A wind power prediction method and device

Abstract

The application provides a wind power prediction method and device, obtains a predicted wind speed with a prediction time resolution; inputs the predicted wind speed with the prediction time resolution into a pre-constructed target time resolution wind speed ultra-short-term prediction model to obtain a target time resolution predicted wind speed; inputs the target time resolution predicted wind speed into a pre-constructed wind power prediction model to obtain a wind power prediction value corresponding to the target time resolution predicted wind speed; the prediction time resolution is greater than the target time resolution; the target time resolution wind speed ultra-short-term prediction model determines the target time resolution predicted wind speed based on a distribution characteristic similarity relationship between a target time resolution measured wind speed and the prediction time resolution predicted wind speed, and couples the target time resolution measured wind speed and the prediction time resolution predicted wind speed, thereby improving the target time resolution of the wind power prediction and the application range of the wind power prediction value.

Description

Technical Field

[0001] This invention relates to the field of new energy technology, specifically to a method and apparatus for predicting wind power output. Background Technology

[0002] With the increasing utilization rate of new energy sources, the cumulative installed capacity of wind power is growing. To enable wind power to play a vital role in power supply, environmental protection, and low-carbon emission reduction, wind power forecasting has become an important means of integrating wind power into dispatch operations. Currently, the time resolution of wind power forecasting is generally 15 minutes or more. Based on this, methods with applicability, robustness, and linear extrapolation are used to improve the time resolution of wind power forecasting. However, linear extrapolation methods require the assumption of slow weather changes, and their wind power forecasting error increases with the forecast lead time, especially when there are sudden changes in wind power system structure or weather. Moreover, there is no upper limit to the wind power forecasting error. Furthermore, the time resolution of numerical weather prediction in the aforementioned existing technologies is 15 minutes or more. Due to atmospheric dynamic and thermodynamic imbalances within 15 minutes, wind speed undergoes nonlinear changes, and numerical weather prediction cannot reflect rapid changes in wind speed within 15 minutes. Therefore, the current time resolution of wind power forecasting is low. Summary of the Invention

[0003] To overcome the shortcomings of large prediction errors and low resolution in the existing technology, the present invention provides a wind power prediction method, comprising:

[0004] Obtain the forecast wind speed with a time resolution;

[0005] The predicted wind speed at the predicted time resolution is input into the pre-constructed ultra-short-term wind speed prediction model at the target time resolution to obtain the predicted wind speed at the target time resolution.

[0006] The predicted wind speed at the target time resolution is substituted into the pre-built wind power prediction model for solution, and the predicted wind power value corresponding to the predicted wind speed at the target time resolution is obtained.

[0007] Wherein, the forecast time resolution is greater than the target time resolution; the target time resolution wind speed ultra-short-term prediction model determines the predicted wind speed at the target time resolution based on the similarity of the distribution characteristics of the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution.

[0008] The construction of the target time resolution wind speed ultra-short-term prediction model includes:

[0009] Based on the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution, the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the forecast wind speed are calculated.

[0010] Determine the distribution location parameters of the measured wind speed and the distribution location parameters of the predicted wind speed, and based on the distribution location parameters of the measured wind speed and the predicted wind speed at the target time resolution, determine the distribution similarity parameters of the measured wind speed and the predicted wind speed at the predicted time resolution.

[0011] The ultra-short-term wind speed prediction model for the target time resolution is constructed based on the shape correlation coefficients of the dynamic optimization functions of the measured wind speed and the dynamic optimization functions of the predicted wind speed, as well as the distribution similarity parameters of the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution.

[0012] The shape correlation coefficients of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed are determined by the following formula:

[0013]

[0014] In the formula, The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed. The difference of the dynamic optimization function corresponding to the predicted wind speed. The difference between the dynamic optimization function corresponding to the measured wind speed. This is the minimum absolute error between the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. ω is the maximum absolute error of the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. The resolution coefficient.

[0015] The and Determine by the following formula:

[0016]

[0017]

[0018] In the formula, Let be the dynamic optimization function corresponding to the measured wind speed at time t. Let be the dynamic optimization function corresponding to the predicted wind speed at time t. Let be the dynamic optimization function corresponding to the measured wind speed at time t+1. This is the dynamic optimization function for the predicted wind speed at time t+1.

[0019] The Determine by the following formula:

[0020]

[0021] In the formula, k is the time interval index, s is the total number of measured wind speed points in the k-th time interval, and x(k+m) represents the measured wind speed at the m-th measured wind speed point in the k-th time interval under the target time resolution.

[0022] The Determine by the following formula:

[0023]

[0024] In the formula, r is the total number of forecast wind speed points in the k-th time interval, and y(k+n) represents the forecast wind speed at the n-th forecast wind speed point in the k-th time interval under the forecast time resolution.

[0025] The determination of the distribution location parameters of the measured wind speed includes:

[0026] Obtain the measured wind speed at the target time resolution, and determine the horizontal coordinate position corresponding to the measured wind speed at the target time resolution based on the measured wind speed at the target time resolution.

[0027] Determine the first mapping relationship between the measured wind speed at the target time resolution and the corresponding abscissa position of the measured wind speed at the target time resolution;

[0028] Based on the first mapping relationship, the measured wind speed at the target time resolution and the corresponding abscissa position are fitted to obtain the distribution position parameters of the measured wind speed.

[0029] The first mapping relationship is determined by the following formula:

[0030] x = a x +b x λ x +σ x

[0031] In the formula, x is the measured wind speed at the target time resolution, and a x b x and σ x Let λ be the location parameter for the distribution of the measured wind speed. x The x-coordinate position corresponding to the measured wind speed at the target time resolution.

[0032] The determination of the location parameters for the forecast wind speed distribution includes:

[0033] Obtain the forecast wind speed at the forecast time resolution, and determine the horizontal coordinate position corresponding to the forecast wind speed at the forecast time resolution based on the forecast wind speed at the forecast time resolution.

[0034] Determine a second mapping relationship between the forecast wind speed at the forecast time resolution and the corresponding abscissa position of the forecast wind speed at the forecast time resolution;

[0035] Based on the second mapping relationship, the predicted wind speed at the predicted time resolution and the corresponding abscissa position are fitted to obtain the distribution position parameters of the predicted wind speed.

[0036] The second mapping relationship is determined by the following formula:

[0037] y = a y +b y λ y +σ y

[0038] In the formula, y is the forecast wind speed at the forecast time resolution, and a y b y σ y Let λ be the location parameter for the predicted wind speed distribution. y The x-coordinate position corresponding to the forecast wind speed at the forecast time resolution is given.

[0039] The determination of distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution, based on the distribution location parameters of the measured wind speed and the distribution location parameters of the predicted wind speed, includes:

[0040] The probability density function of the measured wind speed is determined based on the location parameters of the measured wind speed distribution, and the probability density function of the predicted wind speed is determined based on the location parameters of the predicted wind speed distribution.

[0041] Based on the probability density function of the measured wind speed, the probability density function of the predicted wind speed, and the similarity relationship of the distribution characteristics between the probability density functions of the measured wind speed and the predicted wind speed at the target time resolution, the distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution are determined.

[0042] The similarity relationship of the distribution characteristics between the probability density function of the measured wind speed and the probability density function of the predicted wind speed is determined by the following formula:

[0043]

[0044] In the formula, μ(x|a x b x ;σ x 2 Let g(y|a) be the probability density function of the measured wind speed. y b y ;σ y2 Let be the probability density function of the predicted wind speed, z be the index of the predicted member, Z be the total number of predicted members, and w be the probability density function of the predicted wind speed. z Let be the distribution similarity parameter between the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution for the z-th forecast member.

[0045] The target time-resolution wind speed ultra-short-term prediction model is determined by the following formula:

[0046]

[0047] In the formula, y target Forecast wind speed at target time resolution. The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed.

[0048] The wind power prediction model is constructed using the following formula:

[0049] wp=net{∑ρ 3 *y target +ε}

[0050] In the formula, wp is the wind power prediction value at the target time resolution, net is the activation function of the neural network, ρ is the air density, and ε is the maximum error between the wind power prediction value at the target time resolution and the preset wind power value.

[0051] The step of inputting the predicted wind speed at the target time resolution into the pre-built wind power prediction model for solution includes:

[0052] Based on the predicted wind speed at the target time resolution, a deep learning algorithm is used to solve the wind power prediction model.

[0053] On the other hand, the present invention also provides a wind power prediction device, comprising:

[0054] The acquisition module is used to acquire the forecast wind speed with a forecast time resolution;

[0055] The determination module is used to input the forecast wind speed at the forecast time resolution into the pre-constructed ultra-short-term wind speed prediction model at the target time resolution to obtain the forecast wind speed at the target time resolution.

[0056] The solution module is used to input the predicted wind speed at the target time resolution into the pre-built wind power prediction model for solution, and obtain the predicted wind power value corresponding to the predicted wind speed at the target time resolution.

[0057] Wherein, the forecast time resolution is greater than the target time resolution; the target time resolution wind speed ultra-short-term prediction model determines the predicted wind speed at the target time resolution based on the similarity of the distribution characteristics of the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution.

[0058] The device further includes a first construction module, the first construction module comprising:

[0059] The first determining unit is used to calculate the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the forecast wind speed based on the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution.

[0060] The second determining unit is used to determine the distribution location parameters of the measured wind speed and the distribution location parameters of the predicted wind speed, and to determine the distribution similarity parameters of the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution based on the distribution location parameters of the measured wind speed and the distribution location parameters of the predicted wind speed.

[0061] The third determining unit is used to construct the ultra-short-term prediction model for wind speed at the target time resolution based on the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed, the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed, and the distribution similarity parameters of the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution.

[0062] The third determining unit determines the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed according to the following formula:

[0063]

[0064] In the formula, The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed. The difference of the dynamic optimization function corresponding to the predicted wind speed. The difference between the dynamic optimization function corresponding to the measured wind speed. This is the minimum absolute error between the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. ω is the maximum absolute error of the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. The resolution coefficient;

[0065] The and Determine by the following formula:

[0066]

[0067]

[0068] In the formula, Let be the dynamic optimization function corresponding to the measured wind speed at time t. Let be the dynamic optimization function corresponding to the predicted wind speed at time t. Let be the dynamic optimization function for the measured wind speed at time t+1. This is a dynamic optimization function for predicting wind speed at time t+1.

[0069] The third determining unit is determined by the following formula.

[0070]

[0071] In the formula, k is the time interval index, s is the total number of measured wind speed points in the k-th time interval, and x(k+m) represents the measured wind speed at the m-th measured wind speed point in the k-th time interval under the target time resolution.

[0072] The third determining unit is determined by the following formula.

[0073]

[0074] In the formula, r is the total number of forecast wind speed points in the k-th time interval, and y(k+n) represents the forecast wind speed at the n-th forecast wind speed point in the k-th time interval under the forecast time resolution.

[0075] The second determining unit is specifically used for:

[0076] Obtain the measured wind speed at the target time resolution, and determine the horizontal coordinate position corresponding to the measured wind speed at the target time resolution based on the measured wind speed at the target time resolution.

[0077] Determine the first mapping relationship between the measured wind speed at the target time resolution and the corresponding abscissa position of the measured wind speed at the target time resolution;

[0078] Based on the first mapping relationship, the measured wind speed at the target time resolution and the corresponding abscissa position are fitted to obtain the distribution position parameters of the measured wind speed.

[0079] The second determining unit determines the first mapping relationship according to the following formula:

[0080] x = a x +b x λ x +σ x

[0081] In the formula, x is the measured wind speed at the target time resolution, and a x b x and σ x Let λ be the location parameter for the distribution of the measured wind speed. x The x-coordinate position corresponding to the measured wind speed at the target time resolution.

[0082] The second determining unit is specifically used for:

[0083] Obtain the forecast wind speed at the forecast time resolution, and determine the horizontal coordinate position corresponding to the forecast wind speed at the forecast time resolution based on the forecast wind speed at the forecast time resolution.

[0084] Determine a second mapping relationship between the forecast wind speed at the forecast time resolution and the corresponding abscissa position of the forecast wind speed at the forecast time resolution;

[0085] Based on the second mapping relationship, the predicted wind speed at the predicted time resolution and the corresponding abscissa position are fitted to obtain the distribution position parameters of the predicted wind speed.

[0086] The second determining unit determines the second mapping relationship according to the following formula:

[0087] y = a y +b y λ y +σ y

[0088] In the formula, y is the forecast wind speed at the forecast time resolution, and a y b y σ y Let λ be the location parameter for the predicted wind speed distribution. y The x-coordinate position corresponding to the forecast wind speed at the forecast time resolution is given.

[0089] The second determining unit is specifically used for:

[0090] The probability density function of the measured wind speed is determined based on the location parameters of the measured wind speed distribution, and the probability density function of the predicted wind speed is determined based on the location parameters of the predicted wind speed distribution.

[0091] Based on the probability density function of the measured wind speed, the probability density function of the predicted wind speed, and the similarity relationship of the distribution characteristics between the probability density functions of the measured wind speed and the predicted wind speed at the target time resolution, the distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution are determined.

[0092] The second determining unit determines the similarity relationship of distribution characteristics between the probability density function of the measured wind speed and the probability density function of the predicted wind speed using the following formula:

[0093]

[0094] In the formula, μ(x|a x b x ;σ x 2 Let g(y|a) be the probability density function of the measured wind speed. y b y ;σ y 2 Let be the probability density function of the predicted wind speed, z be the index of the predicted member, Z be the total number of predicted members, and w be the probability density function of the predicted wind speed. z Let a be the distribution similarity parameter between the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution for the z-th forecast member. x b x and σ x Let a be the location parameter for the distribution of the measured wind speed. y b y σ y Here, x represents the distribution location parameter of the predicted wind speed, y represents the measured wind speed at the target time resolution, and y represents the predicted wind speed at the predicted time resolution.

[0095] The third unit constructs the target time-resolution wind speed ultra-short-term prediction model using the following formula:

[0096]

[0097] In the formula, y target Forecast wind speed at target time resolution. The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed.

[0098] The wind power prediction device provided in this application embodiment further includes a second construction module, which constructs a wind power prediction model according to the following formula:

[0099] wp=net{∑ρ 3 *y target +ε}

[0100] In the formula, wp is the wind power prediction value at the target time resolution, net is the activation function of the neural network, ρ is the air density, and ε is the maximum error between the wind power prediction value at the target time resolution and the preset wind power value.

[0101] The solution module is specifically used for:

[0102] Based on the predicted wind speed with target time resolution, a deep learning algorithm is used to solve the wind power prediction model.

[0103] The technical solution provided by this invention has the following beneficial effects:

[0104] The wind power prediction method provided in this application obtains the forecast wind speed at the forecast time resolution; inputs the forecast wind speed at the forecast time resolution into a pre-constructed ultra-short-term wind speed prediction model at the target time resolution to obtain the predicted wind speed at the target time resolution; substitutes the predicted wind speed at the target time resolution into the pre-constructed wind power prediction model for solving to obtain the wind power prediction value corresponding to the predicted wind speed at the target time resolution; the forecast time resolution is greater than the target time resolution; the ultra-short-term wind speed prediction model at the target time resolution determines the predicted wind speed at the target time resolution based on the similarity of the distribution characteristics of the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution, thus coupling the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution to improve the target time resolution of wind power prediction;

[0105] The parameters in the wind power prediction model of this application have strong stability and practicality, and can better simulate the changes in wind speed extreme points, providing a basis for effectively improving the time resolution of wind power prediction.

[0106] In the process of constructing the wind power prediction model, this application fully couples the measured wind speed and the predicted wind speed, realizes wind power prediction with the target time resolution, and improves the application range of the wind power prediction value.

[0107] This application obtains distribution similarity parameters between measured wind speed and predicted wind speed based on the probability density function of measured wind speed, the probability density function of predicted wind speed, and the similarity relationship of the distribution characteristics between the probability density functions of measured wind speed and predicted wind speed at the target time resolution. This allows for a more comprehensive and in-depth analysis of measured and predicted wind speeds, providing a reliable data foundation for wind power prediction. Attached Figure Description

[0108] Figure 1 This is a flowchart of the wind power prediction method in the embodiments of this application;

[0109] Figure 2 This is a structural diagram of the wind power prediction device in the embodiments of this application. Detailed Implementation

[0110] The present application will now be described in further detail with reference to the accompanying drawings.

[0111] Example 1

[0112] Embodiment 1 of the present invention provides a wind power prediction method, the specific flowchart of which is as follows: Figure 1 As shown, the specific process includes:

[0113] S101: Obtain the forecast wind speed with a time resolution;

[0114] S102: Input the forecast wind speed at the forecast time resolution into the pre-built ultra-short-term wind speed prediction model at the target time resolution to obtain the predicted wind speed at the target time resolution.

[0115] S103: Substitute the predicted wind speed at the target time resolution into the pre-built wind power prediction model for solution to obtain the predicted wind power value corresponding to the predicted wind speed at the target time resolution.

[0116] Among them, the forecast time resolution is greater than the target time resolution; the ultra-short-term wind speed prediction model at the target time resolution determines the predicted wind speed at the target time resolution based on the similarity of the distribution characteristics of the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution.

[0117] In this embodiment of the application, the forecast time resolution is 15 minutes and the target time resolution is 5 minutes.

[0118] The specific process for constructing the ultra-short-term wind speed prediction model with target time resolution is as follows:

[0119] Based on the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution, the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the forecast wind speed are calculated.

[0120] Determine the location parameters of the measured wind speed and the location parameters of the forecast wind speed, and based on the location parameters of the measured wind speed and the location parameters of the forecast wind speed, determine the similarity parameters of the distribution of the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution.

[0121] A target time-resolution wind speed ultra-short-term prediction model is constructed based on the shape correlation coefficients of the dynamic optimization functions of measured wind speed and forecast wind speed, as well as the distribution similarity parameters of the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution.

[0122] The shape correlation coefficient between the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed is determined by the following formula:

[0123]

[0124] In the formula, The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed. The difference of the dynamic optimization function corresponding to the predicted wind speed. The difference between the dynamic optimization function corresponding to the measured wind speed. This is the minimum absolute error between the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. ω is the maximum absolute error of the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. The resolution coefficient.

[0125] Among them and Determine by the following formula:

[0126]

[0127]

[0128] In the formula, Let be the dynamic optimization function corresponding to the measured wind speed at time t. Let be the dynamic optimization function corresponding to the predicted wind speed at time t. Let be the dynamic optimization function corresponding to the measured wind speed at time t+1. This is the dynamic optimization function for the predicted wind speed at time t+1.

[0129] Determine by the following formula:

[0130]

[0131] In the formula, k is the time interval index, s is the total number of measured wind speed points in the k-th time interval, and x(k+m) represents the measured wind speed at the m-th measured wind speed point in the k-th time interval under the target time resolution.

[0132] Determine by the following formula:

[0133]

[0134] In the formula, r is the total number of forecast wind speed points in the k-th time interval, and y(k+n) represents the forecast wind speed at the n-th forecast wind speed point in the k-th time interval under the forecast time resolution.

[0135] The determination of the location parameters of the measured wind speed distribution includes:

[0136] Obtain the measured wind speed at the target time resolution, and determine the horizontal coordinate position corresponding to the measured wind speed at the target time resolution based on the measured wind speed at the target time resolution.

[0137] Determine the first mapping relationship between the measured wind speed at the target time resolution and the corresponding abscissa position of the measured wind speed at the target time resolution;

[0138] Based on the first mapping relationship, the measured wind speed at the target time resolution and the corresponding abscissa position are fitted to obtain the distribution location parameters of the measured wind speed. In this embodiment, the method for fitting the measured wind speed at the target time resolution and the corresponding abscissa position can be the least squares method. Of course, other fitting methods can also be used to fit the measured wind speed at the target time resolution and the corresponding abscissa position.

[0139] The first mapping relationship is determined by the following formula:

[0140] x = a x +b x λ x +σ x

[0141] In the formula, x is the measured wind speed at the target time resolution, and a x b x and σ x For the location parameters of the measured wind speed, more specifically, a x It is the intercept of a linear function (i.e., the first mapping relationship) of the measured wind speed and location information with respect to the target time resolution, b x It is the slope of a linear function (i.e., the first mapping relationship) of the measured wind speed and location information with respect to the target time resolution, σ x It is the fitted distribution parameter of the measured wind speed; λ x The x-coordinate position corresponds to the measured wind speed at the target time resolution.

[0142] Determining the location parameters for the forecast wind speed distribution includes:

[0143] Obtain the forecast wind speed at the forecast time resolution, and determine the horizontal coordinate position corresponding to the forecast wind speed at the forecast time resolution based on the forecast wind speed at the forecast time resolution.

[0144] Determine the second mapping relationship between the forecast wind speed at the forecast time resolution and the corresponding abscissa position of the forecast wind speed at the forecast time resolution;

[0145] Based on the second mapping relationship, the predicted wind speed at the predicted time resolution and its corresponding abscissa position are fitted to obtain the distribution location parameters of the predicted wind speed. In this embodiment, the method for fitting the predicted wind speed at the predicted time resolution and its corresponding abscissa position can be the least squares method. Of course, other fitting methods can also be used to fit the predicted wind speed at the predicted time resolution and its corresponding abscissa position.

[0146] The second mapping relationship is determined by the following formula:

[0147] y = a y +b y λ y +σ y

[0148] In the formula, y represents the forecast wind speed at the forecast time resolution, and a y b y σ y To predict the location parameters of wind speed distribution, more specifically, a y It is the intercept of the linear function (i.e., the second mapping relationship) of the forecast wind speed and location information, b y It is the slope of a linear function relating to the predicted wind speed and location information, σ y It is the fitted distribution parameter of the forecast wind speed; λ y The x-axis position corresponds to the forecast wind speed at the forecast time resolution.

[0149] Based on the location parameters of the measured wind speed distribution and the location parameters of the predicted wind speed distribution, distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution are determined, including:

[0150] The probability density function of the measured wind speed is determined based on the location parameters of the measured wind speed distribution, and the probability density function of the predicted wind speed is determined based on the location parameters of the predicted wind speed distribution.

[0151] Based on the probability density functions of measured wind speed, predicted wind speed, and the similarity relationship of their distribution characteristics, the distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution are determined.

[0152] In this embodiment, the probability density function of measured wind speed and the probability density function of predicted wind speed are used as the basis. Data mining methods are used to obtain the similarity of distribution characteristics between the probability density function of predicted wind speed and the probability density function of measured wind speed and predicted wind speed. This provides a more comprehensive and in-depth analysis of measured wind speed and predicted wind speed, and provides a reliable data foundation for wind power prediction.

[0153] Correlation matching applied to wind speed data enables more comprehensive and in-depth analysis of wind speed data, improving the quality of wind speed forecast data;

[0154] The similarity relationship of the distribution characteristics between the probability density function of the measured wind speed and the probability density function of the predicted wind speed is determined by the following formula:

[0155]

[0156] In the formula, μ(x|a x b x ;σ x 2 Let g(y|a) be the probability density function of the measured wind speed. y b y ;σ y 2 Let be the probability density function of the forecast wind speed, z be the forecast member index, Z be the total number of forecast members, and w be the forecast wind speed. z Let be the distribution similarity parameter between the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution for the z-th forecast member.

[0157] The target time resolution wind speed ultra-short-term prediction model is determined by the following formula:

[0158]

[0159] In the formula, y target Forecast wind speed at target time resolution. The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed.

[0160] The wind power prediction model is constructed using the following formula:

[0161] wp=net{∑ρ 3 *y target +ε}

[0162] In the formula, wp is the wind power prediction value at the target time resolution, net is the activation function of the neural network, ρ is the air density, and ε is the maximum error between the wind power prediction value at the target time resolution and the preset wind power value. In the embodiments of this application, ε is less than or equal to 0.05, and the unit is 100%.

[0163] In this embodiment, net can be the sigmoid activation function of the deep learning algorithm in the neural network, or other activation functions in the neural network.

[0164] In the process of constructing the wind power prediction model, the embodiments of this application fully couple the measured wind speed and the predicted wind speed, thereby improving the application scope of the wind power prediction value.

[0165] On the other hand, the parameters of the wind power prediction model (including a) x b x σ x w z It has strong stability and practicality, and can better simulate the changes in wind speed extreme points, effectively improving the time resolution of wind power prediction.

[0166] The predicted wind speed at the target time resolution is input into a pre-built wind power prediction model for solution, including:

[0167] Based on the predicted wind speed with target time resolution, a deep learning algorithm is used to solve the wind power prediction model.

[0168] In this embodiment of the application, after obtaining the wind power prediction value at the target time resolution, the wind power prediction accuracy can be calculated using the following formula:

[0169]

[0170] In the formula, RMSE represents the wind power prediction accuracy, and wp i wp is the predicted wind power value for the i-th time resolution. i-act Let be the measured wind power value of the i-th time resolution, and h be the total number of measured or predicted wind speeds at the target time resolution. It should be noted that the total number of measured wind speeds at the target time resolution is equal to the total number of predicted wind speeds at the target time resolution.

[0171] The target time resolution of the wind power prediction value obtained by the wind power prediction method provided in this application embodiment is higher than that obtained by the existing linear extrapolation method. The target time resolution of the wind power prediction value obtained by this application embodiment can reach 5 minutes. By increasing the target time resolution, the wind power prediction error can be greatly reduced, and the wind power error will not increase with the increase of prediction time. Even if there are sudden changes in wind power system structure or weather (i.e., different climates), the wind power prediction error will not increase significantly. In addition, since the prediction method provided in this application embodiment can achieve a target time resolution of 5 minutes, even if there is an imbalance in atmospheric dynamics and thermodynamics within 15 minutes and the wind speed undergoes a nonlinear change process, the wind power prediction method provided in this application embodiment can still reflect the rapid change of wind speed within 15 minutes, thereby obtaining a more accurate wind power prediction value.

[0172] Example 2

[0173] Embodiment 2 of the present invention provides a wind power prediction device, such as Figure 2 As shown, it specifically includes:

[0174] The acquisition module is used to acquire the forecast wind speed with a forecast time resolution;

[0175] The determination module is used to input the forecast wind speed at the forecast time resolution into the pre-built ultra-short-term wind speed prediction model at the target time resolution to obtain the predicted wind speed at the target time resolution.

[0176] The solution module is used to input the predicted wind speed at the target time resolution into the pre-built wind power prediction model for solution, and obtain the predicted wind power value corresponding to the predicted wind speed at the target time resolution.

[0177] In this embodiment of the application, the solution module specifically uses a deep learning algorithm to solve the wind power prediction model based on the predicted wind speed at the target time resolution.

[0178] The forecast time resolution is greater than the target time resolution; in this embodiment, the forecast time resolution can be 15 minutes and the target time resolution can be 5 minutes.

[0179] The target time resolution wind speed ultra-short-term prediction model determines the predicted wind speed at the target time resolution based on the similarity of the distribution characteristics between the measured wind speed at the target time resolution and the predicted wind speed at the forecast time resolution.

[0180] The wind power prediction device provided in this application embodiment further includes a first construction module, which constructs a target time resolution wind speed ultra-short-term prediction model according to the following process:

[0181] The first determining unit is used to calculate the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the forecast wind speed based on the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution.

[0182] The second determining unit is used to determine the distribution location parameters of the measured wind speed and the distribution location parameters of the forecast wind speed, and to determine the distribution similarity parameters of the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution based on the distribution location parameters of the measured wind speed and the distribution location parameters of the forecast wind speed.

[0183] The third determining unit is used to construct a target time resolution wind speed ultra-short-term prediction model based on the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed, the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed, and the distribution similarity parameters of the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution.

[0184] The third determining unit determines the shape correlation coefficients of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed using the following formula:

[0185]

[0186] In the formula, The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed. The difference of the dynamic optimization function corresponding to the predicted wind speed. The difference between the dynamic optimization function corresponding to the measured wind speed. This is the minimum absolute error between the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. ω is the maximum absolute error of the difference between the dynamic optimization function corresponding to the predicted wind speed and the dynamic optimization function corresponding to the measured wind speed. The resolution coefficient;

[0187] and Determine by the following formula:

[0188]

[0189]

[0190] In the formula, Let be the dynamic optimization function corresponding to the measured wind speed at time t. Let be the dynamic optimization function corresponding to the predicted wind speed at time t. Let be the dynamic optimization function for the measured wind speed at time t+1. This is a dynamic optimization function for predicting wind speed at time t+1.

[0191] The third determining unit is determined by the following formula.

[0192]

[0193] In the formula, k is the time interval index, s is the total number of measured wind speed points in the k-th time interval, and x(k+m) represents the measured wind speed at the m-th measured wind speed point in the k-th time interval under the target time resolution.

[0194] The third determining unit is determined by the following formula.

[0195]

[0196] In the formula, r is the total number of forecast wind speed points in the k-th time interval, and y(k+n) represents the forecast wind speed at the n-th forecast wind speed point in the k-th time interval under the forecast time resolution.

[0197] The second determining unit is specifically used for:

[0198] Obtain the measured wind speed at the target time resolution, and determine the horizontal coordinate position corresponding to the measured wind speed at the target time resolution based on the measured wind speed at the target time resolution.

[0199] Determine the first mapping relationship between the measured wind speed at the target time resolution and the corresponding abscissa position of the measured wind speed at the target time resolution;

[0200] Based on the first mapping relationship, the measured wind speed at the target time resolution and the corresponding abscissa position are fitted to obtain the distribution position parameters of the measured wind speed.

[0201] The second determining unit determines the first mapping relationship according to the following formula:

[0202] x = a x +b x λ x +σ x

[0203] In the formula, x is the measured wind speed at the target time resolution, and a x b x and σ x λ represents the location parameter of the measured wind speed distribution. x The x-coordinate position corresponds to the measured wind speed at the target time resolution.

[0204] The second determining unit is specifically used for:

[0205] Obtain the forecast wind speed at the forecast time resolution, and determine the horizontal coordinate position corresponding to the forecast wind speed at the forecast time resolution based on the forecast wind speed at the forecast time resolution.

[0206] Determine the second mapping relationship between the forecast wind speed at the forecast time resolution and the corresponding abscissa position of the forecast wind speed at the forecast time resolution;

[0207] Based on the second mapping relationship, the predicted wind speed at the forecast time resolution and the corresponding abscissa position are fitted to obtain the distribution location parameters of the predicted wind speed.

[0208] The second determining unit determines the second mapping relationship according to the following formula:

[0209] y = a y +b y λ y +σ y

[0210] In the formula, y represents the forecast wind speed at the forecast time resolution, and a y b y σ y λ is the location parameter for predicting the distribution of wind speed. y The x-axis position corresponds to the forecast wind speed at the forecast time resolution.

[0211] The second determining unit is specifically used for:

[0212] The probability density function of the measured wind speed is determined based on the location parameters of the measured wind speed distribution, and the probability density function of the predicted wind speed is determined based on the location parameters of the predicted wind speed distribution.

[0213] Based on the probability density functions of measured wind speed, predicted wind speed, and the similarity relationship of their distribution characteristics, the distribution similarity parameters between the measured wind speed at the target time resolution and the predicted wind speed at the predicted time resolution are determined.

[0214] The second determining unit determines the similarity relationship of distribution characteristics between the probability density function of the measured wind speed and the probability density function of the predicted wind speed using the following formula:

[0215]

[0216] In the formula, μ(x|a x b x ;σ x 2 Let g(y|a) be the probability density function of the measured wind speed. y b y ;σ y 2 Let be the probability density function of the forecast wind speed, z be the forecast member index, Z be the total number of forecast members, and w be the forecast wind speed. z Let a be the distribution similarity parameter between the measured wind speed at the target time resolution and the forecast wind speed at the forecast time resolution for the z-th forecast member. x b x and σ x For the location parameters of the measured wind speed, a y b y σ y Here, x represents the location parameters for the predicted wind speed, y represents the measured wind speed at the target time resolution, and y represents the predicted wind speed at the predicted time resolution.

[0217] The third unit constructs the target time-resolution wind speed ultra-short-term prediction model using the following formula:

[0218]

[0219] In the formula, y target Forecast wind speed at target time resolution. The shape correlation coefficients are the dynamic optimization functions corresponding to the measured wind speed and the dynamic optimization functions corresponding to the predicted wind speed.

[0220] The wind power prediction device provided in this application embodiment further includes a second construction module, which constructs a wind power prediction model according to the following formula:

[0221] wp=net{∑ρ 3 *y target +ε}

[0222] In the formula, wp is the wind power prediction value at the target time resolution, net is the activation function of the neural network, ρ is the air density, and ε is the maximum error between the wind power prediction value at the target time resolution and the preset wind power value.

[0223] For ease of description, the various parts of the above device are described separately as modules or units based on their functions. Of course, in implementing this application, the functions of each module or unit can be implemented in one or more software or hardware components.

[0224] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0225] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0226] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0227] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0228] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Those skilled in the art can still make modifications or equivalent substitutions to the specific implementation of the present invention by referring to the above embodiments. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention are within the protection scope of the present invention pending approval.

Claims

1. A wind power prediction method, characterized by, The method comprises the following steps: obtaining a predicted wind speed at a prediction time resolution; inputting the predicted wind speed at the prediction time resolution into a pre-constructed target time resolution wind speed ultra-short-term prediction model to obtain a predicted wind speed at a target time resolution; inputting the predicted wind speed at the target time resolution into a pre-constructed wind power prediction model to obtain a predicted wind power value corresponding to the predicted wind speed at the target time resolution; wherein the prediction time resolution is greater than the target time resolution; the target time resolution wind speed ultra-short-term prediction model determines the predicted wind speed at the target time resolution based on a distribution similarity relationship between the measured wind speed at the target time resolution and the predicted wind speed at the prediction time resolution; the construction of the target time resolution wind speed ultra-short-term prediction model comprises the following steps: based on the measured wind speed at the target time resolution and the predicted wind speed at the prediction time resolution, calculating a shape correlation coefficient of a dynamic optimization function corresponding to the measured wind speed and a dynamic optimization function corresponding to the predicted wind speed; determining a distribution position parameter of the measured wind speed and a distribution position parameter of the predicted wind speed, and determining a distribution similarity parameter of the measured wind speed at the target time resolution and the predicted wind speed at the prediction time resolution based on the distribution position parameter of the measured wind speed and the distribution position parameter of the predicted wind speed; based on the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed, and the distribution similarity parameter of the measured wind speed at the target time resolution and the predicted wind speed at the prediction time resolution, constructing the target time resolution wind speed ultra-short-term prediction model.

2. The wind power prediction method according to claim 1, characterized in that, The shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed is determined according to the following formula: wherein, is a shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the predicted wind speed, is a difference of the dynamic optimization function corresponding to the predicted wind speed, is a difference of the dynamic optimization function corresponding to the measured wind speed, is a minimum value of an absolute error of the difference of the dynamic optimization function corresponding to the predicted wind speed and the difference of the dynamic optimization function corresponding to the measured wind speed, is a maximum value of an absolute error of the difference of the dynamic optimization function corresponding to the predicted wind speed and the difference of the dynamic optimization function corresponding to the measured wind speed, and ω is a resolution coefficient of.

3. The wind power prediction method according to claim 2, characterized in that, The and is determined by the formula: In the formula, is a dynamic optimization function corresponding to the measured wind speed at time t, is a dynamic optimization function corresponding to the predicted wind speed at time t, is a dynamic optimization function corresponding to the measured wind speed at time t+1, is a dynamic optimization function corresponding to the predicted wind speed at time t+1.

4. The wind power prediction method according to claim 3, characterized in that, The is determined by the formula: wherein k is a time interval index, s is the total number of measured wind speed points in the kth time interval, and x(k+m) represents the measured wind speed at the mth measured wind speed point in the kth time interval at the target time resolution; The is determined by the formula: wherein r is the total number of predicted wind speed points in the kth time interval, and y(k+n) represents the predicted wind speed at the nth predicted wind speed point in the kth time interval at the prediction time resolution.

5. The wind power prediction method according to claim 1, characterized in that, The determination of the distribution position parameter of the measured wind speed comprises the following steps: obtaining the measured wind speed at the target time resolution, and determining the abscissa position corresponding to the measured wind speed at the target time resolution based on the measured wind speed at the target time resolution; determining a first mapping relationship between the measured wind speed at the target time resolution and the abscissa position corresponding to the measured wind speed at the target time resolution; based on the first mapping relationship, fitting the measured wind speed at the target time resolution and the abscissa position corresponding to the measured wind speed at the target time resolution to obtain the distribution position parameter of the measured wind speed.

6. The wind power prediction method according to claim 5, characterized in that, The first mapping relationship is determined according to the following formula: x = a x + b x λ x + σ x In the formula, x is the measured wind speed of the target time resolution, a x , b x , and σ x are the distribution position parameters of the measured wind speed, and λ x is the abscissa position corresponding to the measured wind speed of the target time resolution.

7. The wind power prediction method according to claim 6, characterized in that, The determination of the distribution position parameter of the predicted wind speed comprises the following steps: obtaining the predicted wind speed at the prediction time resolution, and determining the abscissa position corresponding to the predicted wind speed at the prediction time resolution based on the predicted wind speed at the prediction time resolution; determining a second mapping relationship between the predicted wind speed at the prediction time resolution and the abscissa position corresponding to the predicted wind speed at the prediction time resolution; The distribution position parameter of the predicted wind speed of the prediction time resolution is fitted based on the second mapping relationship, to obtain a distribution position parameter of the predicted wind speed.

8. The wind power prediction method according to claim 7, characterized in that, The second mapping relationship is determined according to the following formula: y = a y + b y λ y + σ y where y is the predicted wind speed of the prediction time resolution, a y , b y , σ y is the distribution position parameter of the predicted wind speed, and λ y is the horizontal coordinate position corresponding to the predicted wind speed of the prediction time resolution.

9. The wind power prediction method according to claim 8, characterized in that, The distribution similarity parameter between the measured wind speed of the target time resolution and the predicted wind speed of the prediction time resolution is determined based on the distribution position parameter of the measured wind speed and the distribution position parameter of the predicted wind speed, including: The probability density function of the measured wind speed is determined based on the distribution position parameter of the measured wind speed, and the probability density function of the predicted wind speed is determined based on the distribution position parameter of the predicted wind speed; The distribution similarity parameter between the measured wind speed of the target time resolution and the predicted wind speed of the prediction time resolution is determined based on the probability density function of the measured wind speed, the probability density function of the predicted wind speed, and the distribution feature similarity relationship between the probability density function of the measured wind speed and the probability density function of the predicted wind speed.

10. The wind power prediction method according to claim 9, characterized in that, The distribution feature similarity relationship between the probability density function of the measured wind speed and the probability density function of the predicted wind speed is determined according to the following formula: where μ(x|a x ; b x ; σ x 2 ) is the probability density function of the observed wind speed, g(y|a y ; b y ; σ y 2 ) is the probability density function of the forecast wind speed, z is the forecast member index, Z is the total number of forecast members, and w z is the distribution similarity parameter between the observed wind speed at the target time resolution and the forecast wind speed at the forecast time resolution for the zth forecast member.

11. The wind power prediction method according to claim 10, characterized in that, The target time resolution wind speed ultra-short-term prediction model is determined according to the following formula: In the formula, y target is the forecast wind speed of the target time resolution, is the shape correlation coefficient of the dynamic optimization function corresponding to the measured wind speed and the dynamic optimization function corresponding to the forecast wind speed.

12. The wind power prediction method according to claim 11, characterized in that, The wind power prediction model is constructed according to the following formula: wp = net {∑ρ 3 *y target +ε} In the formula, wp is the wind power prediction value of the target time resolution, net is the activation function of the neural network, ρ is the air density, and ε is the maximum error of the wind power prediction value of the target time resolution and the preset wind power value.

13. The wind power prediction method according to claim 1, characterized in that, The target time resolution wind speed is brought into the wind power prediction model constructed in advance to solve, including: The wind power prediction model is solved by using a deep learning algorithm based on the target time resolution predicted wind speed.

14. A wind power prediction device, characterized by Including: An acquisition module is configured to acquire a predicted wind speed of a prediction time resolution; A determination module is configured to input the predicted wind speed of the prediction time resolution into a target time resolution wind speed ultra-short-term prediction model constructed in advance to obtain a target time resolution predicted wind speed; A solving module is configured to bring the target time resolution predicted wind speed into a wind power prediction model constructed in advance to solve, to obtain a wind power prediction value corresponding to the target time resolution predicted wind speed; The prediction time resolution is greater than the target time resolution; and the target time resolution wind speed ultra-short-term prediction model is determined based on a distribution feature similarity relationship between a measured wind speed of the target time resolution and a predicted wind speed of the prediction time resolution to obtain the target time resolution predicted wind speed. The determination module includes the following steps: Based on the measured wind speed under the target time resolution and the predicted wind speed under the prediction time resolution, a shape correlation coefficient of a dynamic optimization function corresponding to the measured wind speed and a dynamic optimization function corresponding to the predicted wind speed is calculated; The distribution position parameter of the measured wind speed and the distribution position parameter of the predicted wind speed are determined, and the distribution similarity parameter between the measured wind speed of the target time resolution and the predicted wind speed of the prediction time resolution is determined based on the distribution position parameter of the measured wind speed and the distribution position parameter of the predicted wind speed. The target time resolution wind speed ultra-short-term prediction model is constructed based on the shape correlation coefficient of the dynamic optimization function of the measured wind speed and the dynamic optimization function of the predicted wind speed and the distribution similarity parameter of the measured wind speed of the target time resolution and the predicted wind speed of the predicted time resolution.

Images