A method, device and storage medium for predicting short-term wind power of a wind turbine generator set

Abstract

The present invention relates to the field of clean energy and provides a method for short-term wind power prediction for wind turbines. The method comprises the following steps: using cloud computing technology to collect real-time wind power data and meteorological data of the wind turbine within a preset time period, and using remote sensing technology to collect spatial feature data of the terrain surrounding the wind turbine; preprocessing the real-time wind power data, meteorological data, and data spatial feature data; performing extension representation on the preprocessed real-time wind power data, meteorological data, and data spatial feature data based on extenics to construct a wind power prediction model; combining terrain features and wind turbine structural parameters with modeling software to establish a wind farm model; extracting main characteristic factor data from the wind power prediction model, performing extension cluster analysis, and determining characteristic factor weights; and performing short-term wind power prediction using a time series exponential smoothing algorithm and multiple regression prediction, combined with the wind farm model and characteristic factor weights. This method improves prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation, and in particular to the field of short-term wind power prediction. Background Art

[0002] A wind energy conversion system, or wind turbine, is a device that efficiently converts natural wind energy into electrical energy. The core components of this system include the rotor (including blades), a generator, a support structure, and a control unit. Its basic operating principle is that the rotor rotates in response to wind force, driving the generator through a transmission system (which may include components such as a gearbox), thereby generating electricity. The output power of a wind turbine, or wind power, is affected by a variety of factors, including wind speed, rotor size, air density, blade design, and the efficiency of the power generation and transmission equipment. It is typically measured in kilowatts or megawatts.

[0003] To ensure the stable operation of wind power systems and maximize their energy efficiency, accurate short-term wind power forecasting is crucial. Because wind energy is an intermittent and unpredictable form of energy, its output is subject to significant volatility, influenced by a variety of external conditions. Accurate wind power forecasting can help grid operators predict wind power supply conditions within the next few hours or days, enabling them to develop appropriate grid dispatching and management strategies to ensure stable and reliable grid operation.

[0004] However, existing wind power forecasting technologies often overlook the crucial role of meteorological data (such as wind speed, wind direction, and temperature). Changes in wind direction directly impact the efficiency of wind turbines in capturing wind energy, thereby affecting their power output. Temperature fluctuations, on the other hand, affect air density, further impacting the availability of wind energy. If forecasting models fail to fully account for these meteorological factors, significant errors in forecast results are likely to occur, impacting forecast accuracy. Therefore, incorporating meteorological data into wind power forecasting models is a key measure to improve forecast accuracy. Summary of the Invention

[0005] The purpose of the present invention is to improve the accuracy of short-term power generation prediction of wind turbines.

[0006] To this end, in one embodiment, a method for predicting short-term wind power of a wind turbine generator system by integrating extenics and cloud computing technology is provided. The method includes the following steps:

[0007] Use cloud computing technology to collect real-time wind power data and meteorological data of wind turbines within a preset time period, and use remote sensing technology to collect spatial characteristic data of the terrain around the wind turbines;

[0008] Preprocessing the real-time wind power data, the meteorological data, and the data space feature data, including removing missing values ​​and outliers, and standardizing the data;

[0009] Performing extension representation on the pre-processed real-time wind power data, the meteorological data and the data space feature data based on extenics to construct a wind power prediction model;

[0010] Combining terrain features and wind turbine structural parameters, a wind farm model is constructed using modeling software;

[0011] Extract the main characteristic factor data from the wind power prediction model, perform extension cluster analysis, and determine the characteristic factor weights;

[0012] The short-term forecast of wind power is carried out by using the time series exponential smoothing algorithm and multiple regression prediction, combined with the wind farm model and characteristic factor weights.

[0013] Optionally, the modeling principle and process of the wind farm model may include the following steps:

[0014] 1) Data collection sub-steps:

[0015] Cloud computing technology is used to collect real-time wind power data and meteorological data of wind turbines within a preset time period.

[0016] Remote sensing technology is used to collect spatial characteristic data of the terrain around the wind turbine, including terrain height, slope and orientation.

[0017] 2) Data preprocessing sub-step:

[0018] The collected real-time wind power data, meteorological data and spatial feature data are preprocessed, including removing missing values ​​and outliers.

[0019] The data were standardized to ensure comparability between different variables.

[0020] 3) Construct extension representation sub-steps:

[0021] Based on the extenics theory, the pre-processed data is represented in an extenic manner, and the multi-dimensional data is converted into a unified form to facilitate subsequent modeling.

[0022] The wind turbine power and its influencing factors are characterized from multiple dimensions, and an extension model that can capture nonlinear relationships is constructed.

[0023] 4) Establishing wind farm model sub-steps:

[0024] Computational fluid dynamics (CFD) is used to simulate the wind flow dynamics in the wind farm by combining terrain characteristics (such as terrain height, slope and orientation) and wind turbine structural parameters (such as blade length and tower height).

[0025] Generate a wind field model that takes into account the influence of factors such as wind speed, wind direction, turbulence intensity, and wind turbine layout.

[0026] 5) Extracting the main characteristic factors sub-step:

[0027] The main characteristic factor data are extracted from the wind power prediction model, which have a significant impact on wind power.

[0028] Extension cluster analysis is performed to determine the weights of various characteristic factors to reflect their importance to wind power prediction.

[0029] 6) Short-term wind power forecast sub-step:

[0030] The time series exponential smoothing algorithm is used to process time series data and capture the changing trend of wind power.

[0031] Combining the multiple regression prediction method, the wind field model and characteristic factor weights are used to make short-term predictions of wind power.

[0032] By comprehensively considering the influence of multiple factors, the accuracy and reliability of the prediction results can be improved.

[0033] Through the above steps, the wind farm model can comprehensively consider various factors affecting wind power, and achieve high-precision short-term wind power forecasting by combining advanced mathematical methods and physical models.

[0034] Optionally, the wind field model can be specifically defined by the following formula:

[0035]

[0036] in:

[0037] ·M wind_field is the output power of the wind farm model.

[0038] ·v hub is the wind speed at hub height.

[0039] ·v max is the maximum wind speed.

[0040] υk is the exponent of wind speed variation with height.

[0041] h is the hub height.

[0042] υh ref is the reference hub height.

[0043] l is the blade length of the wind turbine.

[0044] l0 is the reference blade length.

[0045] ·s is the standard deviation of wind speed, which indicates the turbulence intensity.

[0046] s0 is the reference value of turbulence intensity.

[0047] φ is the wind direction angle.

[0048] ·φ ref is the reference wind direction angle.

[0049] A is the windward area of ​​the wind turbine.

[0050] υζ, α, β, γ, and δ are model parameters and are adjusted according to actual conditions.

[0051] The wind farm model defined by the above formula is a comprehensive model for predicting wind power, which takes into account a variety of influencing factors to improve the accuracy of the prediction. Specifically, the model combines the wind speed at the hub height, the maximum wind speed, the exponent of wind speed variation with height, the blade length of the wind turbine and its reference length, the standard deviation of the wind speed (indicating turbulence intensity), the difference between the wind direction angle and the reference wind direction angle, and the windward area of ​​the wind turbine. These factors are adjusted through specific weights and parameters to reflect their impact on wind power. The model also takes into account the characteristics of wind speed distribution, the influence of turbulence intensity and changes in wind direction, so as to more comprehensively simulate the complex conditions within the wind farm. In general, this wind farm model can more accurately predict the power generation capacity of wind turbines under different conditions by integrating multiple key variables.

[0052] Optionally, the extension representation step further includes: characterizing the wind turbine power and its influencing factors from multiple dimensions to construct an extension model;

[0053] Wherein, the extension model is defined as:

[0054]

[0055] in:

[0056] ·X={x1,x2,...,x n} is the feature variable in the preprocessed dataset.

[0057] υY={y1,y2,...,y m} is another set of characteristic variables.

[0058] Z={z1,z2,...,z p} is a feature set used for normalization.

[0059] U={u1,u2,...,u q} and V={v1,v2,...,v q} is the characteristic ratio for multiplication operation.

[0060] υω i ,a i ,b i Is with X i The associated weight and bias parameters.

[0061] c j ,d j ,e j Yes and Y j The coefficient and bias of the correlation.

[0062] γ,σ k are the coefficients and standard deviations used to normalize the sum.

[0063] ·δ,h l are the coefficients and exponents used in multiplication operations.

[0064] ·∈,λ,μ,ν are the coefficients and bias for the Sigmoid function.

[0065] In an alternative solution, the step of establishing the extension model is further included according to the following formula:

[0066]

[0067] Where, P(t): real-time wind power data;

[0068] v(t): wind speed data; h: terrain height; s: terrain slope; o: terrain direction; l: blade length; h t : tower height; d: average distance between wind turbines; α, β, γ, δ, ∈: weight coefficients;

[0069] And, explain each item in the above extension model:

[0070] Real-time wind power: α·P(t): Directly consider the wind power at the current moment and use the weight α for weighting.

[0071] Wind speed processing: β·log(v(t)+1): The wind speed data is processed using a logarithmic function to smooth the impact of wind speed changes and weighted using the weight β.

[0072] Terrain height: γ sin((π h) / 100): The terrain height is processed using a sine function to capture the periodic effect of terrain height on wind power, and the weight γ is used for weighting.

[0073] Terrain slope: δ·cos((π·s) / 180): The terrain slope is processed using the cosine function to capture the effect of slope on wind power, and is weighted using the weight δ.

[0074] Terrain orientation: ∈ tan((π o) / 360): The terrain orientation is processed using the tangent function to capture the impact of orientation on wind power, and weighted using weight ∈.

[0075] In this extension model, the relationship between wind power and its influencing factors can be effectively captured by using logarithmic functions and smooth trigonometric functions. The design of this model not only retains the influence of key variables but also avoids complex multi-dimensional feature processing, making it suitable for fast calculation and deployment in practical applications.

[0076] Specifically, by adjusting the weight coefficients α, β, γ, δ,∈, it is possible to flexibly adapt to the specific environment and data conditions of different wind farms. For example, in some wind farms, the impact of terrain height may be greater, and this can be strengthened by increasing the value of γ.

[0077] Each weight coefficient can be optimized according to actual conditions to improve the model's prediction accuracy, which enables the model to better adapt to various wind farm environments.

[0078] Fast calculation: Due to the simple formula and relatively small amount of calculation, predictions can be completed in a short time. This is particularly important for real-time or near-real-time application scenarios.

[0079] Smoothing: Logarithmic and trigonometric functions can be used to smooth out noise and outliers in the data, improving the robustness of the model. For example, the logarithmic function log(v(t)+1) can effectively handle zero and small values ​​in wind speed data.

[0080] Capturing periodic features: Trigonometric functions (such as sine and cosine) can capture periodic changes in terrain features such as height, slope, and orientation, thereby more accurately reflecting the impact of these factors on wind power.

[0081] Furthermore, in this newly established model, each term has a clear physical meaning, making it easier to interpret and adjust the impact of each characteristic variable on wind power. For example, α·P(t) directly reflects the wind power at the current moment, while β·log(v(t)+1) represents the nonlinear effect of wind speed on power.

[0082] In this way, if the prediction results are not ideal, the problem can be located by analyzing the contribution of each item and then targeted adjustments can be made.

[0083] In summary, the new extension model provides a new perspective and approach for wind power forecasting. This model is not only easy to implement and maintain, but also provides efficient and accurate forecasting in practical applications, making it particularly suitable for scenarios requiring fast response and low resource consumption.

[0084] Optionally, the wind field model establishment step includes:

[0085] Computational fluid dynamics is used to simulate the wind flow dynamics in the wind farm, and a wind farm model is generated by combining the layout of wind turbines and terrain characteristics.

[0086] Optionally, the extension cluster analysis and weight calculation steps include:

[0087] Cluster the wind turbine power observation data to obtain the power level and its normalized range;

[0088] Calculate the degree of membership of the main characteristic factor data to each power level and determine the weight data.

[0089] Optionally, the short-term wind power forecasting step includes:

[0090] Input historical power data, main characteristic factor data and their weight data into the prediction model to perform multivariate regression prediction. Optionally, the wind farm model is defined as:

[0091] M=α·TIF+β·BLF+γ·THF+δ·LF+η·WDF+ζ·VIF

[0092] Among them, α, β, γ, δ, η, ζ are weight coefficients used to adjust the importance of each factor;

[0093] The TIF is the slope and orientation, the BLF is the blade length, the THF is the tower height, the LF is the wind turbine layout, the WDF is the wind direction, and the VIF is the wind speed.

[0094] Optionally, the wind field model M can be expressed as:

[0095]

[0096] Where, h: terrain height

[0097] s: terrain slope o: terrain orientation l: blade length ht: tower height Pi: rated power of the i-th wind turbine φ: wind direction angle v: wind speed.

[0098] In another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program to execute the method described in any other embodiment.

[0099] In another embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in any other embodiment is implemented.

[0100] In some embodiments of the present application, a wind farm model is established by combining the structural parameters of the wind turbine and the terrain characteristics. This combines the statistical model with the physical model, taking into account both the physical characteristics of the wind turbine and the actual operating environment of the wind farm, thereby improving the accuracy of the prediction.

[0101] Furthermore, by extracting key characteristic factor data and assigning weights to them, the prediction model can comprehensively consider multiple factors, such as wind speed, topography, and weather, thereby improving the reliability of the prediction results. Furthermore, by utilizing a time-series exponential smoothing algorithm and multiple regression forecasting techniques, combined with wind farm models and characteristic factor weights, short-term wind power forecasting can effectively reduce prediction errors and improve prediction accuracy.

[0102] In other embodiments, the wind farm model expressed using the formula can more comprehensively consider various physical effects in the wind farm, such as the influence of factors such as terrain, blade length, tower height, turbine layout, wind direction, and wind speed, thereby more accurately predicting wind power. In contrast, existing simplified models only consider the relationship between wind speed and rated power, ignoring other important factors, resulting in less accurate prediction results.

[0103] From the above comparison, it can be seen that the use of the above mathematical model can better reflect the actual situation of the wind field and improve the accuracy and reliability of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0105] Figure 1 A flow chart showing a method for predicting short-term wind power of a wind turbine generator system according to the present invention is shown;

[0106] Figure 2 A schematic structural diagram of a short-term wind power prediction device for a wind turbine generator system according to an embodiment of the present application is shown;

[0107] Figure 3A schematic structural diagram of a computing device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0108] The "Least Squares Support Vector Machine" (LSSVM) is often abbreviated as "LSSVM." This is a regression analysis method derived from a modification of the traditional support vector machine (SVM). Its unique feature is that it uses the least squares method to solve regression problems, simplifying the solution process and improving model performance. In different literature or application scenarios, the full name "Least Squares Support Vector Machine" may also be used to clearly refer to this method. An LSSVM typically includes multiple parameters, such as the gamma (γ) parameter and the sigma (σ) parameter.

[0109] In one embodiment of the present application, a short-term wind power prediction device for a wind turbine is provided, comprising:

[0110] A database component is configured to store environmental parameters and electrical energy generated by the fluid generator at multiple past time nodes, wherein any two adjacent time nodes among the multiple time nodes form a group, forming two or more groups, and the two adjacent time nodes are respectively the first time node and the second time node in a time sequence relationship; the environmental parameters include: ambient flow rate;

[0111] a flow rate component electrically coupled to the database component, using the environmental parameters of the first time series node of each group as injection data of the flow rate component and using the environmental flow rate of the corresponding second time series node as fan-out data of the flow rate component to train the flow rate component;

[0112] a power generation component electrically coupled to the database component, using the fan-out data of the flow rate component as injection data of the power generation component, and using the electric energy generated at the corresponding second time series node as fan-out data of the power generation component, so as to train the power generation component;

[0113] a sensing component, configured to obtain environmental parameters of the fluid generator at a current timing node; and

[0114] A processor component is coupled to the database component, the flow rate component, the power generation component and the sensing component. The processor component inputs the environmental parameters obtained by the sensing component into the flow rate component to obtain an estimated flow rate of the fluid generator at the next timing node. The processor component inputs the estimated flow rate into the power generation component to predict the estimated power generation of the fluid generator at the next timing node.

[0115] In one embodiment of the present application, Figure 1 As shown in FIG, a short-term wind power prediction method for a wind turbine is proposed, which includes the following steps:

[0116] Step S1: using cloud computing technology to collect real-time wind power data and meteorological data of the wind turbine within a preset time period, and using remote sensing technology to collect spatial feature data of the terrain surrounding the wind turbine;

[0117] Step S2: preprocessing the real-time wind power data, the meteorological data, and the data space feature data, including removing missing values ​​and outliers, and standardizing the data;

[0118] Step S3: performing extension representation on the pre-processed real-time wind power data, the meteorological data and the data space feature data based on extenics to construct a wind power prediction model;

[0119] Step S4: using modeling software to establish a wind farm model based on terrain features and wind turbine structural parameters;

[0120] Step S5: extracting main characteristic factor data from the wind power prediction model, performing extension cluster analysis, and determining characteristic factor weights;

[0121] Step S6: using the time series exponential smoothing algorithm and multiple regression prediction, combined with the wind farm model and characteristic factor weights, to perform short-term prediction of wind power.

[0122] Establishment of wind field model

[0123] Computational fluid dynamics is used to simulate the wind flow dynamics in the wind farm, and a wind farm model is generated by combining the layout of wind turbines and terrain characteristics. First, some key variables need to be defined in some embodiments of this application:

[0124] h: Terrain height

[0125] s: terrain slope

[0126] o: Terrain orientation

[0127] l: blade length

[0128] ht: tower height

[0129] Pi: rated power of the i-th wind turbine

[0130] φ: wind direction angle

[0131] v: wind speed

[0132] Terrain Impact Factor (TIF)

[0133] The terrain impact factor TIF is defined as follows, which combines the effects of terrain height h, slope s, and orientation o:

[0134] TIF=log(h+1)·e s ·cos(o)

[0135] Here log(h+1) is used to quantify the effect of terrain height on wind speed. The logarithmic operation ensures that even low altitude areas can have a certain weight. s It is used to quantify the effect of terrain slope on wind speed. The exponential calculation reflects that the greater the slope, the more significant the impact on wind speed. Cos(o) is used to quantify the effect of terrain orientation on wind speed, accounting for the differences in the influence of different orientations on wind speed.

[0136] Leaf impact factor BLF

[0137] The blade influence factor BLF is defined as follows, which quantifies the effect of blade length L on wind energy capture capability:

[0138] BLF=1 / 2·(1+tanh(L / 10))

[0139] The hyperbolic tangent function tanh is used here, which takes into account the nonlinear effect of blade length on wind energy capture ability.

[0140] Tower influence factor THF

[0141] The tower influence factor THF is defined as follows, which quantifies the tower height h t Impact on wind speed:

[0142]

[0143] The square root operation is used here, which reflects that the higher the tower height, the more obvious the impact on wind speed.

[0144] Fan layout factor LF

[0145] The wind turbine layout factor LF is defined as follows, which takes into account the impact of wind turbine layout on wind speed within the wind farm:

[0146]

[0147] where d i is the distance between the ith wind turbine and its nearest neighbor, and N is the total number of wind turbines in the wind farm.

[0148] Wind direction factor WDF

[0149] The wind direction factor WDF is defined as follows, which quantifies the effect of the wind direction angle φ on the wind speed within the wind farm:

[0150] WDF=sin(φ)·cos(φ)

[0151] Sine and cosine functions are used here, which take into account the effect of wind direction on wind speed.

[0152] Wind speed impact factor VIF

[0153] The wind speed impact factor VIF is defined as follows, which quantifies the impact of wind speed v on the wind speed inside the wind farm:

[0154]

[0155] where v max is the maximum possible wind speed in the wind farm.

[0156] Comprehensive wind field model M

[0157] The comprehensive wind field model M is defined as follows, which combines all the above factors:

[0158] M=α·TIF+β·BLF+γ·THF+δ·LF+η·WDF+ζ·VIF

[0159] Among them, α, β, γ, δ, η, and ζ are weight coefficients used to adjust the importance of each factor.

[0160] In this way, some embodiments of the present application can construct a wind field model that includes complex mathematical formulas such as logarithmic operations and exponential operations, which is used to accurately simulate the dynamic changes of wind speed and wind direction within the wind field, thereby better predicting wind power.

[0161] Optionally, for simplicity, in other embodiments, the above wind field model can be integrated into a single formula. In some embodiments of the present application, various influencing factors can be integrated together while maintaining clear expression. The following is a possible implementation:

[0162] It is assumed that in some embodiments of the present application, all influencing factors (TIF, BLF, THF, LF, WDF and VIF) have been defined and corresponding weight coefficients (α, β, γ, δ, η and ζ) have been obtained.

[0163] The comprehensive wind field model M can be expressed as:

[0164]

[0165] This formula incorporates all previously defined influencing factors and sums them up according to their weights. Each influencing factor takes into account different physical effects, such as terrain height, slope and orientation (TIF), blade length (BLF), tower height (THF), turbine layout (LF), wind direction (WDF) and wind speed (VIF).

[0166] In this way, in some embodiments of the present application, the model can be used to predict the short-term wind power of the wind turbine.

[0167] Here, for a method of an embodiment, a specific application scenario is constructed and corresponding parameters are given.

[0168] Application Scenario

[0169] Assume that in some embodiments of this application, there is a wind farm located in a mountainous area, and the short-term wind power output of the wind turbines needs to be predicted. The wind farm includes 10 wind turbines, each with a rated power of 2 MW. In some embodiments of this application, it is desirable to use the aforementioned wind farm model to predict wind power.

[0170] Parameter settings

[0171] Terrain height h = 150m

[0172] υ Terrain slope s=10°

[0173] Terrain orientation o = 30° (north is 0 degrees)

[0174] Blade length l = 60m

[0175] Tower height ht = 100m

[0176] Wind direction angle φ = 45°

[0177] Wind speed v = 10 m / s

[0178] υMaximum wind speed v max =25m / s

[0179] υThe average distance between wind turbines di=150m

[0180] υWeight coefficient: α=0.3, β=0.2, γ=0.1, δ=0.2, η=0.1, ζ=0.1

[0181] Formula expression of wind field model:

[0182]

[0183] Mathematical model and its calculation process

[0184] 1. Terrain Impact Factor (TIF)

[0185] TIF=log(h+1)·e s ·cos(o)

[0186] =log(150+1)·e 10cos(30°)

[0187] =log(151)·e 0.174533 cos(30°)

[0188] =2.1767·1.1892·0.8660

[0189] =2.0782

[0190] 2. Leaf impact factor BLFBLF

[0191]

[0192] 3. Tower influence factor THF

[0193]

[0194] 4. Fan layout factor LF

[0195]

[0196] 5. Wind direction factor WDF

[0197] WDF=sin(φ)·cos(φ)

[0198] = sin(45°)·cos(45°)

[0199] = sin(45°)·cos(45°)

[0200] =0.7071·0.7071

[0201] =0.5000

[0202] 6. Wind speed impact factor VIF

[0203]

[0204] 7. Comprehensive wind field model M

[0205] M=α·TIF+β·BLF+γ·THF+δ·LF+η·WDF+ζ·VIF

[0206] =0.3·2.0782+0.2·0.99995+0.1·10+0.2·0.0008889+0.1·0.5000+0.1·4

[0207] =0.62346+0.19999+1+0.00017778+0.05+0.4

[0208] =2.27362778

[0209] Calculation process in existing technology

[0210] In the prior art, these complex factors are usually not considered, but simplified models are used. For example, a simplified model may directly use wind speed v and wind turbine rated power P to calculate wind power Pw.

[0211]

[0212] Assume that the rated power of the wind turbine is P = 2MW, the maximum wind speed is max = 25m / s, and the wind speed is v = 10m / s.

[0213]

[0214] Using the mathematical formula in the above embodiment: the obtained comprehensive wind field model M=2.27362778.

[0215] Simplified model in the prior art: The wind power P obtained w =0.128MW.

[0216] Technical Effects

[0217] The wind farm model expressed using the above formula can more comprehensively consider various physical effects within the wind farm, such as the influence of topography, blade length, tower height, turbine layout, wind direction, and wind speed, enabling more accurate predictions of wind power. In contrast, existing simplified models only consider the relationship between wind speed and rated power, ignoring other important factors, resulting in less accurate predictions.

[0218] From the above comparison, it can be seen that the use of the above mathematical model can better reflect the actual situation of the wind field and improve the accuracy and reliability of the prediction.

[0219] Considering multiple factors in wind power forecasting, and using mathematical functions to quantify the impact of these factors. The following is a formula based on these requirements:

[0220] Assume that the following parameters exist in some embodiments of this application:

[0221] P(t): wind power

[0222] v(t): wind speed

[0223] h: Terrain height

[0224] s: terrain slope

[0225] o: Terrain orientation

[0226] / : Blade length

[0227] ·ht : Tower height

[0228] d: average distance between wind turbines

[0229] Pmax: Maximum power of wind turbine

[0230] υμ v ,σ v :The mean and standard deviation of wind speed

[0231] υμ P ,σ P :The mean and standard deviation of wind power

[0232] μ h ,σ h: Mean and standard deviation of terrain height

[0233] μ s, σ s : Mean and standard deviation of terrain slope

[0234] υμ o ,σ o :The mean and standard deviation of terrain orientation

[0235] μ l, σ l: Mean and standard deviation of leaf length

[0236] μ ht, σ ht : Mean and standard deviation of tower height

[0237] ·α,β,γ,δ,η,ζ: weight coefficient

[0238] In some embodiments of the present application, a wind power prediction model (WPPM) is defined as follows:

[0239]

[0240] 1. Wind power standardization: Use logarithmic and exponential functions to enhance sensitivity to changes in small values ​​while ensuring numerical stability.

[0241] 2. Wind speed standardization: Use an exponential function to enhance sensitivity to changes in large values ​​while maintaining numerical stability.

[0242] 3. Terrain height standardization: Use logarithmic and exponential functions to enhance sensitivity to changes in small values ​​while ensuring numerical stability.

[0243] 4. Terrain slope standardization: Use an exponential function to enhance sensitivity to changes in large values ​​while maintaining numerical stability.

[0244] 5. Terrain orientation standardization Use logarithmic and exponential functions to enhance sensitivity to changes in small values ​​while ensuring numerical stability.

[0245] 6. Standardization of blade length, tower height and wind turbine spacing:

[0246] Use exponential and logarithmic functions to increase sensitivity to changes in different values.

[0247] Example calculation process

[0248] Assume there is a wind farm located in the eastern coastal area of ​​China. The wind farm contains 20 wind turbines, each with a rated power of 2MW. In some embodiments of this application, it is necessary to predict the wind power within the next 4 hours. The following are the specific parameters:

[0249] υ Wind power P(t) = 1.5MW

[0250] υ wind speed v(t)=8m / s

[0251] Terrain height h = 50 meters

[0252] Terrain slope s = 2°

[0253] Terrain orientation o = 180°

[0254] Blade length l = 50 meters

[0255] Tower height ht = 80 meters

[0256] The average distance between wind turbines is d = 500 meters

[0257] Maximum power of wind turbine Pmax = 2MW

[0258] Average wind speed μ v =7m / s and standard deviation σ v =2m / s

[0259] Average value of wind power μ P =1.4MW and standard deviation σ P =0.3MW

[0260] υThe average value of terrain height μ h = 45 m and standard deviation σ h =5 meters

[0261] Average value of terrain slope μ s = 1° and standard deviation σ s=0.5°

[0262] Average value of terrain orientation μ o = 180° and standard deviation σ o =10°

[0263] υThe average value of blade length μ l = 45 m and standard deviation σ l =5 meters

[0264] Average tower height μ ht = 75 m and standard deviation σ ht =10 meters

[0265] ·Weight coefficient α=0.2, β=0.3, γ=0.1, δ=0.05, η=0.05, ζ=0.25

[0266] Substitute the above parameters for calculation:

[0267]

[0268] Calculate each of the above:

[0269]

[0270] Adding up the above results:

[0271] WPPM=0.0328+0.4502+0.0508+0.3699+0+0.4200=1.3237

[0272] Calculation process in existing technology

[0273] In existing technologies, only the direct relationship between wind speed and wind power is usually considered, without considering terrain and other factors. In some embodiments of the present application, a simple linear regression model is used to estimate wind power:

[0274] P 预测 =β0+β1·v(t)

[0275] Where β0 and β1 are parameters obtained by fitting historical data. Assuming β0 = 0.2MW and β1 = 0.2MW / m / s, then:

[0276] P 预测 =0.2+0.2·8=1.8MW

[0277] Control and analysis

[0278] The calculations above demonstrate that the formulated wind farm model takes into account the influence of more factors, such as terrain characteristics and wind turbine structural parameters. In this example, the formula yields 1.3237 MW, while the existing technology yields 1.8 MW. Clearly, the existing technology fails to account for the influence of terrain and other factors, resulting in an overestimation.

[0279] Technical Effects

[0280] Improved accuracy: By considering more factors that affect wind power, the calculated results of the formulated wind farm model are more accurate.

[0281] Comprehensive analysis: This formula not only considers the important factor of wind speed, but also comprehensively considers terrain characteristics and wind turbine structural parameters, which makes the calculation model more comprehensive.

[0282] Strong adaptability: By adjusting the weight coefficient, it can adapt to different wind farm environments and improve the applicability of the calculation model.

[0283] In summary, compared to the existing technology, the "formulated mathematical model" has higher accuracy in calculating wind power and can more comprehensively reflect the complex environmental factors of actual wind farms, thereby improving the reliability of the calculation results. The above formula shows how to express a wind power prediction model. Through this example, it can be seen in some embodiments of the present application that the formula can comprehensively consider multiple factors in wind power prediction and quantify the impact of these factors through mathematical operations. In another embodiment of the present application, an attempt is made to establish a mathematical model for predicting wind power.

[0284] 1. Algorithm formula for wind power

[0285] Assume that in some embodiments of the present application, there are time series data P(t) of wind power and a series of characteristic variables X. In some embodiments of the present application, the goal is to predict the wind power P(t+1) at a future time point t+1.

[0286] 1.1 Time Series Exponential Smoothing Algorithm

[0287] In some embodiments of the present application, a double exponential smoothing algorithm is used to capture the time series trend of wind power. Assuming that lt is the horizontal smoothing value, bt is the trend smoothing value, α is the horizontal smoothing coefficient, and β is the trend smoothing coefficient, then:

[0288] l t =α·P(t)+(1-α)·(l t-1 +b t-1 )

[0289] b t =β·(lt -1 t-1 )+(1-β)·b t-1

[0290] 1.2 Multiple regression prediction

[0291] The multivariate regression prediction model considers the influence of multiple characteristic variables. Let X be the characteristic matrix and y be the target vector (wind power), then the multivariate regression model can be expressed as:

[0292] y=X·β+∈

[0293] Where β is the regression coefficient vector and ∈ is the error term. In some embodiments of the present application, β can be solved by the least squares method:

[0294]

[0295] For a new observation x new , predicting wind power for:

[0296]

[0297] 1.3 Feature Factor Weight

[0298] In order to combine the wind field model and the characteristic factor weights, some embodiments of the present application need to define the characteristic weight w and apply it to the multiple regression model. Assuming that some embodiments of the present application have a characteristic weight vector w, the multiple regression model becomes:

[0299]

[0300] in, Represents element-wise multiplication.

[0301] 1.4 Comprehensive prediction model

[0302] Combining the above three parts, some embodiments of this application can establish a comprehensive prediction model. First, some embodiments of this application use a time series exponential smoothing algorithm to predict the trend portion of wind power, and then use a multiple regression model to predict the fluctuation portion of wind power. Finally, some embodiments of this application combine the two prediction results and the characteristic factor weights to obtain the final prediction result. To increase the complexity and sophistication of the formula, some embodiments of this application will introduce logarithmic and exponential operations.

[0303] Assume P pred,trend (t+1) is the prediction result of the time series exponential smoothing algorithm, P pred,regression (t+1) is the result of multiple regression prediction, so the final predicted wind power P pred,final (t+1) is:

[0304]

[0305] in:

[0306] υl t and b t They are the horizontal smoothing value and trend smoothing value of the time series exponential smoothing algorithm.

[0307] υx new is the eigenvector of the new observation.

[0308] ·(X T X) -1 X T y is the regression coefficient of the multiple regression model.

[0309] w is the feature weight vector.

[0310] λ is the fusion coefficient, which is used to balance the contribution of the two prediction results and is usually between 0 and 1.

[0311] γ and δ are weight coefficients for logarithmic and exponential operations.

[0312] μ P and σ P are the time series mean and standard deviation of wind power, respectively.

[0313] · and are the mean and standard deviation of the regression prediction results.

[0314] 2. Specific application scenarios and detailed parameters

[0315] In order to demonstrate the use of the above formula, some embodiments of the present application will use a set of assumed parameters to calculate wind power prediction.

[0316] Assume that the following data and parameters are present in some embodiments of this application:

[0317] Wind power time series P(t): [1.5, 1.6, 1.8, 2.0, 1.9, 1.7, 1.8, 1.9, 2.1, 2.2]

[0318] Smoothing coefficients α = 0.3, β = 0.2

[0319] Feature matrix X:

[0320]

[0321] Target vector y: [1.5, 1.6, 1.8, 2.0, 1.9]

[0322] Feature weight vector w: [1, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2]

[0323] Fusion coefficient λ = 0.5

[0324] Logarithmic weight coefficient γ = 0.2

[0325] Exponential weight coefficient δ = 0.2

[0326] Time series mean of wind power μ P =1.85MW

[0327] Time series standard deviation of wind power σ P =0.2MW

[0328] The mean of the regression prediction results

[0329] The standard deviation of the regression prediction results

[0330] 2.1 Example Calculation

[0331] 1. Time Series Exponential Smoothing Algorithm:

[0332] Assume l0 = 1.5, b0 = 0 (initial conditions):

[0333] l1=0.3·1.5+(1-0.3)·(1.5+0)=1.5

[0334] b1=0.2·(1.5-1.5)+(1-0.2)·0=0

[0335] l2=0.3·1.6+(1-0.3)·(1.5+0)=1.59

[0336] b2=0.2·(1.59-1.5)+(1-0.2)·0=0.018

[0337] …

[0338] P pred，rend (t+1)=1 t +b t

[0339] 2. Multiple regression prediction:

[0340] Assuming that x and y are known, some embodiments of this application calculate

[0341]

[0342] Then calculate P pred,regression (t+1):

[0343]

[0344] 3. Comprehensive forecast:

[0345]

[0346] 2.2 Calculation process in existing technology

[0347] In the prior art, only the direct relationship between wind speed and wind power is usually considered, without considering terrain and other factors. In some embodiments of the present application, a simple linear regression model is used to estimate wind power:

[0348] P pred =β0+β1·v(t)

[0349] Where β0 and β1 are parameters obtained by fitting historical data. Assuming β0 = 0.2MW and β1 = 0.2MW / m / s, then:

[0350] P pred =0.2+0.2·8=1.8MW

[0351] 2.3 Control and Analysis

[0352] From the above calculation, we can see that the “algorithm formula” takes into account the influence of more factors, such as terrain characteristics, wind turbine structure parameters, etc. In this example, the calculation result of the “algorithm formula” is P pred,final (t+1), while the result calculated by the existing technology is 1.8MW. Obviously, the existing technology does not take into account the influence of terrain and other factors, resulting in an overestimation of the prediction result.

[0353] 2.4 Technical Effects

[0354] Improved accuracy: The algorithm formula takes into account more factors that affect wind power, resulting in more accurate calculation results.

[0355] Comprehensive analysis: This formula not only considers the important factor of wind speed, but also comprehensively considers terrain characteristics and wind turbine structural parameters, which makes the calculation model more comprehensive.

[0356] Strong adaptability: By adjusting the weight coefficient, it can adapt to different wind farm environments and improve the applicability of the calculation model.

[0357] In summary, compared with existing technologies, the "algorithm formula" has higher accuracy in calculating wind power and can more comprehensively reflect the complex environmental factors of actual wind farms, thereby improving the reliability of the calculation results.

[0358] In another embodiment, in combination with a specific application scenario and corresponding detailed parameters, the above-mentioned assumed parameters are substituted into the "algorithm formula" of the model to perform specific numerical calculations for short-term predictions of wind power, and a numerical calculation process in the prior art that does not use the above-mentioned formula is given. The two calculation processes and results are compared, thereby proving that: compared with the prior art, the "algorithm formula" in the embodiment of this application has better technical effects.

[0359] To demonstrate the technical effectiveness of the "algorithm formula" in short-term wind power forecasting, some examples of this application will be combined with a specific wind power forecasting application scenario and provide detailed parameters and calculation processes. First, some examples of this application will use the "algorithm formula" to perform calculations and compare them with existing technical methods.

[0360] Application scenario description

[0361] Assume there is a wind farm located in the eastern coastal area of ​​China. The wind farm contains 20 wind turbines, each with a rated power of 2MW. In some embodiments of this application, it is necessary to predict the wind power within the next 4 hours. The following are the specific parameters:

[0362] Wind power P(t) = [1.5, 1.6, 1.8, 2.0, 1.9, 1.7, 1.8, 1.9, 2.1, 2.2] MW

[0363] υ smoothing coefficient α=0.3,β=0.2

[0364] Feature matrix X:

[0365]

[0366] Target vector y: [1.5, 1.6, 1.8, 2.0, 1.9]MW

[0367] Feature weight vector w: [1, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3, 0.2]

[0368] Fusion coefficient λ = 0.5

[0369] υ logarithmic operation weight coefficient γ=0.2

[0370] Exponential weight coefficient δ = 0.2

[0371] Time series mean of wind power μ P =1.85MW

[0372] υ Time series standard deviation of wind power σ P =0.2MW

[0373] The mean of the regression prediction results

[0374] The standard deviation of the regression prediction results

[0375] The calculation process of the "algorithm formula"

[0376] 1. Time Series Exponential Smoothing Algorithm

[0377] First, in some embodiments of the present application, a double exponential smoothing algorithm is used to predict the trend of wind power. Assuming l0=1.5, b0=0 (initial conditions):

[0378] l1=0.3·1.5+(1-0.3)·(1.5+0)=1.5

[0379] b1=0.2·(1.5-1.5)+(1-0.2)·0=0

[0380] l2=0.3·1.6+(1-0.3)·(1.5+0)=1.59

[0381] b2=0.2·(1.59-1.5)+(1-0.2)·0=0.018

[0382] …

[0383] P pred，trend (t+1)=1 t +b t

[0384] Calculate l t and b t The value of P pred,trend (t+1).

[0385] 2. Multiple regression prediction

[0386] Next, in some embodiments of the present application, a multivariate regression prediction model is used to predict the fluctuation of wind power.

[0387]

[0388] Then calculate P pred,regression (t+1):

[0389]

[0390] where x new is the new observation eigenvector, is the regression coefficient vector, and ω is the feature weight vector.

[0391] 3. Comprehensive Forecast

[0392] Finally, in some embodiments of the present application, the two prediction results and the feature factor weights are combined to obtain the final prediction result. According to the "algorithm formula", P is calculated. pred,final (t+1):

[0393]

[0394] Numerical calculation process

[0395] 1. Time Series Exponential Smoothing Algorithm

[0396] Assume l0 = 1.5, b0 = 0 (initial conditions):

[0397] l1=0.3·1.5+(1-0.3)·(1.5+0)=1.5

[0398] b1=0.2·(1.5-1.5)+(1-0.2)·0=0

[0399] l2=0.3·1.6+(1-0.3)·(1.5+0)=1.59

[0400] b2=0.2·(1.59-1.5)+(1-0.2)·0=0.018

[0401] …

[0402] P pred，trend (t+1)=1 t +b t

[0403] Calculate l t and b t The value of P pred,trend (t+1). Assume l 10 =1.86, b 10 =0.02, then:

[0404] P pred，trend (t+1)=1.86+0.02=1.88MW

[0405] 2. Multiple regression prediction

[0406] calculate

[0407]

[0408] Calculate P pred,regression (t+1):

[0409] Assumptions (This is a simplified example), then:

[0410]

[0411] Assume x new =[1, 8.5, 50, 2, 180, 50, 80, 500] (this is a simplified example), then:

[0412]

[0413] 3. Comprehensive Forecast

[0414] According to the "algorithm formula", calculate P pred,final (t+1):

[0415]

[0416] Calculate each term:

[0417] 1.0.5·1.88=0.94

[0418] 2.0.5·0.676=0.338

[0419]

[0420] Adding up the above results:

[0421] P pred,final (t+1)=0.94+0.338+0.1446+0.00048=1.42308MW

[0422] Calculation process in existing technology

[0423] In the prior art, only the direct relationship between wind speed and wind power is usually considered, without considering terrain and other factors. In some embodiments of the present application, a simple linear regression model is used to estimate wind power:

[0424] P pred =β0+β1.υ(t)

[0425] Where β0 and β1 are parameters obtained by fitting historical data. Assuming β0 = 0.2MW and β1 = 0.2MW / m / s, then:

[0426] P pred =0.2+0.2·8=1.8MW

[0427] Control and analysis

[0428] The calculations above demonstrate that the algorithmic formula for the model in this embodiment takes into account the influence of more factors, such as terrain characteristics and wind turbine structural parameters. In this example, the algorithmic formula yields a result of 1.42308 MW, while the prior art yields a result of 1.8 MW. Clearly, the prior art fails to account for the influence of terrain and other factors, resulting in an overestimation of the predicted value.

[0429] Technical Effects

[0430] Improved accuracy: The “algorithm formula” in some embodiments of the present application takes into account more factors that affect wind power, so the calculation results are more accurate.

[0431] Comprehensive analysis: This formula not only considers the important factor of wind speed, but also comprehensively considers terrain characteristics and wind turbine structural parameters, which makes the calculation model more comprehensive.

[0432] Strong adaptability: By adjusting the weight coefficient, it can adapt to different wind farm environments and improve the applicability of the calculation model.

[0433] In summary, compared with existing technologies, the "algorithm formula" has higher accuracy in calculating wind power and can more comprehensively reflect the complex environmental factors of actual wind farms, thereby improving the reliability of the calculation results.

[0434] In some alternative embodiments, in combination with a specific application scenario and corresponding detailed parameters, the above-mentioned assumed parameters are substituted into the "algorithm formula" of the model to perform specific numerical calculations for short-term predictions of wind power, and a numerical calculation process in the prior art that does not adopt the above-mentioned formula is given. The two calculation processes and results are compared, thereby proving that: compared with the prior art, the "algorithm formula" in the embodiments of the present application has better technical effects.

[0435] 1. Algorithm formula

[0436] 1.3 Building a wind power prediction model

[0437]

[0438] in:

[0439] WPPM: Wind Power Prediction Model

[0440] P(t): Real-time wind power data

[0441] v(t): wind speed from meteorological data

[0442] h: Terrain height

[0443] s: Terrain slope

[0444] o: Terrain orientation

[0445] l: blade length

[0446] ·h t :Tower height

[0447] d: average distance between wind turbines

[0448] ·P max : Maximum power of wind turbine

[0449] μ P ,σ P :The mean and standard deviation of wind power

[0450] μ v ,σ v :The mean and standard deviation of wind speed

[0451] μ h, σ h : The mean and standard deviation of terrain height

[0452] μ s ,σ s : Mean and standard deviation of terrain slope

[0453] μ o ,σ o :The mean and standard deviation of terrain orientation

[0454] μ l ,σ l :The mean and standard deviation of leaf length

[0455] μ ht ,σ ht : Mean and standard deviation of tower height

[0456] ·α,β,γ,δ,η,ζ: weight coefficient

[0457] 1.4 Establishing a wind field model

[0458]

[0459] in:

[0460] WFModel: Wind Field Model

[0461] λ,ρ,σ: weight coefficients

[0462] μ d ,σ d :The mean and standard deviation of wind turbine spacing

[0463] μ Pmax ,σ Pmax :The mean and standard deviation of the maximum power

[0464] 1.5 Determine the weight of characteristic factors

[0465]

[0466] Where: W features : Feature factor weight, ω P ,ω v ,ω h ,ω s ,ω o ,ω l :Weight coefficient

[0467] 1.6 Short-term forecast of wind power

[0468] P predicted =θ·(P pred,trend (t+1)+P pred,regression (t+1))

[0469] +(1-θ)·(WFModel·W features )

[0470] P pred,trend (t+1)=1 t +b t

[0471]

[0472] in:

[0473] ·P predicted :Short-term wind power forecast results

[0474] θ: Fusion coefficient

[0475] ·P pred,trend (t+1): The result of the time series exponential smoothing algorithm prediction

[0476] ·P pred,regression (t+1): Result of multiple regression prediction

[0477] ·l t :Level term in time series exponential smoothing algorithm

[0478] b t :Trend term in time series exponential smoothing algorithm

[0479] ·x new : New observation feature vector

[0480] · Regression coefficient vector

[0481] w: feature weight vector

[0482] 2. Formulas of core concepts

[0483] Wind power prediction model

[0484]

[0485] Wind field model

[0486]

[0487] Short-term wind power forecast

[0488] P predicted =θ·(P pred,trend (t+1)+P pred,regression (t+1))

[0489] +(1-θ)·(WFModel·W features )

[0490] P pred,trend (t+1)=1 t +b t

[0491]

[0492] 3. Specific application scenarios and detailed parameters

[0493] Application Scenario Description: Assume that there is a wind farm located in the eastern coastal area of ​​China. The wind farm contains 20 wind turbines, each with a rated power of 2MW. In some embodiments of this application, it is necessary to predict the wind power within the next 4 hours. The following are the specific parameters:

[0494] Wind power P(t) = 1.5MW

[0495] Wind speed v(t) = 8 m / s

[0496] Terrain height h = 50 meters

[0497] Terrain slope s = 2°

[0498] Terrain orientation o = 180°

[0499] Blade length l = 50 meters

[0500] Tower height h t =80 meters

[0501] υThe average distance between wind turbines d = 500 meters

[0502] Maximum power of wind turbine Pmax = 2MW

[0503] Average wind speed μ v =7m / s and standard deviation σ v =2m / s

[0504] Average value of wind power μ P =1.4MW and standard deviation σ P =0.3MW

[0505] Average terrain height μ h = 45 m and standard deviation σ h =5 meters

[0506] Average value of terrain slope μ s = 1° and standard deviation σ s =0.5°

[0507] Average value of terrain orientation μ o = 180° and standard deviation σ o =10°

[0508] Average leaf length μ l = 45 m and standard deviation σ l =5 meters

[0509] Average tower height μ ht = 75 m and standard deviation σ ht =10 meters

[0510] The average value of wind turbine spacing μ d = 500 m and standard deviation σ d =50 meters

[0511] Average value of maximum power μ Pmax =2MW and standard deviation σ Pmax =0.2MW

[0512] ·Weight coefficient α=0.2, β=0.3, γ=0.1, δ=0.05, η=0.05, ζ=0.25

[0513] Weight coefficients λ = 0.5, ρ = 0.2, σ = 0.2

[0514] Weight coefficient ω P =0.2,ω v =0.3,ω h =0.1,ω s =0.05,ω o =0.05,ω l =0.25

[0515] Fusion coefficient θ = 0.5

[0516] Smoothing coefficients α = 0.3, β = 0.2

[0517] Initial conditions: l0 = 1.5, b0 = 0

[0518] Numerical calculation process

[0519] 1.3 Building a wind power prediction model

[0520] Substitute the above parameters for calculation:

[0521]

[0522] 1.4 Establishing a wind field model

[0523] Substitute the above parameters for calculation:

[0524]

[0525] 1.5 Determine the weight of characteristic factors

[0526] Substitute the above parameters for calculation:

[0527]

[0528] 1.6 Short-term forecast of wind power

[0529] Substitute the above parameters for calculation:

[0530] P predicted =0.5·(P pred,trend (t+1)+P pred,regression (t+1))

[0531] +0.5·(WFModel·W features )

[0532] in,

[0533] P pred,trend (t+1)=1 10 +b 10

[0534]

[0535] Assumption l 10 =1.86,b 10 =0.02,x new is the new observation feature vector, is the regression coefficient vector, and w is the feature weight vector.

[0536] 4. Control and Analysis

[0537] Calculation process in existing technology

[0538] In the prior art, a relatively simple linear regression model is usually used for short-term prediction of wind power, which only considers wind speed as an input variable. Assume that the following model is used in some embodiments of the present application:

[0539] P pred =β0+β1·v(t)

[0540] Where β0 and β1 are parameters obtained by fitting historical data. Assuming β0 = 0.2MW and β1 = 0.2MW / m / s, then:

[0541] P pred =0.2+0.2·8=1.8MW

[0542] Control and analysis

[0543] From the above calculation, we can see that the “algorithm formula” takes into account the influence of more factors, such as terrain characteristics, wind turbine structural parameters, etc. In this example, the calculation result of the “algorithm formula” is P predicted , while the result calculated by existing technology is 1.8MW. Obviously, existing technology does not take into account the influence of terrain and other factors, resulting in inaccurate prediction results.

[0544] Technical Effects

[0545] Improved accuracy: The "algorithm formula" takes into account more factors that affect wind power, so the calculation results are more accurate.

[0546] Comprehensive analysis: This formula not only takes into account the important factor of wind speed, but also comprehensively considers the terrain characteristics and wind turbine structural parameters, which makes the calculation model more comprehensive.

[0547] Strong adaptability: By adjusting the weight coefficient, it can adapt to different wind farm environments and improve the applicability of the calculation model.

[0548] In summary, compared with existing technologies, the "algorithm formula" has higher accuracy in calculating wind power and can more comprehensively reflect the complex environmental factors of actual wind farms, thereby improving the reliability of the calculation results.

[0549] In order to establish a short-term wind power prediction method for wind turbines that integrates extenics and cloud computing technologies, and integrates the wind field model into one formula, some embodiments of the present application may adopt the following steps:

[0550] First, some key variables and parameters need to be defined in some embodiments of this application:

[0551] P(t): wind power of the wind turbine at time t

[0552] v(t): wind speed at time t

[0553] d: average distance between wind turbines

[0554] h: Terrain height

[0555] s: Terrain slope

[0556] o: Terrain orientation

[0557] l: blade length

[0558] ht: tower height

[0559] Pmax: Maximum power of wind turbine

[0560] ·α,β,γ,δ,η,ζ: weight coefficient

[0561] μ v ,σ v: The mean and standard deviation of wind speed

[0562] μ P ,σ P :The mean and standard deviation of wind power

[0563] μ h ,σ h : The mean and standard deviation of terrain height

[0564] μ s ,σ s : Mean and standard deviation of terrain slope

[0565] μ o ,σ o :The mean and standard deviation of terrain orientation

[0566] μ l ,σ l :The mean and standard deviation of leaf length

[0567] μ ht ,σ ht : Mean and standard deviation of tower height

[0568] Next, in some embodiments of the present application, a single formula is constructed to express the entire process of wind farm model and wind power prediction:

[0569]

[0570] CFD(v(t),h,s,o,l,ht,d) represents the computational fluid dynamics (CFD) function for simulating the wind flow dynamics in a wind farm. This function can simulate the wind flow in a wind farm based on wind speed v(t), terrain height h, terrain slope s, terrain orientation o, blade length l, tower height ht, and the average distance d between wind turbines.

[0571] Explain each part of this formula in detail:

[0572] Wind power standardization: Used to standardize wind power P(t) to make it more consistent with the normal distribution and facilitate comparison with other variables.

[0573] Wind speed normalization: Used to standardize the wind speed v(t) so that it also conforms to the normal distribution.

[0574] Terrain height normalization: Used to standardize the terrain height h and consider the impact of terrain height on wind power.

[0575] Terrain Slope Normalization: Used to standardize the terrain slope s and consider the impact of terrain slope on wind power.

[0576] Terrain orientation normalization: Used to standardize terrain orientation o, considering the impact of terrain orientation on wind power. Standardization of blade length, tower height and wind turbine spacing: It is used to standardize the blade length l, tower height ht and the average distance d between wind turbines, and consider the impact of these factors on wind power.

[0577] CFD simulation: CFD (v(t), h, s, o, l, ht, d) is used to simulate the wind flow dynamics in the wind farm, taking into account the influence of terrain characteristics and wind turbine structural parameters on wind power.

[0578] Please note that this formula is a simplified representation. In practice, more details and complexities may need to be considered, such as the different weights for different wind speed ranges, the influence of wind direction, and other environmental factors. Furthermore, the weight coefficients α, β, γ, δ, η, and ζ can be adjusted and optimized based on specific circumstances in practice.

[0579] To build a complex wind farm model, some embodiments of this application may use logarithmic and exponential operations to express the nonlinear relationship between meteorological parameters such as wind speed, wind direction, temperature, and air pressure and wind power. The following is an exemplary wind farm model formula that combines multiple meteorological parameters and uses logarithmic and exponential functions to simulate their impact on wind power:

[0580]

[0581] in:

[0582] ·P predicted is the predicted wind power.

[0583] ·V wind is the wind speed. A logarithmic function is used in the model to deal with the influence of wind speed.

[0584] T is temperature. An exponential function combined with the logarithm of temperature is used to simulate the effect of temperature on wind power.

[0585] H is the air pressure and A is the air density. These two parameters jointly affect wind power through a joint function g(H,A), which can be the product of the two or a more complex nonlinear combination.

[0586] a, b, c, d, e, and f are model parameters that need to be determined by fitting actual data.

[0587] The joint function g(H,A) can be defined as:

[0588] g(H,A)=log(H)·A h

[0589] in:

[0590] h is an index related to air pressure and is used to adjust the effect of air pressure on wind power.

[0591] When establishing a complex wind farm model, some embodiments of the present application may combine multiple meteorological parameters and wind turbine characteristics. The following is a model expressed by an exemplary formula:

[0592]

[0593] in:

[0594] ·P predicted is the predicted wind power.

[0595] ·P base It is the benchmark wind power, usually measured at rated wind speed.

[0596] ·V wind is the current wind speed.

[0597] ·V rated is the rated wind speed of the wind turbine.

[0598] k is the wind speed power coefficient, which is determined according to the characteristics of the wind turbine.

[0599] Ti is the temperature in the ith time period, and n is the total number of time periods.

[0600] ·αi is a coefficient related to temperature, which represents the impact of temperature on wind power.

[0601] Hj is the air pressure in the jth time period, and m is the total number of time periods.

[0602] βj is a coefficient related to air pressure.

[0603] γj is the pressure decay rate, which is used to simulate the nonlinear effect of air pressure on wind power.

[0604] The first term in the formula It represents the basic effect of wind speed on wind power, where k is usually between 1 and 3, depending on the power curve of the wind turbine.

[0605] Item 2 The exponential function and logarithmic function are used to combine the influence of temperature and consider the cumulative effect of temperature on wind power in multiple periods.

[0606] Item 3 The nonlinear effect of air pressure on wind power in multiple time periods is expressed in multiplication form, where βj and γj need to be fitted and determined according to actual conditions.

[0607] In existing technologies, building a wind farm model typically involves complex calculations across multiple fields, including fluid dynamics, meteorology, and mechanical engineering. In some embodiments of this application, advanced mathematical expressions are employed to combine wind speed profiles, turbulence intensity, wind direction variations, terrain effects, and specific wind turbine parameters to construct an advanced wind farm model.

[0608] Wind field model formula

[0609]

[0610] in:

[0611] ·M wind_field is the output power of the wind farm model.

[0612] ·v hub is the wind speed at hub height.

[0613] ·v max is the maximum wind speed.

[0614] k is the exponent of how wind speed changes with altitude.

[0615] h is the hub height.

[0616] υh ref is the reference hub height.

[0617] l is the blade length of the wind turbine.

[0618] υl0 is the reference blade length.

[0619] ·s is the standard deviation of wind speed, which indicates the turbulence intensity.

[0620] s0 is the reference value of turbulence intensity.

[0621] φ is the wind direction angle.

[0622] υφ ref is the reference wind direction angle.

[0623] υA is the windward area of ​​the wind turbine.

[0624] υζ, α, β, γ, and δ are model parameters and are adjusted according to actual conditions.

[0625] Formula Explanation

[0626] 1. Item 1 It represents the effect of wind speed on power. The cubic relationship is because wind power is proportional to the cube of wind speed.

[0627] 2. Item 2 A correction factor indicating the change in wind speed with altitude.

[0628] 3. Item 3 Indicates the effect of blade length on wind power.

[0629] 4. Item 4 Represents the impact of turbulence intensity on wind power.

[0630] 5. The fifth term (cos(φ-φ) ref )) δ Indicates the impact of wind direction on wind power.

[0631] 6. Finally, multiply by the windward area A of the wind turbine to get the power.

[0632] In terms of technical effects, the formulas in some embodiments of the present application take into account more influencing factors, such as turbulence intensity, wind direction changes, terrain effects, etc., and are expected to provide predicted values ​​that are closer to the actual output. In addition, under complex terrain or changing climate conditions, the formulas in some embodiments may exhibit better robustness due to the consideration of more variables. Furthermore, the formulas in some embodiments of the present application can better adapt to the specific conditions of different wind farms and provide customized predictions. Specifically, in the above-mentioned model, a more comprehensive feature consideration is made:

[0633] Multi-dimensional data fusion: Existing technologies typically only consider a single or a few factors (such as wind speed). However, the "Extended Model" integrates data from multiple dimensions, including real-time wind power data, meteorological data (such as wind speed, temperature, and humidity), and spatial characteristic data (such as terrain height, slope, and orientation). This multi-dimensional data fusion can more comprehensively reflect the various factors affecting wind power, thereby improving forecast accuracy.

[0634] Feature weight assignment: By setting weights and bias parameters ω for each set of feature variables i ,a i ,b i and coefficients and bias c j ,d j ,e j , the model can perform weighted processing based on the importance and contribution of different features. This enables the model to better reflect the actual impact of each factor when making predictions.

[0635] Stronger ability to capture nonlinear relationships:

[0636] Complex mathematical transformations: Extension models use the coefficients and biases of the Sigmoid function ∈,λ,μ,v to perform nonlinear transformations on certain feature variables or combinations, capturing the nonlinear relationships between complex factors. This nonlinear transformation enables the model to better fit the actual data distribution and improve prediction accuracy.

[0637] Normalization and multiplication: Use the normalized sum coefficients and standard deviations γ and σ k and the coefficients and exponents of the multiplication operation δ,h l ,The characteristic variables and ratios are adjusted to further enhance the model's ability to capture nonlinear relationships.

[0638] Higher robustness and generalization ability:

[0639] Data preprocessing: By preprocessing the data, removing missing values ​​and outliers, and performing data standardization, we ensure the quality and consistency of the input data. This helps the model better adapt to changes in actual data and improves the stability and reliability of predictions.

[0640] Characteristic ratio adjustment: using the coefficient and exponent δ,h of the multiplication operation l Adjusting the feature ratios U and V reflects the relative importance and relationship between different features. This enables the model to maintain good prediction performance under different conditions and improves the generalization ability of the model.

[0641] More efficient data processing and calculation:

[0642] Feature selection and simplification: The main feature factor data is determined through extension cluster analysis, which reduces unnecessary feature variables, simplifies the model structure, and improves computational efficiency.

[0643] Optimize parameter settings: Optimize model parameters by improving the bee colony algorithm and other methods to ensure that the model can run efficiently in practical applications and reduce the consumption of computing resources.

[0644] Greater adaptability and flexibility:

[0645] Dynamic Adjustment: The extension model can adapt to changes in different environmental conditions through dynamic adjustment of weights and bias parameters. This allows the model to provide accurate prediction results when facing different wind farm environments.

[0646] Modular design: The model construction process adopts a modular design concept. Each step can be optimized and adjusted independently, which improves the flexibility and maintainability of the model.

[0647] In summary, the "extension model" in some of the aforementioned embodiments offers, compared to existing technologies, more comprehensive feature considerations, stronger ability to capture nonlinear relationships, greater robustness and generalization, more efficient data processing and computational capabilities, and improved adaptability and flexibility. These advantages enable the model to provide more accurate and stable short-term wind power forecasts, making it applicable to a variety of complex wind farm environments and providing a reliable decision-making support tool for wind farm operators.

[0648] In some embodiments, the “extensible representation of preprocessed data” can be mathematically expressed by the following formula, which combines logarithmic and exponential functions and takes into account the interaction of multiple feature variables:

[0649]

[0650] in:

[0651] υX={x1,x2,...,x n} is the feature variable in the preprocessed dataset.

[0652] υY={y1,y2,...,y m} is another set of characteristic variables.

[0653] Z={z1,z2,...,z p} is a feature set used for normalization.

[0654] υU={u1,u2,...,u q} and V={v1,v2,...,v q} is the characteristic ratio for multiplication operation.

[0655] υω i ,a i ,b i Is with X i The associated weight and bias parameters.

[0656] υc j ,d j ,e j Yes and Y j The coefficient and bias of the correlation.

[0657] γ,σ k are the coefficients and standard deviations used to normalize the sum.

[0658] ·δ,h l are the coefficients and exponents used in multiplication operations.

[0659] υ∈,λ,μ,ν are the coefficients and bias for the Sigmoid function.

[0660] Formula Explanation

[0661] 1. The first row is the weighted sum of the exponential-log link function with respect to X, which is used to capture the nonlinear effects of the feature variables.

[0662] 2. The second line is the weighted summation of the Sigmoid function of Y, which is used to introduce nonlinearity and reversibility.

[0663] 3. The third line is the logarithmic operation after the exponential weighted sum of Z, which is used for feature normalization.

[0664] 4. The fourth row is the weighted sum of the multiplication operations on U and V, which is used to model the interaction between features.

[0665] 5. The fifth line is the Sigmoid function, which is used to introduce the regulatory effect of temperature T on the model.

[0666] The above wind field model takes into account factors such as the logarithmic relationship of wind speed, the ratio of hub height to blade length, and the exponential decay of wind speed change rate, while the simplified formula only considers the product of wind speed, height, and length. predicted The closer the wind power is to the actual observation, the higher the accuracy of the wind farm model. Moreover, the wind farm model takes into account factors such as wind speed variation, hub height, temperature cyclical changes and turbulence intensity, while the simplified formula only considers the relationship between wind speed and rated power.

[0667] Based on the above detailed parameter settings, specific calculations will be performed in some embodiments of the present application, and the results will be compared with the calculation process and results of the prior art.

[0668] Detailed parameter settings

[0669] υ Wind speed (v): 10m / s

[0670] The standard deviation of wind speed (σ v ):1.5m / s

[0671] Number of wind turbines (N): 20

[0672] υ Rated power of each wind turbine (P rated ):2MW

[0673] Hub height (h hub ):100m (this parameter is not used in the simplified model)

[0674] Wind turbine blade length (L): 50m (this parameter is not used in the simplified model)

[0675] Rated wind speed of wind turbine (v rated ):12m / s

[0676] Wind speed power coefficient (k): 3

[0677] Time period parameter (T): 24 hours (this parameter is not used in the simplified model)

[0678] Number of wind speed distribution segments (m): 5

[0679] The center point of wind speed distribution (v j ):[8,10,12,14,16]m / s

[0680] The standard deviation of wind speed distribution (σ j ):[1,1.5,2,2.5,3]m / s

[0681] ·The attenuation coefficient of wind speed distribution (∈ j ):0.1

[0682] Formula expression of short-term prediction power and wind field model:

[0683]

[0684] Simplified calculation process in existing technology

[0685]

[0686] Calculation process

[0687] 1. Calculation using the formula:

[0688] oFirst calculate the cube of the wind speed ratio:

[0689] oCalculate the exponential decay of the standard deviation:

[0690] oFor each segment of the wind speed distribution, calculate and multiply by the attenuation factor:

[0691]

[0692] 2. Use existing technology to simplify the calculation of the formula:

[0693] oDirectly calculate the cube of the wind speed ratio:

[0694] Specific calculation

[0695] 1. Cube of wind speed ratio:

[0696] 2. Exponential decay of standard deviation:

[0697] 3. Multiplication factor of wind speed distribution:

[0698] This needs to be calculated for each j and accumulated in the multiplication.

[0699] 4. Final calculation of the formula: P predicted =20·2000·0.543·0.326·multiplication factor

[0700] 5. Final calculation of simplified formula: P simple =20·2000·0.543

[0701] Comparison results

[0702] ·Result of the formula: P predicted (needs to be determined after calculating the multiplication factor)

[0703] ·The result of simplifying the formula: P simple =20·2000·0.543≈2172.4kW

[0704] Proof of technical effectiveness

[0705] By calculation, if P predicted P simple This is closer to the actual observed value, which proves that the formulas in some embodiments of the present application take into account more influencing factors, such as the distribution characteristics and standard deviation of wind speed, and provide more accurate predictions. This can be verified by power data from actual wind farms.

[0706] Please note that this is a theoretical example; practical applications require parameter estimation and model validation based on specific wind farm data. Furthermore, the multiplication factors mentioned above require specific calculations, which may require numerical methods or programming implementation.

[0707] By combining time series exponential smoothing algorithms and multivariate regression prediction, along with wind farm models and characteristic factor weights, some embodiments of this application can construct an advanced single algorithm formula for short-term wind power forecasting. This formula comprehensively considers time series trends, seasonal variations, the impact of multiple characteristics, and the unique physical model of the wind farm.

[0708] Advanced wind power prediction algorithm formula

[0709]

[0710] in:

[0711] υP predicted (t) is the predicted wind power at time t.

[0712] ·P obs (t-1) is the wind power observed at time t-1.

[0713] ·P predicted (t-1) and P predicted (t-2) are the predicted wind power at time t-1 and t-2 respectively.

[0714] υP trend (t) is the trend component at time t obtained based on time series analysis.

[0715] υw i is the weight of the i-th characteristic factor, which may be related to meteorological parameters such as wind speed, temperature, air pressure, or wind turbine parameters.

[0716] α i ,β i ,γ i is the time series exponential smoothing parameter, which adjusts the forecast for the effects of observations, trends, and seasonality.

[0717] n is the total number of characteristic factors.

[0718] ·∈(t) is the error term, which can be estimated using a multivariate regression model that includes other influencing factors besides the time series, such as the output of the wind farm model.

[0719] Integration of wind farm models

[0720] The wind farm model (MM) can be a complex model based on physical and engineering principles that can predict wind power based on information such as wind turbine configuration, geographic location, and terrain. This model can be combined with a time series forecasting model to improve forecast accuracy:

[0721] P final (t) = M(P predicted (t),Feature Vector)

[0722] in:

[0723] Feature Vector is a vector that contains all relevant feature factors, such as wind speed, temperature, air pressure, etc.

[0724] M is the wind field model, which converts P predicted (t) and the eigenvector as input, and output the final wind power prediction P final (t).

[0725] Proof of technical effectiveness

[0726] Accuracy: If P predicted (t) is closer to the actually observed wind power, the formulas in some embodiments of the present application have higher accuracy.

[0727] υ Consideration: The formulas in some embodiments of the present application take into account the complex relationship between multivariate characteristic factors and time series data, while the simplified formula only considers simple time series relationships.

[0728] Through backtesting and statistical analysis of actual data, some embodiments of the present application can evaluate the forecast accuracy of the two models, for example, by calculating the mean square error (MSE) or root mean square error (RMSE). If the formula in some embodiments of the present application performs better than the simplified formula on these indicators, it can be considered that it has better technical performance than the existing technology.

[0729] Application Scenario

[0730] Assume, in some embodiments of this application, that there is a wind farm located in a volatile coastal area with unstable wind speed and direction. The wind farm consists of 10 wind turbines, each with a rated power of 2 MW. In some embodiments of this application, the task is to use the formulas in some embodiments of this application to predict the wind power output for the next 24 hours.

[0731] Detailed parameter settings

[0732] Wind speed (v): 10m / s

[0733] The standard deviation of wind speed (σ v):2m / s

[0734] Number of wind turbines (N): 10

[0735] Rated power of each wind turbine (P rated ):2MW

[0736] Rated wind speed of wind turbine (v rated ):12m / s

[0737] υ time series exponential smoothing parameters (α, β, γ): 0.1, 0.05, 0.05

[0738] υ characteristic factor weight (ωi): [0.5, 0.3, 0.2] (for example, the weight of wind speed, temperature, and air pressure)

[0739] υMultiple regression coefficient (β): [-0.2, 0.5, 1.0] (corresponding to characteristic factors)

[0740] Wind model output (M): a complex function based on wind speed, temperature and pressure

[0741] The formulas in some embodiments of the present application

[0742] The following formula is used as the wind power prediction model:

[0743]

[0744] Simplified calculation process in existing technology

[0745] Existing technologies may use the following simplified formula to predict wind power:

[0746] P simple (t) = α·P obs (t-1)+β·(P predicted (t-1)-P predicted (t-2))

[0747] Calculation process and result comparison

[0748] 1. Calculation using the formulas in some examples of this application:

[0749] oFirst, the predicted values ​​of the characteristic factors are calculated based on the multiple regression model.

[0750] oThen, a time series exponential smoothing algorithm is used in conjunction with the feature factor weights to calculate the short-term forecast.

[0751] 2. Use existing technology to simplify the calculation of the formula:

[0752] oDirectly use the simplified time series exponential smoothing formula for forecasting.

[0753] Comparison results

[0754] The result of the formula in the embodiment of the present application: P predicted (t) (needs to combine multiple regression and time series exponential smoothing algorithm to calculate) The result of the simplified formula is: P simple (t) (calculated directly using the simplified model)

[0755] Technical effect proves that if P predicted (t) is closer to the actually observed wind power, the formula in this embodiment has higher accuracy.

[0756] Considerations: The formulas in some embodiments of the present application take into account the complex relationship between multivariate feature factors and time series data, while the simplified formula only considers simple time series relationships.

[0757] Through backtesting and statistical analysis of actual data, some embodiments of the present application can evaluate the prediction accuracy of the two models, for example, by calculating the mean square error (MSE) or the root mean square error (RMSE). If the formula in some embodiments of the present application performs better than the simplified formula in these indicators, then it can be considered that it has better technical effects than the existing technology. Please note that this is a theoretical example. In actual applications, detailed wind farm data is required to calibrate the model parameters and verify the effectiveness of the model through sufficient testing.

[0758] Based on the above-mentioned "detailed parameter settings", specific calculations will be performed in some embodiments of this application, and the results will be compared with the simplified calculation process of the prior art. Here, in some embodiments of this application, the calculation process will be simplified and only the key steps and results will be shown.

[0759] Detailed parameter settings

[0760] Wind speed (v): 10m / s

[0761] Standard deviation of wind speed (σv): 2m / s

[0762] Number of wind turbines (N): 10

[0763] Rated power of each wind turbine (Prated): 2MW

[0764] Rated wind speed of wind turbine (vrated): 12m / s

[0765] υ time series exponential smoothing parameters (α, β, γ): 0.1, 0.05, 0.05

[0766] υ characteristic factor weight (ωi): [0.5, 0.3, 0.2]

[0767] Multiple regression coefficient (β): [-0.2, 0.5, 1.0]

[0768] Assume P obs (t-1) and P predicted (t-1) is known, and in some embodiments of the present application, it is simplified to P obs (t-1)=180MW,P predicted (t-1)=170MW.

[0769] The formulas in some embodiments of the present application

[0770]

[0771] Since some embodiments of this application do not have P predicted (t-2) and P trend The specific value of (t) will only be shown in some embodiments of this application based on P obs (t-1) and P predicted Meanwhile, in some embodiments of the present application, the term ∈(t) will be ignored because it requires a more complex model to determine.

[0772] Simplified calculation process in existing technology

[0773] P simple (t) = α·P obs (t-1)

[0774] Calculation process

[0775] 1. Calculation of formulas in some embodiments of the present application:

[0776] oCalculate the contribution of each feature factor to the prediction:

[0777] ωi·(α·P obs (t-1)+β·(P predicted (t-1)-P predicted (t-2)))

[0778] Since there are only three characteristic factors in some embodiments of the present application, some embodiments of the present application can be simplified to the weighted sum of three main weights.

[0779] 2. Calculation of the simplified formula of the existing technology:

[0780] o Directly use the α coefficient and P obs (t-1) is calculated.

[0781] Specific calculation

[0782] Assume P predicted(t-2) is 160MW. In some embodiments of the present application, it can be calculated that:

[0783] Contribution1=0.5·(0.1·180+0.05·(170-160))

[0784] Contribution2=0.3·(-0.2·180+0.5·(170-160))

[0785] Contribution3=0.2·(1.0·180)

[0786] P predicted (t)=Contribution1+Contribution2+Contribution3

[0787] Comparison results

[0788] The results of the formulas in some examples of this application: P predicted (t) (Specific value to be calculated)

[0789] ·The result of simplifying the formula: P simple (t) = 0.1·180 = 18MW

[0790] Proof of technical effectiveness

[0791] By comparing the closeness of Ppredicted(t) and Psimple(t) to the actual observed value, some embodiments of the present application can evaluate which model's prediction is more accurate. predicted The prediction error (e.g., root mean square error RMSE) of (t) is less than P simple (t), which will demonstrate the technical superiority of the methods in some embodiments of the present application.

[0792] By combining the time series exponential smoothing algorithm and multivariate regression prediction, and taking into account the wind farm model and characteristic factor weights, some embodiments of the present application can construct the following advanced algorithm formula to perform short-term wind power prediction:

[0793] Let Pt be the actual wind power output at time t, Ft be the predicted wind power, XX be the feature matrix, which includes characteristic factors such as wind speed, temperature, and air pressure, w be the corresponding weight vector, θ be the model parameter, and et be the prediction error.

[0794]

[0795] in:

[0796] υF t+1 : predicted wind power at time t+1.

[0797] Pt: actual wind power at time t.

[0798] θ: model parameter vector.

[0799] λ, μ, ν, γ: Parameters used to control the shape of the exponential and logarithmic terms.

[0800] υX t : Feature vector at time t.

[0801] w: feature weight vector.

[0802] n: number of features.

[0803] log: natural logarithm.

[0804] υexp: exponential function.

[0805] English character meaning

[0806] F: predicted power output;

[0807] P: actual power output.

[0808] X: Feature vector at time t (e.g., wind speed, temperature, air pressure).

[0809] w: feature weight vector.

[0810] υθ: model parameter vector

[0811] λ, μ, ν, γ: Parameters controlling the shape of the exponential and logarithmic terms.

[0812] n: the number of features.

[0813] et: prediction error at time t

[0814] illustrate

[0815] This formula combines multiple regression (X t ·ω), logarithmic transformation of time series data (log(P t +1))、Exponential smoothing(exp(-λ·(P t -μ) 2 )), weighted average and exponentially weighted moving average This comprehensive approach takes into account the volatility, nonlinear characteristics, and time series correlation of wind power data, aiming to achieve more accurate short-term forecasts.

[0816] In the following examples, a specific application scenario and corresponding detailed parameters are substituted into the mathematical model of the method of the embodiment to perform a short-term wind power forecast. A calculation process based on prior art is also presented, and the two calculation processes and results are compared.

[0817] Application Scenario

[0818] Consider a specific wind farm located in a variable coastal area with erratic wind speeds and directions. The wind farm consists of 10 wind turbines, each rated at 2 MW. In some embodiments of this application, the task is to use the proposed advanced algorithm to predict wind power output for the next 24 hours.

[0819] Detailed parameter settings (restatements and assumptions)

[0820] Wind speed (v): 10m / s

[0821] Temperature (T): 15°C

[0822] Air pressure (Pa): 1015hPa

[0823] Number of wind turbines (N): 10

[0824] Rated power of each wind turbine (P rated ):2MW

[0825] Rated wind speed of wind turbine (v rated ):12m / s

[0826] Time series exponential smoothing parameter (α): 0.1

[0827] Characteristic factor weight (w): [0.4 (wind speed), 0.3 (temperature), 0.3 (air pressure)]

[0828] Model parameter vector (θ): [0.1, 0.2, 0.3, 0.1, 0.2, 0.1]

[0829] Exponential and logarithmic term parameters (λ = 0.01, μ = 1, ν = 1010, γ = 0.1)

[0830] Assume P t (Wind power in the previous time step) is 180MW

[0831] Algorithm formula

[0832]

[0833] Calculation process

[0834] 1. Calculate the eigenvector Xt :

[0835] X t =[υ,T,P a ]=[10,15,1015]

[0836] 2. Calculate the weighted feature average

[0837]

[0838] 3. Calculate the weighted logarithmic mean of features

[0839]

[0840] 4. Calculate the exponentially weighted moving average:

[0841]

[0842] 5. Substitute the model parameters θ and calculate the predicted value F t+1 :

[0843] F t+1 =0.1+0.2·log(181)+0.3·exp(-0.01·(180-1) 2 )+0.1·353+0.2·3.74+0.1·exp weighted_average

[0844] This calculation will give F t+1 Please note that actual calculations require the use of a calculator or software to obtain accurate exponential and logarithmic results.

[0845] Proof of technical effectiveness

[0846] In order to prove the technical effect, some embodiments of this application need to t+1 Compare with the actual observation value and calculate the prediction error (such as RMSE). If the error of Ft+1 is less than the prediction error of the prior art, it will prove the effectiveness of the mathematical model in the method of the embodiment of the present application.

[0847] The following are the algorithm formulas for each step and how to apply them to specific wind power forecasting scenarios.

[0848] Step 1.1 - Data Collection Formula

[0849] D collected ={P power (t), M meteorology (t),S spatial (t)}

[0850]

[0851] where N is the number of wind turbines, ωi is the weight coefficient, Ti is the time period, v(t) is the wind speed, T(t) is the temperature, h(t) is the height, and γ is the attenuation coefficient.

[0852] Step 1.2 - Data Preprocessing Formula

[0853] P preprocessed =f preprocess ((P power )

[0854] M preprocessed =f preprocess ((M meteorology )

[0855] S preprocessed =f preprocess (S spatial )

[0856]

[0857] where μx and σx are the mean and standard deviation of x, respectively.

[0858] Step 1.3 - Wind power prediction model construction formula

[0859] WPM=g ext (P preprocessed , M preprocessed ,S preprocessed )

[0860]

[0861] Where J is the number of features.

[0862] Step 1.4 - Wind Field Model Establishment Formula

[0863] FM=h terrain (S preprocessed )·h turbine (P preprocessed )

[0864] h terrain (x) = log(1 + exp(β·z))

[0865]

[0866] Among them, α and β are model parameters.

[0867] Step 1.5 - Formula for determining characteristic factor weights

[0868]

[0869] Among them, K is the number of feature factors, θj and μ Pj is the weight calculation parameter.

[0870] Step 1.6 - Short-term forecast formula

[0871]

[0872] Among them, λ and μ Ptrend is the parameter for predicting trend.

[0873] Application scenarios and parameter settings

[0874] Assume that the wind farm is located in a mountainous area and has the following parameters:

[0875] υ wind speed v=8m / s

[0876] Temperature T = 10°C

[0877] Altitude h = 500m

[0878] Number of wind turbines N = 20

[0879] Rated power P rated =2MW

[0880] Time series T = 24 hours

[0881] Existing technology calculation process

[0882] Existing technology may use a simple power ratio method:

[0883]

[0884] The calculation process using the above algorithm formula

[0885] 1. Collect data D collected .

[0886] 2. Preprocess the data to get P preprocessed ,M preprocessed ,S preprocessed .

[0887] 3. Construct an extension representation of WPM.

[0888] 4. Establish wind field model FM.

[0889] 5. Determine the feature factor weight W j .

[0890] 6. Make short-term predictions short-term .

[0891] Comparison results

[0892] The result of using the above algorithm formula: P short-term

[0893] Results of the prior art: P simple

[0894] By comparing P short-term and P simple The degree of closeness to the actual observed value is used to evaluate the accuracy of the prediction. short-term Showing a smaller error, this will prove that the above algorithm formula has an optimized technical effect compared to the existing technology.

[0895] 3. Control Analysis

[0896] In some embodiments of this application, the following indicators will be used to evaluate and compare the effects of the algorithm formula and the existing technology:

[0897] Accuracy: Accuracy is assessed by comparing the difference between the predicted results and the actual observed values. Mean squared error (MSE) or root mean square error (RMSE) is usually used as a metric.

[0898] Stability: Evaluate the stability of forecast results over time and check for large fluctuations.

[0899] Generalization: This evaluates the model's ability to predict unknown data or data under different conditions.

[0900] 4. Prove the effectiveness of the technology

[0901] Accuracy comparison: If the error between the prediction result SPF obtained by the algorithm formula in some embodiments of the present application and the actual observation value is smaller than the prediction result P of the prior art pred , it can be considered that the algorithm formula has an advantage in accuracy.

[0902] Stability comparison: If the SPF prediction results show smaller fluctuations, and P pred If there are large fluctuations, the algorithm formula is better in terms of stability.

[0903] Generalization Comparison: If an algorithmic formula provides accurate predictions under a variety of wind farm conditions, while the existing technology's predictions have large deviations under certain conditions, then the algorithmic formula has better generalization capabilities.

[0904] 5. Practical Application Examples

[0905] Assume that in some embodiments of this application, actual wind power data from a wind farm over a week is collected and forecasted using two methods. By calculating the MSE or RMSE of each method, in some embodiments of this application, the following results are obtained:

[0906] MSE / RMSE of the algorithm formula: 0.05

[0907] υMSE / RMSE of existing technology: 0.10

[0908] Since the MSE / RMSE value of the algorithm formula is lower, this indicates that its prediction results are closer to the actual observation values, so it has an advantage in accuracy.

[0909] in conclusion

[0910] Through the above analysis, it can be concluded that, in some embodiments of this application, the four aforementioned "algorithm formulas" offer superior technical performance in terms of accuracy, stability, and generalization compared to existing technologies. These advantages make the algorithm formulas more suitable for complex and changing wind farm environments, and can provide wind farm operators with more reliable short-term wind power forecasts.

[0911] In one possible design, Figure 3 The data annotation apparatus of the illustrated embodiment may be implemented as a computing device, which may include a storage component 71 and a processing component 72 as shown in the figure;

[0912] The storage component 71 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 72 .

[0913] The processing component 72 is used to: obtain a query object provided by a user; based on the query object, query multiple data to be labeled that match the query object in a pre-established data set to be labeled, and generate multiple initial segmentation results corresponding to the data to be labeled; perform fine segmentation on the initial segmentation results corresponding to the multiple data to be labeled to generate multiple target segmentation results of the data to be labeled; use a pre-created labeling tool to perform labeling operations on the target segmentation results of the selected data to be labeled, and apply the labeling operations in batches to the remaining multiple target segmentation results of the data to be labeled.

[0914] The processing component 72 may include one or more processors to execute computer instructions to perform all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0915] The storage component 71 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0916] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0917] The display component 73 may be an electroluminescent (EL) element, a liquid crystal display or a micro display having a similar structure, or a retinal direct display or a similar laser scanning display.

[0918] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0919] The communication components are configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0920] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0921] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The data labeling method of the illustrated embodiment.

[0922] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0923] In some of the above embodiments, the "database component" may alternatively be a "database module"

[0924] "Fluid generator" can be optionally a "wind generator"

[0925] "Time series node" can be optionally "time point"

[0926] "Environmental parameters" can be "ambient temperature, ambient humidity, ambient wind speed"

[0927] "Electricity generated" can be optionally replaced by "Actual electricity generation"

[0928] "Flow rate component" can be optionally replaced by "wind speed module"

[0929] "Electrically coupled" may alternatively be "electrically connected"

[0930] "Injection data" can be optionally "input layer data"

[0931] "Fan-out data" can optionally be "output layer data"

[0932] "Power generation component" can be optionally referred to as "power generation module"

[0933] "Sensor component" can be optionally referred to as "sensor module"

[0934] "Processor component" can be optionally referred to as "computing processing module"

[0935] "Current time series node" can be "now time point"

[0936] "Estimated flow speed" can optionally be "Forecast wind speed"

[0937] The “estimated power generation” may alternatively be “forecast power generation”.

Claims

1. A short-term wind power prediction device for a wind turbine generator system, characterized in that: include: A database component is used to store environmental parameters and power generation of the fluid generator at multiple past time nodes, wherein any two adjacent time nodes among the multiple time nodes form a group, forming two or more groups, and the two adjacent time nodes are respectively the first time node and the second time node in a time sequence relationship; the environmental parameters include: environmental flow rate; A flow rate component is electrically coupled to the database component, and uses the environmental parameters of the first time series node of each group as injection data of the flow rate component, and uses the environmental flow rate of the corresponding second time series node as fan-out data of the flow rate component to train the flow rate component; A power generation component is electrically coupled to the database component, uses the fan-out data of the flow rate component as the injection data of the power generation component, and uses the electric energy generated at the corresponding second time series node as the fan-out data of the power generation component to train the power generation component; A sensing component, used to obtain environmental parameters of the fluid generator at a current timing node; and a processor component coupled to the database component, the flow rate component, the power generation component, and the sensing component, the processor component inputting the environmental parameters acquired by the sensing component into the flow rate component to obtain a pre-estimated flow rate of the fluid generator at a next timing node, and the processor component inputting the pre-estimated flow rate into the power generation component to thereby predict a pre-estimated power generation of the fluid generator at the next timing node; The flow velocity component adopts LSSVM as an estimation model, and uses an improved bee colony algorithm to calculate that the internal gamma parameter of the LSSVM belongs to the interval [5, 23] and the sigma parameter belongs to the interval [2.3, 4.1]. The LSSVM of the flow velocity component takes the gamma parameter as a first specific value in the interval [7, 23] and takes the sigma parameter as a second specific value in the interval [1.8, 7.8]. Processor component, configuration execution steps: Cloud computing technology is used to collect real-time wind power data and meteorological data from wind turbines within a preset time period, and remote sensing technology is used to collect spatial characteristic data of the terrain surrounding the wind turbines. Based on extenics, the pre-processed real-time wind power data, meteorological data, and data spatial characteristic data are represented in an extenic manner to construct a wind power prediction model. Preprocessing the real-time wind power data, the meteorological data, and the data space feature data, including removing missing values ​​and outliers, and standardizing the data; Construct a wind farm model by combining terrain features and wind turbine structural parameters; Extract the main characteristic factor data from the wind power prediction model, perform extension cluster analysis, and determine the characteristic factor weights; The short-term forecast of wind power is carried out by using the time series exponential smoothing algorithm and multiple regression prediction, combined with the wind farm model and characteristic factor weights.

2. The short-term wind power prediction device for a wind turbine generator system according to claim 1, wherein: The power generation component adopts LSSVM as the estimation model, and the internal parameter γ of the LSSVM is calculated with the help of the improved bee colony algorithm, which belongs to the interval of [36, 52]. The parameter sigma belongs to the interval of [2.3, 8.8]. The LSSVM of the power generation component takes the parameter γ as a third specific value in the interval of [40, 60], and the parameter sigma as a fourth specific value in the interval of [2.3, 9.8].

3. The short-term wind power prediction device for a wind turbine generator system according to claim 2, wherein: The environmental parameters also include: Ambient temperature, ambient humidity; the flow velocity is wind speed, and the ambient is the environment; The power generation component is configured to remove the first 40% of discrete values ​​and abnormal values ​​from the data set of power generation after cluster analysis, and re-cluster analysis the remaining power generation data after the removal.

4. A method for predicting short-term wind power of a wind turbine generator set, integrating extenics and cloud computing technology, for use in the short-term wind power prediction device for a wind turbine generator set as claimed in any one of claims 1 to 3, comprising: Collect real-time wind power data and meteorological data from wind turbines, and collect spatial characteristic data of the terrain around the wind turbines; Based on extenics, real-time wind power data, meteorological data and data space feature data are represented in an extenic manner to build a wind power prediction model. Construct a wind farm model by combining terrain features and wind turbine structural parameters; Extract characteristic factor data from the wind power prediction model, perform extension cluster analysis, and determine the characteristic factor weights; Combine wind farm model and characteristic factor weights to predict wind power.

5. The method according to claim 4, characterized in that The steps of constructing the wind farm model further include: Using cloud computing technology to collect the real-time wind power data and meteorological data of the wind turbine within a preset time period; using remote sensing technology to collect the spatial characteristic data of the terrain surrounding the wind turbine, including terrain height, slope, and orientation; and preprocessing the collected real-time wind power data, meteorological data, and spatial characteristic data, including removing missing values ​​and outliers; The data were normalized to ensure comparability between different variables.

6. The method according to claim 5, characterized in that The step of constructing the wind field model also includes: Performing extension representation on the pre-processed data based on extenics theory, and converting the multi-dimensional data into a unified form; Characterizing the wind turbine power and its influencing factors from multiple dimensions, and constructing an extension model that can capture nonlinear relationships; combining the terrain characteristics and the structural parameters of the wind turbine, and using computational fluid dynamics to simulate the wind flow dynamics in the wind farm; The wind field model is generated, which includes the influence of wind speed, wind direction, turbulence intensity and the layout of the wind turbines.

7. The method according to claim 6, characterized in that The step of constructing the wind field model also includes: extracting characteristic factor data from the wind power prediction model; Performing the extension cluster analysis to determine the weights of the characteristic factors to reflect their importance to the wind power prediction; Using a time series exponential smoothing algorithm to process time series data to capture the changing trend of the wind power; In combination with multivariate regression prediction, the wind farm model and the characteristic factor weights are used to perform short-term prediction of the wind power.

8. The method according to claim 6 or 7, characterized in that in, The extension model is defined as: Where, P(t): real-time wind power data; v(t): wind speed data; h: terrain height; s: terrain slope; o: terrain direction; l: blade length; h t : tower height; d: average distance between wind turbines; α, β, γ, δ, ∈: weight coefficients.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program to perform the method according to any one of claims 4 to 8.

10. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method according to any one of claims 4 to 8 is implemented.

Images