Wind power plant power prediction method, system and equipment based on adaptive window and time-frequency feature fusion, and medium

Abstract

In order to improve the precision of wind power plant power prediction, the invention discloses a wind power plant power prediction method, system, equipment and medium based on adaptive window and time-frequency feature fusion, the method comprises the following steps: collecting historical wind power data to form a power time-domain function, setting a window width value for selecting a sampling range, calculating a time domain characteristic value in each window, converting a power time domain function into a frequency domain function, calculating a frequency domain characteristic value in each window, and inputting the time domain characteristic values, the frequency domain characteristic values and historical weather forecast data into a power prediction model to perform model training; and inputting future weather forecast data into the trained power prediction model to predict future wind power plant power. The method improves the flexibility of feature extraction, and adapts to non-stationary data; and time-frequency domain features and meteorological data are fused, so that the prediction precision is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to wind farm power prediction technology, in particular to a wind farm power prediction method, system, device and medium based on adaptive window and time-frequency feature fusion. BACKGROUND

[0002] Wind farm power prediction refers to predicting the output power of a wind farm in a future period of time by using various data and technical means. This prediction is crucial for the stable operation and efficient management of power systems.

[0003] The core data of wind farm power prediction is historical power and future weather forecast data, which mainly relies on the accuracy of weather forecasts to improve the accuracy of power prediction. However, due to the complexity and variability of weather forecasting technology, the accuracy of power prediction is not high, which to some extent limits the large-scale grid-connected utilization of wind power. SUMMARY

[0004] The present application provides a wind farm power prediction method, system, device and medium based on adaptive window and time-frequency feature fusion to address the shortcomings of the prior art. An adaptive window adjustment algorithm is proposed to dynamically optimize the window width of time-frequency feature extraction. An improved neural network model is designed to fuse time-frequency features and meteorological data to improve prediction accuracy.

[0005] To achieve the above-mentioned purposes, the following technical solutions are adopted:

[0006] A wind farm power prediction method based on adaptive window and time-frequency feature fusion includes the following steps:

[0007] Collect historical wind power data sets to generate a power time-domain function, set a window width value to select a sampling range, and calculate the time-domain feature value in each window,

[0008] Convert the power time-domain function to a frequency-domain function and calculate the frequency-domain feature value in each window,

[0009] Input the time-domain feature value, frequency-domain feature value, and historical weather forecast data into the power prediction model for model training;

[0010] Input future weather forecast data into the trained power prediction model to predict future wind farm power.

[0011] To optimize the above technical solutions, the following specific measures are taken:

[0012] Further, the expression of the power time-domain function is:

[0013] f(t)=(P1,P2,…,P t )

[0014] where f(t) is a power time-domain function, P t is the historical wind power at time t.

[0015] Further, the setting window width value is specifically:

[0016] According to the local data variation coefficient dynamically adjusting the window width value m, the local data variation coefficient is the standard deviation divided by the mean value of the wind power data in the window, when the local data variation coefficient is greater than or equal to a preset threshold value, the window width value is reduced to Symbol Indicates rounding down, the window width value indicates the number of data in the window.

[0017] Further, the time-domain characteristic value includes the mean value, the variance, the maximum value and the minimum value, calculating the time-domain characteristic value in each window, the formula is as follows:

[0018]

[0019] where, is the mean value of the wind power data in the i-th window, P (i-1)×m+1 Indicates the 1st wind power data in the i-th window, P i×m Indicates the mth wind power data in the i-th window, m indicates the window width value;

[0020]

[0021] where, s i is the variance of the wind power data in the i-th window;

[0022] P i,max = max(P (i-1)×m+1 , P (i-1)×m+2 , …, P i×m )

[0023] where, P i,max is the maximum value of the wind power data in the i-th window, max(·) indicates taking the maximum value;

[0024] P i,min = min(P (i-1)×m+1 , P (i-1)×m+2 , …, P i×m )

[0025] where, P i,min is the minimum value of the wind power data in the i-th window, min(·) indicates taking the minimum value.

[0026] Further, the converting the power time-domain function into a frequency-domain function is specifically:

[0027] The power time domain function is converted into a frequency domain function by Fourier transform;

[0028]

[0029] In the formula, f(t) is a power time domain function, F(ω) is a frequency domain function, ω is a frequency variable, j is an imaginary unit, and t is a time.

[0030] Further, the frequency domain characteristic values include a center of gravity frequency, a root mean square frequency, and a frequency standard deviation, and the calculation of the frequency domain characteristic values in each window is specifically:

[0031]

[0032] In the formula, FC i is the center of gravity frequency of the frequency data in the i-th window, ω i is the frequency of the frequency data in the i-th window, S(ω i ) is the amplitude of the frequency data in the i-th window, and ω is a frequency variable.

[0033]

[0034] In the formula, RMSF i is the root mean square frequency of the frequency data in the i-th window.

[0035]

[0036] In the formula, RVF i is the frequency standard deviation of the frequency data in the i-th window.

[0037] Further, the structure of the power prediction model includes:

[0038] A time-frequency feature branch module that adopts a bidirectional LSTM layer to capture time sequence dynamic correlation;

[0039] A meteorological branch module that extracts a nonlinear relationship through a fully connected layer;

[0040] A feature fusion module that weightedly fuses the outputs of the time-frequency feature branch and the meteorological branch through an attention mechanism;

[0041] An output layer that outputs a predicted power value based on the fused features;

[0042] The input of the time domain characteristic values, the frequency domain characteristic values, and the historical weather forecast data into the power prediction model is specifically that the time domain characteristic values and the frequency domain characteristic values are input into the time-frequency feature branch module, and the historical weather forecast data is input into the meteorological branch module.

[0043] The application further provides a wind farm power prediction system based on adaptive window and time-frequency feature fusion, comprising:

[0044] A training sample collection module is configured to collect historical wind power data sets, set a window width value, select a sampling range, calculate time domain feature values in each window, convert the power time domain function into a frequency domain function, and calculate frequency domain feature values in each window.

[0045] A model training module is configured to input the time domain feature values, the frequency domain feature values and historical weather forecast data into a power prediction model to perform model training.

[0046] A prediction module is configured to input future weather forecast data into the trained power prediction model to predict future wind farm power.

[0047] The application further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the wind farm power prediction method based on adaptive window and time-frequency feature fusion as described above when executing the computer program.

[0048] The application further provides a computer readable storage medium storing a computer program, wherein the computer program enables a computer to execute the wind farm power prediction method based on adaptive window and time-frequency feature fusion as described above.

[0049] The application has the following beneficial effects: the application mines a series of time-frequency domain indexes from historical wind power time series data, uses adaptive windows to improve the flexibility of feature extraction, and adapts to non-stationary data; and uses an improved neural network model to fuse time-frequency domain features and meteorological data, thereby significantly improving prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The application provides a wind farm power prediction method based on adaptive window and time-frequency feature fusion.

[0051] Figure 2 The application provides a structure diagram of a power prediction model.

[0052] Figure 3 The application provides a power prediction result curve diagram. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, but not all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0054] Embodiment one

[0055] The application provides a wind farm power prediction method based on adaptive window and time-frequency feature fusion, and a flowchart of the method is shown in the figure Figure 1 The method comprises the following steps:

[0056] A historical wind power data set is collected to obtain a power time domain function, and the expression of the power time domain function is as follows:

[0057] f(t) = (P1, P2, …, P t )

[0058] In the formula, f(t) is the power time domain function, P t is the historical wind power at t.

[0059] For example, the historical wind power data of a wind farm from January 1, 2024 to June 1, 2024 is collected with a time interval of 15 minutes.

[0060] A window width value is set to select a sampling range, and the window width value is specifically as follows:

[0061] The initial window value m is set to 96, the window width value m is dynamically adjusted according to a local data variation coefficient, the local data variation coefficient is the standard deviation divided by the mean value of the wind power data in the window, and when the local data variation coefficient is greater than or equal to a preset threshold value, the window width value is reduced to Enhance the capture of high-frequency features. Symbol represents rounding down, and the window width value represents the number of data in the window.

[0062] Time domain feature values in each window are calculated, the time domain feature values include a mean value, a variance, a maximum value and a minimum value, the time domain feature values in each window are calculated, and the formula is as follows:

[0063]

[0064] In the formula, P is the mean value of the wind power data in the i th window, P (i-1)×m+1 represents the first wind power data in the i th window, P i×m represents the m th wind power data in the i th window, and m represents the window width value.

[0065]

[0066] In the formula, s i is the variance of the wind power data in the i th window.

[0067] P i,max = max(P (i-1)×m+1,P (i-1)×m+2 ,…,P i×m )

[0068] In the formula, P i,max This represents the maximum value of wind power data within the i-th window, where max(·) indicates taking the maximum value.

[0069] P i,min =min(P (i-1)×m+1 ,P (i-1)×m+2 ,…,P i×m )

[0070] In the formula, P i,min Let min(·) be the minimum value of wind power data within the i-th window.

[0071] Using Fourier transform, the power time-domain function is transformed into a frequency-domain function:

[0072]

[0073] In the formula, f(t) is the power time-domain function, F(ω) is the frequency-domain function, ω is the frequency variable, j is the imaginary unit, and t is the time.

[0074] Calculate the frequency domain eigenvalues ​​within each window. These eigenvalues ​​include the centroid frequency, root mean square frequency, and frequency standard deviation. The specific steps for calculating the frequency domain eigenvalues ​​within each window are as follows:

[0075]

[0076] In the formula, FC i ω is the centroid frequency of the frequency data within the i-th window. i S(ω) is the frequency of the frequency data within the i-th window. i ) represents the amplitude of the frequency data within the i-th window, and v is the frequency variable;

[0077]

[0078] In the formula, RMSF i It is the root mean square frequency of the frequency data within the i-th window;

[0079]

[0080] In the formula, RVF i It is the standard deviation of the frequency data within the i-th window.

[0081] Time-domain feature values, frequency-domain feature values, and historical weather forecast data are input into the power prediction model for model training; the structure of the power prediction model is as follows: Figure 2As shown, it comprises a time-frequency feature branch module, a weather branch module, a feature fusion module and an output layer.

[0082] The time domain feature and the frequency domain feature data enter the time-frequency feature branch module, adopts a bidirectional LSTM layer to capture the time sequence dynamic correlation;The historical weather forecast data (such as wind speed, wind direction, temperature, air pressure, etc.) enters the weather branch module, and the nonlinear relationship is extracted through the full connection layer;The feature fusion module fuses the outputs of the two branches by the attention mechanism, and the output layer outputs the predicted power value.

[0083] The future weather forecast data is input into the trained power prediction model to predict the future wind farm power. Figure 3 As shown.

[0084] Embodiment two

[0085] The application proposes a wind farm power prediction system based on adaptive window and time-frequency feature fusion corresponding to the method of embodiment one, comprising:

[0086] The training sample acquisition module is used to collect historical wind power data set power time domain function, set window width value, used to select sampling range, calculate time domain feature value in each window, convert power time domain function into frequency domain function, and calculate frequency domain feature value in each window;

[0087] The model training module is used to input the time domain feature value, the frequency domain feature value and the historical weather forecast data into the power prediction model for model training.

[0088] The prediction module is used to input the future weather forecast data into the trained power prediction model to predict the future wind farm power.

[0089] The implementation mode of each module and the function of the module in the system is completely consistent with each step of the method of embodiment one, so it will not be repeated here.

[0090] Embodiment three

[0091] The application proposes an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the wind farm power prediction method based on adaptive window and time-frequency feature fusion as described in embodiment one is realized.

[0092] Embodiment four

[0093] The application proposes a computer readable storage medium, which stores a computer program, and the computer program enables the computer to execute the wind farm power prediction method based on adaptive window and time-frequency feature fusion as described in embodiment one.

[0094] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium which can contain or store programs for use by or in connection with an instruction execution system, apparatus or device. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of computer storage medium can include one or more wires, portable computer disks, hard drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), fiber optics, compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0095] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solutions. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0096] The above is only the preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application. It should be noted that, for ordinary skilled persons in the art, some improvements and refinements without departing from the principles of the present application shall be considered within the protection scope of the present application.

Claims

1. A wind farm power prediction method based on adaptive window and time-frequency feature fusion, characterized in that, The method comprises the following steps: a historical wind power data set is collected to obtain a power time domain function, a window width value is set to select a sampling range, and time domain characteristic values in each window are calculated, the power time domain function is converted into a frequency domain function, and frequency domain characteristic values in each window are calculated, the time domain characteristic values, the frequency domain characteristic values and historical weather forecast data are input into a power prediction model for model training; future weather forecast data is input into the trained power prediction model to predict future wind farm power.

2. The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, characterized in that, The expression of the power time domain function is: f(t) = (P1, P2,..., P t ) In the formula, f(t) is a power time domain function, P t is the historical wind power at time t. 3.The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, wherein, The window width value is set as: According to the local data variation coefficient, the window width value m is dynamically adjusted. The local data variation coefficient is the standard deviation of the wind power data in the window divided by the mean value. When the local data variation coefficient is greater than or equal to a preset threshold value, the window width value is reduced to Symbol represents rounding down, and the window width value represents the number of data in the window.

4. The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, characterized in that, The time domain characteristic values include mean value, variance, maximum value and minimum value, and the time domain characteristic values in each window are calculated according to the following formula: In the formula, P is the average value of the wind power data in the i-th window, (i-1)×m+1 P represents the 1st wind power data in the i-th window, i×m P represents the mth wind power data in the i-th window, and m represents the window width value. In the formula, s i is the variance of the wind power data in the i-th window. P i,max = max(P (i-1)×m+1 , P (i-1)×m+2 ,..., P i×m ) In the formula, P i,max is the maximum value of the wind power data in the ith window, and max( ) represents taking the maximum value. P i,min = min(P (i-1)×m+1 ,P (i-1)×m+2 ,…,P i×m ) In the formula, P i,min is the minimum value of the wind power data in the ith window, and min( ) represents taking the minimum value.

5. The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, characterized in that, The power time domain function is converted into a frequency domain function as follows: The power time domain function is converted into a frequency domain function by using Fourier transform. In the formula, f(t) is the power time domain function, F(ω) is the frequency domain function, ω is the frequency variable, j is the imaginary unit, and t is the time.

6. The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, characterized in that, The frequency domain characteristic values include center of gravity frequency, root mean square frequency and frequency standard deviation, and the frequency domain characteristic values in each window are calculated as follows: where FC i is the center frequency of the frequency data within the ith window, ω i is the frequency of the frequency data within the ith window, S(ω i ) is the amplitude of the frequency data within the ith window, and ω is the frequency variable. where RMSF is the root mean square frequency of the frequency data within the ith window; and i is the root mean square frequency of the frequency data within the ith window; and where RVF i is the frequency standard deviation of the frequency data within the ith window.

7. The wind farm power prediction method based on adaptive window and time-frequency feature fusion according to claim 1, characterized in that, The structure of the power prediction model comprises: a time-frequency characteristic branch module that adopts a bidirectional LSTM layer to capture time sequence dynamic correlation; a meteorological branch module that extracts a nonlinear relationship through a fully connected layer; a feature fusion module that fuses the outputs of the time-frequency characteristic branch and the meteorological branch through an attention mechanism; an output layer that outputs a predicted power value based on the fused features. The time domain characteristic values and the frequency domain characteristic values are input into the time-frequency characteristic branch module, and the historical weather forecast data is input into the meteorological branch module.

8. A wind farm power prediction system based on adaptive window and time-frequency feature fusion, characterized in that, The method comprises the following steps: a training sample collection module that collects historical wind power data sets to obtain power time domain functions, sets window width values to select sampling ranges, calculates time domain characteristic values in each window, converts the power time domain functions into frequency domain functions, and calculates frequency domain characteristic values in each window; a model training module that inputs the time domain characteristic values, the frequency domain characteristic values and historical weather forecast data into a power prediction model for model training; a prediction module that inputs future weather forecast data into the trained power prediction model to predict future wind farm power.

9. An electronic device, comprising: The method comprises the following steps: a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the wind farm power prediction method based on adaptive window and time-frequency feature fusion according to any one of claims 1-7 when executing the computer program.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program enables the computer to implement the wind farm power prediction method based on adaptive window and time-frequency feature fusion according to any one of claims 1-7.

Images