CN-121983941-A - Wind power prediction method, device, equipment and storage medium

CN121983941ACN 121983941 ACN121983941 ACN 121983941ACN-121983941-A

Abstract

The application discloses a wind power prediction method, a device, equipment and a storage medium. The method comprises the steps of obtaining wind power data, decomposing the wind power data into a plurality of data subsequences by using a variation modal decomposition technology, predicting each data subsequence by combining a long-period memory network model to obtain a prediction result, and weighting and superposing the prediction results of each data subsequence to obtain a final wind power prediction power value. The method comprises the steps of collecting and sorting wind power data of a given wind power plant, preprocessing the data, decomposing the wind power data by using a variation modal decomposition technology, predicting each data subsequence by using a long-period memory network model, and weighting and superposing prediction results of the data subsequences to obtain a final wind power prediction power value. The method can effectively overcome the defects of the traditional prediction method in the aspect of processing the fluctuation and the randomness of the wind power, so that the wind power plant can more effectively participate in the operation and the scheduling of the power system.

Inventors

DENG XIAOGUO
WANG JUN
NING MENGMENG
HUANG YAN
Liang Xinkun
CHEN JINDIAN
Chi Lixun

Assignees

中国石油天然气股份有限公司
南方石油勘探开发有限责任公司

Dates

Publication Date: 20260505
Application Date: 20241030

Claims (10)

1. A method for predicting wind power, the method comprising: Acquiring wind power data; Decomposing wind power data into a plurality of data subsequences by using a variation modal decomposition technology; Predicting each data subsequence by combining with the long-term and short-term memory network model to obtain a prediction result; and weighting and superposing the prediction results of the data subsequences to obtain a final wind power prediction power value.
2. The wind power prediction method according to claim 1, wherein the step of acquiring wind power data includes: collecting initial wind power data from sensors, monitoring systems and databases of a wind farm; And removing abnormal values and filling missing data from the initial wind power data to obtain wind power data.
3. A method of predicting wind power as claimed in claim 1, wherein the step of decomposing wind power data into relatively smooth sub-sequences using a variational modal decomposition technique comprises: Performing Hilbert transformation on wind power data to obtain a complex time domain signal, and calculating a single-side frequency spectrum of the complex time domain signal; Initializing spectrum parameters of the unilateral spectrum, wherein the spectrum parameters comprise center frequency and bandwidth; Decomposing by searching the initialized spectrum parameters with the minimum difference between the decomposition mode of the complex time domain signal and the residual error to obtain an eigenmode function; Performing inverse transformation on the eigenmode function to obtain time domain mode data values of all modes after decomposition; Reconstructing the time domain modal data to obtain a plurality of data subsequences.
4. A wind power prediction method according to claim 3, wherein the step of obtaining the eigenmode function by searching for initialized spectral parameters of the complex time domain signal that minimize the difference between the decomposition mode and the residual of the signal comprises: constructing a constraint variation equation; Introducing Lagrangian multipliers and second order penalty factors, and converting the constraint variation equation into an unconstrained variation equation; and obtaining the best eigenmode function of the center frequency and the bandwidth by using a traversal optimizing method.
5. The wind power prediction method according to claim 3, wherein the step of predicting each data subsequence by combining with the long-short-term memory network model to obtain a prediction result comprises: capturing time sequence dependency in wind power data through a long-short-term memory neural network; the wind power data are brought into the calculation of the current time step through an input door, and then the characteristics of the wind power data are weighted; Determining memory of time steps before reservation and forgetting through a forgetting door; The memory state in the past is combined with the data subsequence input at present through the forgetting gate to update the internal memory state; generating an output gate by combining the current input data subsequence with the memory state of the previous time step; and generating a predicted power value of each data subsequence through the output gate, and obtaining a predicted result through the predicted power value.
6. The wind power prediction method according to claim 5, wherein the step of generating the predicted power value of each data sub-sequence through the output gate and obtaining the prediction result through the predicted power value comprises: Constructing a plurality of long-term and short-term memory neural networks; And using the mean square error to measure the difference between the predicted power value and the actual value, and outputting the predicted result of each data subsequence through random gradient descent iterative optimization.
7. The wind power prediction method according to claim 6, wherein the step of weighting and superposing the prediction results of the data subsequences to obtain a final wind power prediction power value comprises the steps of: And collecting the prediction results of each data subsequence, and realizing superposition calculation of wind power prediction power values by distributing weights to the prediction results of different data subsequences.
8. A wind power prediction apparatus, the apparatus comprising: the data acquisition module is used for acquiring wind power data; The transformation modal decomposition module is used for decomposing wind power data into a plurality of data subsequences by utilizing a transformation modal decomposition technology; the long-period memory network prediction module is used for predicting each data subsequence by combining with the long-period memory network model to obtain a prediction result; And the integration module is used for weighting and superposing the prediction results of the data subsequences to obtain a final wind power prediction power value.
9. Wind power prediction device, characterized in that it comprises a memory, a processor and a computer program stored on the memory and running on the processor, the computer program being configured to implement the steps of the wind power prediction method according to any of claims 1 to 7.
10. A storage medium, characterized in that the storage medium is a computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the steps of the wind power prediction method according to any one of claims 1 to 7.

Description

Wind power prediction method, device, equipment and storage medium Technical Field The application relates to the technical field of wind power prediction, in particular to a wind power prediction method, a wind power prediction device, wind power prediction equipment and a storage medium. Background Wind energy is increasingly taking an important role in energy production as a clean renewable energy source. However, the instability and volatility of wind power makes the operation and scheduling of wind farms challenging. In actual operation, accurately predicting wind power change is important to optimizing power grid dispatching and energy management. The existing wind power prediction method has some defects. Traditional statistical models often have difficulty capturing complex relationships in wind power time series, resulting in poor prediction accuracy. The method based on the physical model is limited by the complexity of the wind field and the difficult problem of parameter estimation, and the effect of the method in practical application is limited. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present application and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The application mainly aims to provide a wind power prediction method, a device, equipment and a storage medium, which can improve the accuracy and reliability of wind power prediction. In order to achieve the above object, the present application provides a wind power prediction method, which includes: Acquiring wind power data; Decomposing wind power data into a plurality of data subsequences by using a variation modal decomposition technology; Predicting each data subsequence by combining with the long-term and short-term memory network model to obtain a prediction result; and weighting and superposing the prediction results of the data subsequences to obtain a final wind power prediction power value. In an embodiment, the step of obtaining wind power data includes: collecting initial wind power data from sensors, monitoring systems and databases of a wind farm; And removing abnormal values and filling missing data from the initial wind power data to obtain wind power data. In an embodiment, the step of decomposing the wind power data into relatively smooth sub-sequences using a variational modal decomposition technique comprises: Performing Hilbert transformation on wind power data to obtain a complex time domain signal, and calculating a single-side frequency spectrum of the complex time domain signal; Initializing spectrum parameters of the unilateral spectrum, wherein the spectrum parameters comprise center frequency and bandwidth; Decomposing by searching the initialized spectrum parameters with the minimum difference between the decomposition mode of the complex time domain signal and the residual error to obtain an eigenmode function; Performing inverse transformation on the eigenmode function to obtain time domain mode data values of all modes after decomposition; Reconstructing the time domain modal data to obtain a plurality of data subsequences. In an embodiment, the step of obtaining the eigenmode function by searching for an initialized spectrum parameter of the complex time domain signal, which minimizes a difference between a decomposition mode and a residual of the signal, includes: constructing a constraint variation equation; Introducing Lagrangian multipliers and second order penalty factors, and converting the constraint variation equation into an unconstrained variation equation; and obtaining the best eigenmode function of the center frequency and the bandwidth by using a traversal optimizing method. In one embodiment, the step of predicting each data subsequence by combining with the long-short-term memory network model to obtain a prediction result includes: capturing time sequence dependency in wind power data through a long-short-term memory neural network; the wind power data are brought into the calculation of the current time step through an input door, and then the characteristics of the wind power data are weighted; Determining memory of time steps before reservation and forgetting through a forgetting door; The memory state in the past is combined with the data subsequence input at present through the forgetting gate to update the internal memory state; generating an output gate by combining the current input data subsequence with the memory state of the previous time step; and generating a predicted power value of each data subsequence through the output gate, and obtaining a predicted result through the predicted power value. In one embodiment, the step of generating the predicted power value of each data sub-sequence through the output gate, and obtaining the predicted result through the predicted power value includes: Constructing a plurality of long-term and short-term memory neura