Wind power short-term power prediction method in low-temperature cold-tide weather

Abstract

The invention discloses a wind power short-term power prediction method in low-temperature cold-tide weather. In the feature extraction and selection stage, wind power plant surrounding meteorological data is collected, feature correlation values are calculated, and key features are screened. A support vector machine (SVM) model and a radial basis function (RBF) kernel are adopted to establish a meteorological and power mapping model, and the model is optimized through training, verification and evaluation index analysis; and for low-temperature cold-wave weather, comprehensive judgment conditions are set, model parameters are adaptively adjusted according to prediction errors, and model adaptability is improved. Meanwhile, an updating mechanism driven by real-time data is established, data are collected in real time, a model is incrementally trained, feedback adjustment is performed after performance evaluation, and a closed-loop optimization system is formed. According to the method, the accuracy of wind power prediction in the low-temperature cold-tide weather is effectively improved, the model adaptability and stability are enhanced, powerful support is provided for wind power integration and power grid dispatching, the wind power operation efficiency is improved, and the development of the new energy power generation field is promoted.

Description

Technical Field

[0001] The present invention relates to the field of renewable energy power generation, and is particularly focused on the accurate prediction technology of wind power under low temperature and cold wave weather conditions, aiming to improve the grid connection stability and operation efficiency of wind power under extreme weather conditions. Background Art

[0002] In the field of new energy power generation, wind power, as an important renewable energy power generation method, has significant volatility and intermittency in power output. This characteristic is mainly due to the instability of wind energy itself, which is affected by a variety of meteorological factors such as wind speed, wind direction, temperature, air pressure, etc. In the existing technology, although there are many wind power prediction methods, there are still many defects when facing extreme weather such as low temperature and cold wave.

[0003] Traditional wind power forecasting methods are often based on relatively simple statistical models or single machine learning models. When processing data under complex meteorological conditions, the models lack adaptability and accuracy. Some methods based on the average of historical data or simple linear regression cannot effectively capture the nonlinear relationship between meteorological factors and the special impact mechanism of low temperature and cold wave weather on wind power equipment and power output. In low temperature and cold wave weather, the operating performance of wind turbines will change. For example, blade icing may cause a decrease in aerodynamic performance, which in turn affects power output. However, existing technologies make it difficult to accurately incorporate these low temperature and cold wave weather characteristics into the forecasting model, resulting in a large deviation between the forecast results and the actual power output.

[0004] In addition, existing forecasting methods usually lack the ability to effectively utilize real-time data and dynamically adjust. In wind power forecasting, real-time meteorological data and power output data are constantly changing, but most methods cannot be quickly updated and optimized based on new data after the model is established, making the model slow to respond to sudden weather changes (such as the sudden onset of low temperatures and cold waves), and unable to accurately predict the changing trend of wind power, which in turn affects the stability of wind power grid connection and the dispatching decision of the power system.

[0005] In response to these problems, the present invention proposes a wind power short-term power forecasting method that comprehensively considers multiple meteorological factors, has adaptability to low temperatures and cold waves, and has real-time data-driven update capabilities. The method aims to overcome the shortcomings of the existing technology, improve the accuracy and reliability of wind power forecasting under extreme weather conditions, and provide strong guarantees for the stable development of renewable energy power generation. Summary of the invention

[0006] In view of the above deficiencies in the prior art, the purpose of the present invention is to overcome the problem of inaccurate wind power prediction in low temperature cold wave weather in the prior art, and provide strong support for power grid dispatching.

[0007] In order to achieve the above objectives, the present invention adopts the following technical solutions:

[0008] The method for short-term wind power prediction under low temperature cold wave weather includes the following steps:

[0009] S1. Feature extraction and selection: collect meteorological data around the wind farm, including temperature, wind speed, wind direction, etc., and pay special attention to the characteristics of low temperature cold wave weather, such as sudden temperature drop and continuous low temperature. Extract key features related to wind power output from meteorological data and select the features that have the greatest impact on wind power output.

[0010] S2. Establish a meteorological and power mapping model. Establish a meteorological and power mapping model based on the features extracted in S1. The meteorological and power mapping model can map the conditions of meteorological characteristics to wind power output, thereby converting meteorological characteristics into prediction results of wind power output;

[0011] S3. Adaptive adjustment of the model to low temperature and cold waves. Based on the meteorological and power mapping model established in S2, the parameters of the meteorological and power mapping model are dynamically adjusted according to the changes in real-time meteorological data, especially the parameters during low temperature and cold wave weather.

[0012] S4. Real-time data-driven model update. A real-time data-driven model update mechanism is established based on the data collected by S1 and the meteorological and power mapping model established by S2, so that the model can dynamically adjust the prediction according to the latest meteorological data and power output. The real-time update mechanism ensures that the model can quickly respond to environmental changes and maintain the accuracy of the prediction.

[0013] Further, the S1 includes:

[0014] S11. Data collection: obtaining meteorological data, such as temperature, wind speed, wind direction, air pressure and humidity, through a meteorological monitoring station, and recording wind power output data corresponding to the meteorological data;

[0015] S12. Feature calculation:

[0016] Calculate the temperature change rate. Use the sliding window method to calculate the temperature change rate. The formula is:

[0017]

[0018] Where ΔTt is the temperature change rate at the current time t; Tt is the temperature at the current time t; is the temperature Δt1 minutes ago; at the same time, the sliding average of the temperature change rate in the past n1 time intervals is calculated

[0019]

[0020] Calculate the wind speed trend and the first-order difference of wind speed Vt is the current wind speed, is the wind speed value at the time interval Δt1 before the current time t;

[0021] And calculate the sliding average of the wind speed change trend in the past n1 time intervals

[0022]

[0023] Wind direction feature processing: convert wind direction data into vector components in the Cartesian coordinate system, i.e. x = cos(θ) and y = sin(θ), and calculate the change in wind direction vector and Among them, xt represents the value of the wind direction vector in the x direction at time t, represents the value of the wind direction vector in the x direction at the moment t-Δt1, Δx represents the change of the wind direction vector in the x direction at the moment t relative to the previous moment t-Δt1; represents the value of the wind direction vector in the y direction at the moment t, represents the value of the wind direction vector in the x direction at the moment t-Δt1, and Δy represents the change of the wind direction vector in the x direction at the moment t relative to the previous moment t-Δt1;

[0024] Pressure variation characteristics, calculate the first-order difference of pressure:

[0025]

[0026] Among them, ΔPa is the change in air pressure, Pa,t is the air pressure at the current time t, is the air pressure at time t-Δt1;

[0027] Humidity change characteristics, calculate the first-order difference of humidity:

[0028]

[0029] Among them, ΔH is the humidity change, Ht is the humidity at the current time t, is the humidity at time t-Δt1;

[0030] S13. Preliminary screening of features, calculation of the simple linear correlation coefficient r between each feature and wind power output, the formula is:

[0031]

[0032] Pt is the wind power at time t, is the power average value, Ft is the meteorological characteristic value, is the feature average, m is the number of samples; the features whose absolute value of the correlation coefficient r is greater than the threshold are selected as the screening results.

[0033] Furthermore, the establishment of the meteorological and power mapping model includes:

[0034] S21. Model selection and architecture design, the support vector machine (SVM) model is selected to establish the relationship between meteorology and power mapping, and the radial basis function (RBF) kernel is used; the expression of the model is:

[0035]

[0036] Among them, xi is the input meteorological feature vector, xi is a vector formed by combining multiple related types of feature values ​​Ft, xi contains multiple meteorological feature information meaningful to wind power prediction in vector form; x is another sample vector, which has the same structure as xi and also contains multiple meteorological feature values, which is used to represent the meteorological state at another time or condition; yi is the wind power output corresponding to the meteorological feature vector xi, sgn is the sign function, αi is the Lagrange multiplier, K(xi,x) is the kernel function, and b is the bias; the expression of the kernel function K(xi,x) is:

[0037] K(xi,x)=exp(-γxi-x2);

[0038] γ represents the kernel parameter. The larger the value of γ, the narrower the curve of the kernel function, which means that in the feature space, only sample points that are very close to each other are considered similar. On the contrary, the smaller the value of γ, the wider the curve of the kernel function, and sample points farther away will also affect the prediction.

[0039] The training data is divided into a training set and a validation set according to the ratio. For each set of parameter combinations, the model is trained on the training set, and then the mean square error is calculated on the validation set. The parameter combination with the smallest mean square error on the validation set is selected as the final model hyperparameter;

[0040] S22. Model training and verification: Use the training set data to train the model, where the training set data contains the selected meteorological characteristics and the corresponding wind power. The training goal is to minimize the structural risk, which is expressed as:

[0041]

[0042] The constraints are: yi(ω·φ(xi)+b)≥1-ξi(ξi≥0), where ω is the weight vector, b is the bias, ξi is the slack variable, yi is the real wind power, φ(xi) is the function that maps the input samples to a high-dimensional space, and C is the penalty parameter;

[0043] S23. Use the validation set data to evaluate the model performance, divide the training data into training set and validation set according to the ratio, and calculate the mean absolute error (MAE) and the coefficient of determination;

[0044] The mean absolute error formula is:

[0045] The formula for the coefficient of determination is:

[0046] Among them, yi,v is the true power of the i-th sample in the validation set, is the power of the prediction for the i-th sample in the validation set; is the average value of the validation set power;

[0047] S24. Model evaluation index analysis. Analyze the calculated evaluation index. If the mean square error is large, it means that the average square error between the model prediction value and the true value is large. It may be due to overfitting or underfitting of the model, or unreasonable feature selection. At this time, you can further check the data distribution, feature correlation, etc. If the MAE is large, it means that the average absolute error of the model prediction is large, and the accuracy of the model needs to be improved. The determination coefficient R 2 The closer it is to 1, the better the model fits the data;

[0048] S25. Model optimization strategy. According to the analysis results of the evaluation indicators, take corresponding optimization strategies. If the model is found to be overfitting, it can be solved by increasing the penalty parameter C and reducing the number of features. If it is underfitting, consider increasing the amount of training data, reducing the regularization strength, adjusting the kernel parameter γ, or increasing the feature dimension.

[0049] The S3 includes:

[0050] S31. Low temperature cold wave monitoring and judgment: set comprehensive low temperature cold wave judgment conditions, which are not limited to the temperature threshold and the temperature change rate threshold, but also consider the changes in meteorological factors such as air pressure change characteristics and humidity change characteristics. When the average temperature is lower than the temperature threshold and the average temperature change rate is lower than the temperature change rate threshold for several consecutive time intervals, and the average pressure change exceeds a certain threshold and the average humidity change is lower than a certain threshold, it is judged that the low temperature cold wave weather state has entered;

[0051] S32. Model parameter adjustment strategy: when it is determined to be low temperature cold wave weather, the adaptive learning rate adjustment method is used to update the model weight vector ω and bias b. The update formula is:

[0052]

[0053]

[0054]

[0055] η is the adaptive learning rate, α is the learning rate attenuation factor, L is the loss function, ωo is the weight vector of the model before parameter adjustment at the current moment, which contains the weight values ​​related to each meteorological feature; ωn is the weight vector updated by the model after parameter adjustment; bo is the bias value of the model before parameter adjustment at the current moment; bn is the bias value updated by the model after parameter adjustment;

[0056] Among them, the adaptive learning rate η is adjusted according to the prediction error of the current model under low temperature cold wave weather. The prediction error formula of the current model under low temperature cold wave weather is expressed as:

[0057]

[0058] e is the prediction error, which reflects the degree of deviation between the model prediction result and the actual situation. The smaller the error, the higher the prediction accuracy of the model under low temperature and cold wave weather; q is the number of samples under the current low temperature and cold wave weather, yi,c is the true power of the i-th sample, is the prediction power of the model for the i-th sample; in the low temperature cold wave adaptive adjustment of the model, the prediction error e is the basis for adjusting the adaptive learning rate η. If e is large, it means that the current prediction effect of the model is poor, and the parameter adjustment speed needs to be accelerated, that is, η is increased; if e is small, it means that the model prediction effect is good, and the adjustment range should be reduced, that is, η is reduced, so that the model can more stably adapt to the data characteristics under low temperature cold wave weather and improve the prediction performance;

[0059] S33. Evaluation of parameter adjustment effect. After each parameter adjustment, use new data samples under low temperature and cold wave weather to evaluate the model; calculate the adjusted MSE, MAE and other indicators, and compare them with the indicators before adjustment; if the adjusted indicators are improved, it means that the parameter adjustment strategy is effective; if the indicators do not improve or even get worse, it is necessary to readjust the initial value of the learning rate, the attenuation factor or the adjustment parameter update formula; at the same time, observe the fitting curve of the model's predicted power and actual power under low temperature and cold wave weather to see if it is closer to the actual situation, so as to intuitively evaluate the effect of parameter adjustment;

[0060] S34. Continue to optimize the adjustment strategy. According to the results of the parameter adjustment effect evaluation, continue to optimize the adjustment strategy. If a certain parameter adjustment method is found to bring good results in multiple tests, fix it. If it is found that the adjustment strategy fails in certain cases, try to use different parameter adjustment ranges or methods according to the different intensities of low temperature cold waves.

[0061] The S4 includes:

[0062] S41. Real-time data collection, real-time acquisition of the latest meteorological data and wind power output data around the wind farm;

[0063] S42. Model incremental update mechanism, adding new preprocessed data samples to the training data set to form a new training data set; using the new training data set to incrementally train the model, adjusting the coefficients of the corresponding support vectors and other parameters according to the impact of the new samples on the model decision boundary; after each incremental update, recording the updated parameters and the corresponding timestamps, so as to facilitate subsequent analysis of the changing trend of the model parameters and their impact on the prediction performance;

[0064] S43. Model performance evaluation, monitoring and feedback adjustment are deepened. After each model update, the model performance is evaluated using the latest validation set data; indicators such as mean square error and mean absolute error are calculated, and these indicators are compared with the previous model performance. If the model performance drops significantly, an alarm is triggered and the reasons are further analyzed, including abnormal values ​​in the newly collected data, data collection equipment failure, improper model parameter update strategy, etc.;

[0065] According to the analysis results, take corresponding adjustment measures, such as re-collecting abnormal data, repairing equipment, adjusting the learning rate, or re-selecting the incremental learning algorithm.

[0066] In summary, due to the adoption of the above technical solution, the beneficial technical effects of the invention are:

[0067] By comprehensively considering multiple meteorological factors and accurately screening key features, combining advanced SVM models with RBF kernels to build meteorological and power mapping models, the accuracy of wind power forecasting has been greatly improved, effectively reducing forecast errors, and providing more reliable power expectations for wind power operations. At the same time, the unique low-temperature cold wave adaptive adjustment mechanism combines multiple meteorological factors to determine the cold wave state and adaptively update model parameters, enhancing the model's adaptability in extreme low-temperature cold wave weather, ensuring that it can still maintain high forecast performance under severe weather conditions and ensure the stable operation of the wind power system.

[0068] The real-time data-driven model update mechanism realizes the real-time synchronization of the model and the dynamic environment, and continuously optimizes the model to adapt to meteorological changes. Combined with a rigorous performance evaluation and monitoring system, it can promptly discover and solve the problem of model performance degradation, ensure the timeliness and accuracy of the prediction, provide accurate decision-making basis for power grid dispatching, optimize the allocation of power resources, enhance the stability of wind power grid connection, and reduce the impact of wind power fluctuations on the power grid.

[0069] In addition, the closed-loop optimization system feeds back the prediction results to the data link, strengthens the quality control of key feature data, and conducts in-depth analysis of the model through long-term performance tracking, promotes continuous improvement of the model, and continuously improves performance and reliability. It not only extends the service life of wind turbine equipment and reduces operating costs, but also promotes the sustainable and efficient development of the wind power industry as a whole, showing excellent technical advantages and practical value in the complex environment of new energy power generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a logic block diagram of a method for short-term wind power prediction under low temperature and cold wave weather;

[0071] Figure 2 Flowchart of the method for feature extraction and selection;

[0072] Figure 3 A flow chart of the method for establishing a meteorological and power mapping model;

[0073] Figure 4 A flow chart of a method for dynamically adjusting parameters of a meteorological and power mapping model according to changes in real-time meteorological data;

[0074] Figure 5 Flowchart of a method for establishing a real-time data-driven model update mechanism based on data collected by S1 and a meteorological and power mapping model established by S2. DETAILED DESCRIPTION

[0075] In order to make the purpose, technical solution and advantages of the invention more clear, the invention is further described in detail below in combination with the embodiments. It should be understood that the specific embodiments described here are only used to explain the invention and are not used to limit the invention.

[0076] like Figure 1 As shown, the method for short-term wind power prediction under low temperature cold wave weather includes the following steps:

[0077] S1. Feature extraction and selection: collect meteorological data around the wind farm, including temperature, wind speed, wind direction, etc., and pay special attention to the characteristics of low temperature cold wave weather, such as sudden temperature drop and continuous low temperature. Extract key features related to wind power output from meteorological data (temperature change rate, wind speed change trend, etc.), and select the features that have the greatest impact on wind power output.

[0078] S2. Establish a meteorological and power mapping model. Establish a meteorological and power mapping model based on the features extracted in S1. The meteorological and power mapping model can map the conditions of meteorological characteristics to wind power output, thereby converting meteorological characteristics into prediction results of wind power output;

[0079] S3. Adaptive adjustment of the model to low temperature and cold waves. Based on the meteorological and power mapping model established in S2, the parameters of the meteorological and power mapping model are dynamically adjusted according to the changes in real-time meteorological data, especially the parameters during low temperature and cold wave weather.

[0080] S4. Real-time data-driven model update. A real-time data-driven model update mechanism is established based on the data collected by S1 and the meteorological and power mapping model established by S2, so that the model can dynamically adjust the prediction according to the latest meteorological data and power output. The real-time update mechanism ensures that the model can quickly respond to environmental changes and maintain the accuracy of the prediction.

[0081] like Figure 2 As shown, the S1 includes:

[0082] S11. Data collection: obtaining meteorological data, such as temperature, wind speed, wind direction, air pressure and humidity, through a meteorological monitoring station, and recording wind power output data corresponding to the meteorological data;

[0083] S12. Feature calculation:

[0084] Calculate the temperature change rate. Use the sliding window method to calculate the temperature change rate. The formula is:

[0085]

[0086] Where ΔTt is the temperature change rate at the current time t; Tt is the temperature at the current time t; is the temperature Δt1 minutes ago; at the same time, the sliding average of the temperature change rate in the past n1 time intervals is calculated

[0087]

[0088] Calculate the wind speed trend and the first-order difference of wind speed (Unit: m / s / min), Vt is the current wind speed, is the wind speed value at the time interval Δt1 before the current time t;

[0089] And calculate the sliding average of the wind speed change trend in the past n1 time intervals

[0090]

[0091] Wind direction feature processing: convert wind direction data into vector components in the Cartesian coordinate system, i.e. x = cos(θ) and y = sin(θ), and calculate the change in wind direction vector and Among them, xt represents the value of the wind direction vector in the x direction at time t, represents the value of the wind direction vector in the x direction at the moment t-Δt1, Δx represents the change of the wind direction vector in the x direction at the moment t relative to the previous moment t-Δt1; represents the value of the wind direction vector in the y direction at the moment t, represents the value of the wind direction vector in the x direction at the moment t-Δt1, and Δy represents the change of the wind direction vector in the x direction at the moment t relative to the previous moment t-Δt1;

[0092] Pressure variation characteristics, calculate the first-order difference of pressure:

[0093]

[0094] Among them, ΔPa is the change in air pressure, Pa,t is the air pressure at the current time t, is the air pressure at time t-Δt1;

[0095] Humidity change characteristics, calculate the first-order difference of humidity:

[0096]

[0097] Among them, ΔH is the humidity change, Ht is the humidity at the current time t, is the humidity at time t-Δt1;

[0098] S13. Preliminary screening of features, calculation of the simple linear correlation coefficient r between each feature and wind power output, the formula is:

[0099]

[0100] Pt is the wind power at time t, is the average power value, Ft is the meteorological characteristic value (such as temperature change rate, wind speed change trend, etc.), is the feature average value, m is the number of samples; the features whose absolute value of the correlation coefficient r is greater than the threshold are selected as the screening results. The threshold is a suitable value determined through multiple tests and analyses to ensure that the screened features can effectively reflect the changes in wind power without making the model too complicated.

[0101] like Figure 3 As shown, the method for establishing a meteorological and power mapping model includes:

[0102] S21. Model selection and architecture design, the support vector machine (SVM) model is selected to establish the relationship between meteorology and power mapping, and the radial basis function (RBF) kernel is used; the expression of the model is:

[0103]

[0104] Among them, xi is the input meteorological feature vector, xi is a vector formed by combining multiple related types of feature values ​​Ft, xi contains multiple meteorological feature information meaningful to wind power prediction in vector form; x is another sample vector, which has the same structure as xi and also contains multiple meteorological feature values, which is used to represent the meteorological state at another time or condition; yi is the wind power output corresponding to the meteorological feature vector xi, sgn is the sign function, αi is the Lagrange multiplier, K(xi,x) is the kernel function, and b is the bias; the expression of the kernel function K(xi,x) is:

[0105] K(xi,x)=exp(-γxi-x2);

[0106] γ represents the kernel parameter. The larger the value of γ, the narrower the curve of the kernel function, which means that in the feature space, only sample points that are very close to each other are considered similar. On the contrary, the smaller the value of γ, the wider the curve of the kernel function, and sample points farther away will also affect the prediction.

[0107] The training data is divided into a training set and a validation set according to the ratio. For each set of parameter combinations, the model is trained on the training set, and then the mean square error is calculated on the validation set. The parameter combination with the smallest mean square error on the validation set is selected as the final model hyperparameter;

[0108] S22. Model training and verification: Use the training set data to train the model, where the training set data contains the selected meteorological characteristics and the corresponding wind power. The training goal is to minimize the structural risk, which is expressed as:

[0109]

[0110] The constraints are: yi(ω·φ(xi)+b)≥1-ξi(ξi≥0), where ω is the weight vector, b is the bias, ξi is the slack variable, yi is the real wind power, φ(xi) is the function that maps the input samples to a high-dimensional space, and C is the penalty parameter;

[0111] S23. Use the validation set data to evaluate the model performance, divide the training data into training set and validation set according to the ratio, and calculate the mean absolute error (MAE) and the coefficient of determination;

[0112] The mean absolute error formula is:

[0113] The formula for the coefficient of determination is:

[0114] Among them, yi,v is the true power of the i-th sample in the validation set, is the power of the prediction for the i-th sample in the validation set; is the average value of the validation set power;

[0115] S24. Model evaluation index analysis. Analyze the calculated evaluation index. If the mean square error is large, it means that the average square error between the model prediction value and the true value is large. It may be due to overfitting or underfitting of the model, or unreasonable feature selection. At this time, you can further check the data distribution, feature correlation, etc. If the MAE is large, it means that the average absolute error of the model prediction is large, and the accuracy of the model needs to be improved. The determination coefficient R 2 The closer it is to 1, the better the model fits the data. 2 =0.8, it means that the model can explain 80% of the wind power changes, and the remaining 20% ​​of the changes cannot be explained by the model, which requires further exploration of other influencing factors or improvement of the model.

[0116] S25. Model optimization strategy. According to the analysis results of the evaluation indicators, take the corresponding optimization strategy. If the model is found to be overfitting, it can be solved by increasing the regularization intensity (such as increasing the penalty parameter C), reducing the number of features, or using more effective feature selection methods. If it is underfitting, consider increasing the amount of training data, reducing the regularization intensity, increasing the complexity of the model (such as adjusting the kernel parameter γ or increasing the feature dimension), etc. At the same time, you can also try different kernel functions or model structures, compare their performance on the wind power prediction problem, and select the most suitable model configuration. For example, in addition to the RBF kernel function, you can also try the linear kernel function to observe whether the model performance is improved.

[0117] like Figure 4 As shown, the S3 includes:

[0118] S31. Low temperature and cold wave monitoring and judgment: set comprehensive low temperature and cold wave judgment conditions, which are not limited to the temperature threshold and the temperature change rate threshold, but also consider the changes in meteorological factors such as air pressure change characteristics and humidity change characteristics. When the average temperature is lower than the temperature threshold and the average temperature change rate is lower than the temperature change rate threshold for m consecutive time intervals, and the average air pressure change exceeds a certain threshold (indicating that the air pressure change is abnormally drastic and may be accompanied by cold wave weather) and the average humidity change is lower than a certain threshold (dry air conditions are also often related to cold waves), it is judged that the low temperature and cold wave weather state has entered;

[0119] S32. Model parameter adjustment strategy: when it is determined to be low temperature cold wave weather, the adaptive learning rate adjustment method is used to update the model weight vector ω and bias b. The update formula is:

[0120]

[0121]

[0122]

[0123] η is the adaptive learning rate, α is the learning rate attenuation factor, L is the loss function, ωo is the weight vector of the model before parameter adjustment at the current moment, which contains the weight values ​​related to each meteorological feature; ωn is the weight vector updated by the model after parameter adjustment, which replaces ωo for subsequent prediction calculations, reflecting the model's re-evaluation and adjustment of the importance of features based on new information (such as data performance under low temperature and cold wave weather); bo is the bias value of the model before parameter adjustment at the current moment, and the bias plays the role of shifting the decision boundary in the model. Bo and the weight vector jointly determine the position of the decision boundary of the model, thereby affecting the model's prediction of wind power; bn is the bias value updated by the model after parameter adjustment, and bn will replace bo for subsequent prediction calculations, reflecting the model's readjustment of the decision boundary position based on new information (such as data performance under low temperature and cold wave weather);

[0124] Among them, the adaptive learning rate η is adjusted according to the prediction error of the current model under low temperature cold wave weather. The prediction error formula of the current model under low temperature cold wave weather is expressed as:

[0125]

[0126] e is the prediction error, which reflects the degree of deviation between the model prediction result and the actual situation. The smaller the error, the higher the prediction accuracy of the model under low temperature and cold wave weather; q is the number of samples under the current low temperature and cold wave weather, yi,c is the true power of the i-th sample, is the prediction power of the model for the i-th sample; in the low temperature cold wave adaptive adjustment of the model, the prediction error e is the basis for adjusting the adaptive learning rate η. If e is large, it means that the current prediction effect of the model is poor, and the parameter adjustment speed needs to be accelerated, that is, η is increased; if e is small, it means that the model prediction effect is good, and the adjustment range should be reduced, that is, η is reduced, so that the model can more stably adapt to the data characteristics under low temperature cold wave weather and improve the prediction performance;

[0127] S33. Evaluation of parameter adjustment effect. After each parameter adjustment, use new data samples under low temperature and cold wave weather to evaluate the model; calculate the adjusted MSE, MAE and other indicators, and compare them with the indicators before adjustment; if the adjusted indicators are improved, it means that the parameter adjustment strategy is effective; if the indicators do not improve or even get worse, it may be necessary to re-examine the adjustment strategy, such as adjusting the initial value of the learning rate, the attenuation factor, or the formula for adjusting the parameter update; at the same time, observe the fitting curve of the model's predicted power and actual power under low temperature and cold wave weather to see if it is closer to the actual situation, so as to intuitively evaluate the effect of parameter adjustment;

[0128] S34. Continuously optimize the adjustment strategy. According to the results of the parameter adjustment effect evaluation, continuously optimize the adjustment strategy. If a certain parameter adjustment method is found to bring good results in multiple tests, fix it; if it is found that the adjustment strategy fails in certain cases, try new methods, such as using different parameter adjustment ranges or methods according to different low temperature cold wave intensities (such as lower temperature, greater temperature change rate, etc.); in addition, other meteorological factors (such as changes in air pressure, humidity, etc. in low temperature cold wave weather) can be combined to further optimize the parameter adjustment strategy to improve the prediction accuracy and stability of the model in low temperature cold wave weather.

[0129] like Figure 5 As shown, the S4 includes:

[0130] S41. Real-time data collection, real-time acquisition of the latest meteorological data (temperature, wind speed, wind direction, air pressure, humidity, etc.) and wind power output data around the wind farm;

[0131] S42. Model incremental update mechanism, adding new preprocessed data samples to the training data set to form a new training data set; using the new training data set to incrementally train the model, adjusting the coefficients of the corresponding support vectors and other parameters according to the impact of the new samples on the model decision boundary; after each incremental update, recording the updated parameters and the corresponding timestamps, so as to facilitate subsequent analysis of the changing trend of the model parameters and their impact on the prediction performance;

[0132] S43. Model performance evaluation, monitoring and feedback adjustment are deepened. After each model update, the model performance is evaluated using the latest validation set data; indicators such as mean square error and mean absolute error (MAE) are calculated, and these indicators are compared with the previous model performance. If the model performance drops significantly (for example, the MSE increases by more than a certain threshold, which can be set according to the historical performance of the model and actual needs), an alarm is triggered and the cause is further analyzed; the causes include the presence of outliers in the newly collected data (judged by data inspection and outlier detection methods), data collection equipment failure (check equipment status and data stability), improper model parameter update strategy (such as unreasonable learning rate setting), etc.

[0133] According to the analysis results, take corresponding adjustment measures, such as re-collecting abnormal data, repairing equipment, adjusting the learning rate or re-selecting the incremental learning algorithm, etc. At the same time, continuously monitor the prediction performance of the model under different weather conditions (especially low temperature and cold wave weather), analyze the distribution and trend of prediction errors, for example, by drawing error histograms, analyzing the error change curve over time, etc., determine whether it is necessary to further optimize the model's parameter adjustment strategy under specific weather conditions or add targeted features, etc., so as to continuously improve the accuracy and adaptability of the model; in addition, feed back the model's prediction results to the data collection and preprocessing links. For example, if it is found that certain features play a key role in model prediction but the data quality is unstable, the data collection and quality control of these features can be strengthened, so as to form a closed-loop optimization system and continuously improve the performance and reliability of the model; in addition, establish a long-term tracking record of model performance, including evaluation indicators and parameter changes in different time periods, so as to conduct more in-depth model performance analysis and optimization strategy formulation. For example, by analyzing historical data, it is found that the model is prone to performance fluctuations in certain seasons or under specific meteorological conditions, and targeted measures are taken in advance for prevention and optimization.

[0134] The above description is a preferred embodiment of the invention and is not intended to limit the invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the invention should be included in the protection scope of the invention.

Claims

1. A method for predicting short-term wind power in low-temperature cold wave weather, characterized in that: The following steps are involved: S1. Feature extraction and selection: collect meteorological data around the wind farm, including temperature, wind speed, wind direction, etc., and pay special attention to the characteristics of low temperature cold wave weather, such as sudden temperature drop and continuous low temperature. Extract key features related to wind power output from meteorological data and select the features that have the greatest impact on wind power output. S2. Establish a meteorological and power mapping model. Establish a meteorological and power mapping model based on the features extracted in S1. The meteorological and power mapping model can map the conditions of meteorological characteristics to wind power output, thereby converting meteorological characteristics into prediction results of wind power output; S3. Adaptive adjustment of the model to low temperature and cold waves. Based on the meteorological and power mapping model established in S2, the parameters of the meteorological and power mapping model are dynamically adjusted according to the changes in real-time meteorological data, especially the parameters during low temperature and cold wave weather. S4. Real-time data-driven model update. A real-time data-driven model update mechanism is established based on the data collected by S1 and the meteorological and power mapping model established by S2, so that the model can dynamically adjust the prediction according to the latest meteorological data and power output. The real-time update mechanism ensures that the model can quickly respond to environmental changes and maintain the accuracy of the prediction.

2. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 1 is characterized in that: The S1 includes: S11. Data collection: obtaining meteorological data, including temperature, wind speed, wind direction, air pressure and humidity information, through a meteorological monitoring station, and recording wind power output data corresponding to the meteorological data; S12. Feature calculation: Calculate the temperature change rate. Use the sliding window method to calculate the temperature change rate. The formula is: Where ΔTt is the temperature change rate at the current time t; Tt is the temperature at the current time t; is the temperature Δt1 minutes ago; at the same time, the sliding average of the temperature change rate in the past n1 time intervals is calculated Calculate the wind speed trend and the first-order difference of wind speed Vt is the current wind speed, is the wind speed value at the time interval Δt1 before the current time t; And calculate the sliding average of the wind speed change trend in the past n1 time intervals Wind direction feature processing: convert wind direction data into vector components in the Cartesian coordinate system, i.e. x = cos(θ) and y = sin(θ), and calculate the change in wind direction vector and Among them, xt represents the value of the wind direction vector in the x direction at time t, represents the value of the wind direction vector in the x direction at time t-Δt1, Δx represents the change of the wind direction vector in the x direction at time t relative to the previous time t-Δt1; represents the value of the wind direction vector in the y direction at time t, represents the value of the wind direction vector in the x direction at the moment t-Δt1, and Δy represents the change of the wind direction vector in the x direction at the moment t relative to the previous moment t-Δt1; Pressure variation characteristics, calculate the first-order difference of pressure: Among them, ΔPa is the change in air pressure, Pa,t is the air pressure at the current time t, is the air pressure at time t-Δt1; Humidity change characteristics, calculate the first-order difference of humidity: Among them, ΔH is the humidity change, Ht is the humidity at the current time t, is the humidity at time t-Δt1; S13. Preliminary screening of features, calculation of the simple linear correlation coefficient r between each feature and wind power output, the formula is: Pt is the wind power at time t, is the power average value, Ft is the meteorological characteristic value, is the feature average, m is the number of samples; the features whose absolute value of the correlation coefficient r is greater than the threshold are selected as the screening results.

3. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 1 is characterized in that: The establishment of the meteorological and power mapping model comprises: S21. Model selection and architecture design, the support vector machine (SVM) model is selected to establish the relationship between meteorology and power mapping, and the radial basis function (RBF) kernel is used; the expression of the model is: Among them, xi is the input meteorological feature vector, xi is a vector formed by combining multiple related types of feature values ​​Ft, xi contains multiple meteorological feature information meaningful to wind power prediction in vector form; x is another sample vector, which has the same structure as xi and also contains multiple meteorological feature values, which is used to represent the meteorological state at another time or condition; yi is the wind power output corresponding to the meteorological feature vector xi, sgn is the sign function, αi is the Lagrange multiplier, K(xi,x) is the kernel function, and b is the bias; the expression of the kernel function K(xi,x) is: K(xi,x)=exp(-γxi-x2); γ represents the kernel parameter. The larger the value of γ, the narrower the curve of the kernel function, which means that in the feature space, only sample points that are very close to each other are considered similar. On the contrary, the smaller the value of γ, the wider the curve of the kernel function, and sample points farther away will also affect the prediction. The training data is divided into a training set and a validation set according to the ratio. For each set of parameter combinations, the model is trained on the training set, and then the mean square error is calculated on the validation set. The parameter combination with the smallest mean square error on the validation set is selected as the final model hyperparameter; S22. Model training and verification: Use the training set data to train the model, where the training set data contains the selected meteorological characteristics and the corresponding wind power. The training goal is to minimize the structural risk, which is expressed as: S23. Use the validation set data to evaluate the model performance, divide the training data into training set and validation set according to the ratio, and calculate the mean absolute error (MAE) and the coefficient of determination; The mean absolute error formula is: The formula for the coefficient of determination is: Among them, yi,v is the true power of the i-th sample in the validation set, is the power of the prediction for the i-th sample in the validation set; is the average value of the validation set power; S24. Model evaluation index analysis. Analyze the calculated evaluation index. If the mean square error is large, it means that the average square error between the model prediction value and the true value is large. It may be due to overfitting or underfitting of the model, or unreasonable feature selection. At this time, you can further check the data distribution, feature correlation, etc. If the MAE is large, it means that the average absolute error of the model prediction is large, and the accuracy of the model needs to be improved. The determination coefficient R 2 The closer it is to 1, the better the model fits the data; S25. Model optimization strategy. According to the analysis results of the evaluation indicators, take corresponding optimization strategies. If the model is found to be overfitting, it can be solved by increasing the penalty parameter C and reducing the number of features. If it is underfitting, consider increasing the amount of training data, reducing the regularization strength, adjusting the kernel parameter γ, or increasing the feature dimension.

4. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 3 is characterized in that: The structural risk constraint to be minimized is: yi(ω·φ(xi)+b)≥1-ξi(ξi≥0), where ω is the weight vector, b is the bias, ξi is the slack variable, yi is the real wind power, φ(xi) is the function that maps the input samples to a high-dimensional space, and C is the penalty parameter.

5. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 1 is characterized in that: The S3 includes: S31. Low temperature cold wave monitoring and judgment: set comprehensive low temperature cold wave judgment conditions, which are not limited to the temperature threshold and the temperature change rate threshold, but also consider the changes in meteorological factors such as air pressure change characteristics and humidity change characteristics. When the average temperature is lower than the temperature threshold and the average temperature change rate is lower than the temperature change rate threshold for several consecutive time intervals, and the average pressure change exceeds a certain threshold and the average humidity change is lower than a certain threshold, it is judged that the low temperature cold wave weather state has entered; S32. Model parameter adjustment strategy: when it is determined to be low temperature cold wave weather, the adaptive learning rate adjustment method is used to update the model weight vector ω and bias b. The update formula is: η is the adaptive learning rate, α is the learning rate attenuation factor, L is the loss function, ωo is the weight vector of the model before parameter adjustment at the current moment, which contains the weight values ​​related to each meteorological feature; ωn is the weight vector updated by the model after parameter adjustment; bo is the bias value of the model before parameter adjustment at the current moment; bn is the bias value updated by the model after parameter adjustment; S33. Evaluation of parameter adjustment effect. After each parameter adjustment, use new data samples under low temperature and cold wave weather to evaluate the model; calculate the adjusted MSE, MAE and other indicators, and compare them with the indicators before adjustment; if the adjusted indicators are improved, it means that the parameter adjustment strategy is effective; if the indicators do not improve or even get worse, it is necessary to readjust the initial value of the learning rate, the attenuation factor or the adjustment parameter update formula; at the same time, observe the fitting curve of the model's predicted power and actual power under low temperature and cold wave weather to see if it is closer to the actual situation, so as to intuitively evaluate the effect of parameter adjustment; S34. Continue to optimize the adjustment strategy. According to the results of the parameter adjustment effect evaluation, continue to optimize the adjustment strategy. If a certain parameter adjustment method is found to bring good results in multiple tests, fix it. If it is found that the adjustment strategy fails in certain cases, try to use different parameter adjustment ranges or methods according to the different intensities of low temperature cold waves.

6. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 4 is characterized in that: The adaptive learning rate η is adjusted according to the prediction error of the current model under low temperature cold wave weather. The prediction error formula of the current model under low temperature cold wave weather is expressed as: e is the prediction error, which reflects the degree of deviation between the model prediction result and the actual situation. The smaller the error, the higher the prediction accuracy of the model under low temperature and cold wave weather; q is the number of samples under the current low temperature and cold wave weather, yi,c is the true power of the i-th sample, is the prediction power of the model for the i-th sample; in the low temperature cold wave adaptive adjustment of the model, the prediction error e is the basis for adjusting the adaptive learning rate η. If e is larger, it means that the current prediction effect of the model is poor, and the parameter adjustment speed needs to be accelerated, that is, η is increased; if e is smaller, it means that the model prediction effect is better, and the adjustment range should be reduced, that is, η is reduced, so that the model can more stably adapt to the data characteristics under low temperature cold wave weather and improve the prediction performance.

7. The method for predicting short-term wind power in low-temperature cold wave weather according to claim 1 is characterized in that: The S4 includes: S41. Real-time data collection, real-time acquisition of the latest meteorological data and wind power output data around the wind farm; S42. Model incremental update mechanism, adding new preprocessed data samples to the training data set to form a new training data set; using the new training data set to incrementally train the model, adjusting the coefficients of the corresponding support vectors and other parameters according to the impact of the new samples on the model decision boundary; after each incremental update, recording the updated parameters and the corresponding timestamps, so as to facilitate subsequent analysis of the changing trend of the model parameters and their impact on the prediction performance; S43. Model performance evaluation, monitoring and feedback adjustment are deepened. After each model update, the model performance is evaluated using the latest validation set data; indicators such as mean square error and mean absolute error are calculated, and these indicators are compared with the previous model performance. If the model performance drops significantly, an alarm is triggered and the reasons are further analyzed, including abnormal values ​​in the newly collected data, data collection equipment failure, improper model parameter update strategy, etc.; According to the analysis results, take corresponding adjustment measures, such as re-collecting abnormal data, repairing equipment, adjusting the learning rate, or re-selecting the incremental learning algorithm.

Images