target prediction time period based on the wind power prediction data at each target prediction moment corresponding to each target decomposition signal; The device also includes: A data partitioning module, used for obtaining sample data in a second historical time period before inputting the target input matrix into a preset wind power prediction model, and dividing the sample data into training sample data and update sample data according to time, wherein the sample data includes sample wind power real data and sample climate data; A model training module is used to determine a first sample input matrix based on the training sample data, input the first sample input matrix into an initial wind power prediction model, obtain sample wind power prediction data corresponding to a first historical moment, and adjust the model parameters of the initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the first historical moment, and update the initial wind power prediction model, wherein the initial wind power prediction model includes an initial convolutional neural network model and an initial long short-term memory model; update the training sample data through the updated sample data, determine a second sample input matrix, input the second sample input matrix into the updated initial wind power prediction model, obtain sample wind power prediction data corresponding to a second historical moment, and adjust the model parameters of the updated initial wind power prediction model based on the sample wind power real data and the sample wind power prediction data at the second historical moment until the model loss value is less than a preset loss threshold, thereby obtaining the preset wind power prediction model.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

Images