CN-121996915-A - Wind farm electricity limiting data identification and correction method and device

CN121996915ACN 121996915 ACN121996915 ACN 121996915ACN-121996915-A

Abstract

The invention relates to the technical field of water wind power prediction and discloses a method and a device for identifying and correcting wind power station electricity limiting data, wherein the method comprises the steps of adopting a CNN-LSTM model to learn normal wind speed-power curve characteristics, and effectively capturing the correlation characteristics of wind speed and power through the combination of a one-dimensional rolling layer and a pooling layer, so that the identification accuracy and stability of abnormal sections of a wind power station are improved; the wind power plant operation data processing method based on the wind power plant abnormal data analysis and the wind power plant abnormal data analysis has the advantages that complicated wind power plant operation states can be effectively identified and processed through multidimensional wind condition feature analysis and similarity calculation, the limitation that only single features are relied on in the prior art is overcome, the comprehensiveness and accuracy of wind power plant abnormal data identification are improved, the weight of each similar section is determined based on the similarity of the wind power plant states, the reconstruction power value of the limited sections is obtained through a weighted summation mode, and the reliability and accuracy of the wind power plant operation data are improved.

Inventors

ZHAI RAN
LU GUILIN
YIN ZHAOKAI
LV ZHENYU
LIANG LILI
LI WAN
ZHANG XIAOMEI
DONG YIYANG
YANG HENG
LI MENGJIE
ZHANG ZHENYU

Assignees

中国长江三峡集团有限公司

Dates

Publication Date: 20260508
Application Date: 20260120

Claims (11)

1. A method for identifying and correcting wind farm electricity limiting data, comprising: based on the preprocessed wind speed-power time sequence data, adopting a CNN-LSTM model to learn the characteristics of a normal wind speed-power curve; based on the CNN-LSTM model, learning the characteristic of a normal wind speed-power curve, identifying an abnormal section of wind speed-power time series data, and verifying and correcting the abnormal section; And extracting wind condition characteristics in the electricity limiting section, carrying out similarity analysis, and obtaining a reconstruction power value of the electricity limiting section based on the similarity of the fan states.
2. A method of identifying and modifying wind farm electrical power limit data according to claim 1, wherein the preprocessing of the wind speed-power time series data comprises: normalizing the wind speed-power time series data by adopting a quantile normalization method; And filtering the wind speed and power data by adopting a Savitzky-Golay filter.
3. A method of identifying and modifying wind farm electrical power limit data according to claim 1, wherein the process of learning the normal wind speed-power curve characteristics using the CNN-LSTM model comprises: taking the preprocessed wind speed-power time series data as a training set, dividing the continuous wind speed-power time series data into a plurality of samples according to a time window, and preprocessing sample data; capturing the correlation characteristics of wind speed and power at the same time point by adopting one-dimensional convolution; Adopting a pooling layer to add MaxPooling D, compressing the dimension of the associated feature, and reserving the key local feature; inputting the feature vector output by the CNN into an LSTM layer, and capturing the change rule of the feature along with time; Outputting the LSTM layer and then connecting with 3 full-connection layers, and further fusing space-time characteristics; The output layer adopts a linear activation function to output the reconstructed wind speed-power data, and the output data and the input data are in the same dimension.
4. A method of identifying and modifying wind farm electrical limit data according to claim 3, wherein the process of identifying abnormal sections of wind speed-power time series data comprises: Calculating error values of the reconstruction power and the input power, and considering that the window comprises a suspected limit section when the reconstruction error exceeds a preset threshold value; Further analyzing the steep drop characteristic of the power sequence on the basis of detecting the reconstruction error window; calculating the power acceleration of the abnormal section based on the abrupt drop feature, and marking the acceleration smaller than a preset threshold value as a first section; And verifying the first section through SCADA state information, screening out the abrupt power drop caused by extreme weather, and obtaining the electricity limiting section.
5. A method of identifying and modifying wind farm electrical power limit data according to claim 1, wherein the process of extracting wind condition characteristics within the electrical power limit segments and performing similarity analysis comprises: extracting wind condition characteristics in the electricity limiting section; Searching a plurality of normal similar sections with similar wind conditions to the limited electric section in the time range of the same season as the limited electric section, and taking the plurality of normal similar sections as candidate sections; Extracting wind speed characteristics, wind direction characteristics and turbulence intensity characteristics of the electricity limiting section and candidate sections in the same season as the electricity limiting section; calculating Euclidean distance between the wind speed characteristics and wind direction characteristics of the electricity limiting section and the candidate sections and the turbulence intensity to obtain initial similarity of each candidate section; screening a preset number of target candidate sections according to the initial similarity; And calculating the similarity of the wind conditions of the limited electric section and the target candidate section by adopting the DTW to obtain a plurality of similar sections.
6. A method of identifying and modifying wind farm electrical power limits as defined in claim 1, wherein the step of deriving the reconstructed power value of the electrical power limits based on a similarity of the fan states comprises: determining weights of all similar sections based on the similarity of the fan states; And carrying out weighted summation on the power data of the similar section based on the weight value to obtain the reconstruction power value of the limited electric section.
7. A method of identifying and modifying wind farm electrical power limit data as in claim 6, wherein the step of deriving the plurality of reconstructed powers when a period of the electrical power limit time is simultaneously present in a plurality of overlapping base time windows comprises: extracting the reconstruction power values of the limit electricity time in a plurality of sections, positioning the time points corresponding to the limit electricity time for each section containing the limit electricity time, and extracting the reconstruction power of the time points; And fusing candidate values according to the section credibility weights, wherein the section containing the electricity limiting time has different credibility of the representation of the electricity limiting time period, and obtaining the final reconstruction power by using weighted average after weight assignment.
8. A wind farm electricity limiting data identification and correction device, the device comprising: The model training module is used for learning the characteristics of a normal wind speed-power curve by adopting a CNN-LSTM model based on the preprocessed wind speed-power time sequence data; The identifying and correcting module is used for learning the characteristics of a normal wind speed-power curve based on the CNN-LSTM model, identifying an abnormal section of the wind speed-power time sequence data, and verifying and correcting the abnormal section; and the reconstruction module is used for extracting wind condition characteristics in the electricity limiting section and carrying out similarity analysis, and obtaining a reconstruction power value of the electricity limiting section based on the similarity of the fan states.
9. A computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the wind farm electricity limit data identification and correction method of any of claims 1 to 7.
10. A computer-readable storage medium having stored thereon computer instructions for causing a computer to perform the wind farm electricity limit data identification and correction method according to any of claims 1 to 7.
11. A computer program product comprising computer instructions for causing a computer to perform a method of identifying and modifying wind farm electricity limit data according to any of claims 1 to 7.

Description

Wind farm electricity limiting data identification and correction method and device Technical Field The invention relates to the technical field of water wind power prediction, in particular to a wind farm electricity limiting data identification and correction method and device. Background With the rapid development of wind power technology, the wind power installation scale is continuously enlarged and rapidly increased, and the wind power installation scale becomes an important development direction of a novel power system in a modern energy system. Currently, in the running process of a wind farm, the wind farm is limited by multiple factors such as a power grid dispatching mechanism, power consumption capability and the like, and the electricity limiting phenomenon occurs sometimes. The existence of the electricity limiting data not only breaks the continuity and regularity of the normal power data of the wind farm, but also brings a plurality of challenges for subsequent data analysis and processing. Currently, the main methods for processing wind power plant power limit data in the industry mainly comprise two types, namely statistical analysis based on historical data and a simple mathematical model. According to the statistical analysis method based on the historical data, simple mean filling or similar pattern matching is usually carried out on the electricity limiting data according to the distribution characteristics of the historical synchronous power data. However, the method cannot fully consider the dynamic influence of key meteorological factors such as wind speed, wind direction, turbulence intensity and the like on power, and is difficult to adapt to complex and changeable operating environments of a wind farm, so that the deviation between a data reconstruction result and actual conditions is large. However, a simple mathematical model (such as a linear regression model) is difficult to accurately capture the complex relationship behind the electricity limiting data due to the high nonlinearity and uncertainty of the power variation of the wind farm, and the accurate reconstruction of the electricity limiting data cannot be realized. Disclosure of Invention In view of the above, the invention provides a method and a device for identifying and correcting wind farm electricity limiting data, which are used for solving the problems of low accuracy of identifying the electricity limiting data and poor reconstruction accuracy. The invention provides a wind farm electricity limiting data identification and correction method, which comprises the steps of adopting a CNN-LSTM model to learn normal wind speed-power curve characteristics based on preprocessed wind speed-power time series data, adopting the CNN-LSTM model to learn normal wind speed-power curve characteristics based on the CNN-LSTM model, identifying abnormal sections of the wind speed-power time series data, verifying and correcting the abnormal sections, extracting wind condition characteristics in the electricity limiting sections, carrying out similarity analysis, and obtaining a reconstruction power value of the electricity limiting sections based on the similarity of fan states. According to the wind power station wind power generation method, the CNN-LSTM model is adopted to learn the characteristics of a normal wind speed-power curve, and the correlation characteristics of wind speed and power can be effectively captured through the combination of the one-dimensional rolling layer and the pooling layer, so that the identification accuracy and stability of abnormal sections of a wind power station are improved; the wind power plant operation data processing method based on the wind power plant abnormal data analysis and the wind power plant abnormal data analysis has the advantages that complicated wind power plant operation states can be effectively identified and processed through multidimensional wind condition feature analysis and similarity calculation, the limitation that only single features are relied on in the prior art is overcome, the comprehensiveness and accuracy of wind power plant abnormal data identification are improved, the weight of each similar section is determined based on the similarity of the wind power plant states, the reconstruction power value of the limited sections is obtained through a weighted summation mode, and the reliability and accuracy of the wind power plant operation data are improved. In an alternative implementation, the preprocessing of the wind speed-power time series data comprises normalizing the wind speed-power time series data by a quantile normalization method and filtering the wind speed and power data by a Savitzky-Golay filter. In an alternative implementation mode, a process of learning normal wind speed-power curve features by adopting a CNN-LSTM model comprises the steps of taking preprocessed wind speed-power time series data as a training set, segmenting continuous wind