A real-time error estimation method for wind power grid-connected power prediction
Patent Information
- Application Number
- CN202310901618.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-21
- Publication Date
- 2026-09-11
- Estimated Expiration
- 2043-07-21
AI Technical Summary
[0003]目前风电预测误差估计方法主要是对预测误差的概率密度特性曲线进行拟合，然而这种数理统计的方法具有一定的局限性和不确定性，主要表现在三个方面：1.单一分布函数只能对特定风电场预测误差的概率密度特性曲线进行拟合，不具有较好的普适性
[0037] (1) This invention uses the wind speed fluctuation process and rise and fall characteristics as constraints, and constructs a prediction error polarity judgment model based on the DTW algorithm to realize the judgment of prediction error polarity;
Smart Images
Figure CN116937682B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of new energy power generation technology, and in particular to a real-time error estimation method for wind power grid connection power prediction. Background Technology
[0002] With the continuous growth of global energy demand and the environmental impact of traditional fossil fuels, renewable energy has gradually become key to solving energy and environmental problems. Wind energy, as an important renewable energy source, has enormous potential. However, the randomness and volatility of wind mean that wind power is highly uncontrollable, and wind power forecasting still has significant errors. Without sufficient reserves to compensate for these forecasting errors, severe wind load reduction (WC) or even load shedding (LS) problems may occur. Therefore, accurate estimation of wind power forecasting errors is of paramount importance.
[0003] Current methods for estimating wind power forecasting errors mainly involve fitting the probability density curve of the forecasting error. However, this mathematical and statistical approach has certain limitations and uncertainties, primarily in three aspects: 1. A single distribution function can only fit the probability density curve of the forecasting error for a specific wind farm, lacking good universality. 2. The probability density characteristics of wind power forecasting errors are often dynamically changing, influenced by different factors, and may have different distribution characteristics at different times, making real-time estimation impossible. 3. Wind power forecasting error estimation based on probability density functions can only provide the distribution range of the error, but cannot obtain the polarity of the forecasting error. Summary of the Invention
[0004] To address the problems existing in the prior art, this invention proposes a real-time error estimation method for wind power grid connection power prediction.
[0005] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:
[0006] A real-time error estimation method for wind power grid-connected power prediction includes the following steps:
[0007] S1. Collect wind speed data for wind power generation over a year, normalize the wind speed data to a range of (0, 1), and divide the wind speed fluctuation process into small fluctuation process, medium fluctuation process, and large fluctuation process using the peak probability density distribution equal division method.
[0008]
[0009] Among them, v maxε0 represents the maximum peak value during a wind speed fluctuation process, ε1 and ε2 represent the judgment thresholds for the wind speed fluctuation process, and W0, W1, W2, and W3 represent the zero-output wind speed process, small fluctuation process, medium fluctuation process, and large fluctuation process, respectively.
[0010] S2. Based on the wind speed difference between time i and time i+1, the characteristics of wind speed rise and fall are derived:
[0011]
[0012] Where m is the wind speed change threshold, and m > 0; when the wind is stable at time i+1, it means that the wind speed has not changed significantly compared with the previous time, and thus the wind speed rise and fall characteristics are consistent with those of the previous time.
[0013] S3. Based on S1 and S2, determine the wind speed fluctuation process and rise / fall characteristics for day k and the nearly r days preceding day k, and statistically analyze the wind power prediction error polarity at the corresponding times; using the constraint that the sampling times are in the same fluctuation process and have the same rise / fall characteristics, divide the time series of day k and the nearly r days preceding day k into several subsequences {K1, K2, ... K}. n},{R1,R2,…R m Each subsequence consists of several time points, K n ={k1,k2,…k s}, R m ={r1,r2,…r t}; Based on the DTW algorithm for K n ,R m Similarity is calculated, and the historical time subsequences R with the smallest cumulative distance D(s,t) are selected based on their cumulative distance D(s,t). m The polarity of the prediction error is used as the time subsequence K of day k. n The polarity of the prediction error;
[0014] D(s,t)=d(s,t)+min{D(s-1,t-1),D(s-1,t),D(s,t-1)} (3);
[0015] Where s is the wind speed at time i in the time subsequence of day k, and t is the wind speed at time j in a time subsequence of the nearest r days before day k;
[0016] S4. Calculate the actual wind power output and the predicted output for the nearest r days before day k. The prediction error amplitude at each time point is:
[0017]
[0018] in, These are the actual wind power output and the day-ahead predicted wind power output at the i-th moment, respectively.
[0019] Prediction accuracy is represented by the weighted average absolute error of the N time steps preceding time i on day j.
[0020]
[0021] Where θ is the inertia factor, ranging from (0, 1), and exists between time i and time i-1;
[0022] Theoretical prediction error magnitude at time i on day j
[0023]
[0024] Among them, v i Let v be the wind speed at time i on day j. i-1 Let i be the wind speed at time i-1. To contribute to the prediction at time i, This represents the actual output at time i-1;
[0025] S5. With a sampling time interval of 15 minutes, calculate the prediction accuracy for N = 2 to 96, and then calculate the Pearson correlation coefficient R between the prediction accuracy and the prediction error amplitude using 96 sample points per day. j Select the number of samples N that results in the highest correlation coefficient. opt As the optimal value for each calculation;
[0026]
[0027] Among them, E i and Let be the prediction accuracy and its average value at time i, respectively. and Let $i$ be the magnitude of the prediction error at time $i$ and $i$ is the average value of the error.
[0028] S6. Calculate the theoretical prediction error magnitude at each time point. Based on the optimal number of samples N opt Prediction accuracy E i A model for estimating the magnitude of wind power prediction errors was constructed using multiple linear regression.
[0029]
[0030] Among them, E i Let be the prediction accuracy at time i. Let b0, b1, and b2 be the theoretical prediction error magnitude at time i, and b0, b1, and b2 be the regression coefficients. b0, b1, and b2 are solved using the least squares method.
[0031] As a further preferred technical solution of the present invention, when the wind speed increases from low to high and becomes an upwind, due to the existence of inertia, the actual rotational speed increases slowly, and the actual power generation increases slowly. Therefore, the power corresponding to the wind speed during this period is lower than the standard wind power curve, and the prediction error shows a negative deviation trend. When the wind speed decreases from high to low and becomes a downwind, the prediction error shows a positive deviation trend. At the same time, in order to improve the similarity of the time series, the peak probability density distribution equal division method is used to divide the wind speed fluctuation process into small fluctuation, medium fluctuation, and large fluctuation processes. Using the wind speed fluctuation process and rise and fall characteristics as constraints, based on the DTW algorithm, the historical time subsequence most similar to each time subsequence of the k-th day is found as the basis for judging the polarity of the prediction error.
[0032] Furthermore, as a preferred embodiment of the present invention, the predicted power at time i on day j differs significantly from the predicted power at the previous N times due to wind speed variations, resulting in a large difference in the magnitude of the prediction error; through calculation... The value of is used to measure the similarity between the predicted power at time i and time it. The larger the value, the more similar the wind power output at the two times, and the more similar the prediction error amplitude. Therefore, As the weight of the prediction error amplitude at time i-1; at the same time, considering that the inertia of the wind turbine is large and the output of the wind turbine is similar at adjacent times, an inertia factor θ is added to the weight of the prediction error amplitude at time i-1, and the specific calculation formula is shown in equation (5).
[0033] Furthermore, as a preferred embodiment of the present invention, the model of the wind turbine output power is shown in equation (9):
[0034]
[0035] in, The actual output power of the wind turbine is given by ρ, air density is given by A, and the swept area of the wind turbine is given by C. p v is the wind turbine power coefficient. i Wind speed; actual output power at time i-1 Wind speed v i-1 Wind speed v at time i i The theoretical actual output power at time i can be obtained as follows: Therefore, the theoretical prediction error magnitude at time i on day j is shown in equation (6).
[0036] The real-time error estimation method for wind power grid-connected power prediction described in this invention, compared with the prior art, has the following technical advantages:
[0037] (1) This invention uses the wind speed fluctuation process and rise and fall characteristics as constraints, and constructs a prediction error polarity judgment model based on the DTW algorithm to realize the judgment of prediction error polarity;
[0038] (2) This invention proposes two features that are highly correlated with the magnitude of the prediction error: prediction accuracy E i theoretical prediction error amplitude It improved the accuracy of the prediction error magnitude estimation;
[0039] (3) This invention only requires analysis and calculation of historical data of wind turbine units. The data is easy to obtain and the amount of calculation is small. At the same time, it also takes into account the recent performance of wind power output, and has good universality and real-time performance.
[0040] (4) The estimability of wind power prediction error can improve the reliability of wind power grid connection forecasts and at the same time improve the effectiveness of energy storage system capacity utilization for adjusting the accuracy of wind power grid connection data. Attached Figure Description
[0041] Figure 1 This is a flowchart illustrating the present invention;
[0042] Figure 2 This is a schematic diagram of the wind speed fluctuation process in this invention;
[0043] Figure 3 This is a wind speed similarity comparison chart based on the DTW algorithm in this invention;
[0044] Figure 4 The optimal number of samples N in this invention is... opt At that time, a statistical chart showing the correlation between prediction accuracy and prediction error magnitude;
[0045] Figure 5 This is a comparison chart of the estimated prediction error and the actual error in this invention;
[0046] Figure 6 This is a comparison chart of the power plan deviations before and after the optimization of the grid-connected power reporting plan in this invention;
[0047] Figure 7 This is a schematic diagram of the change in energy storage capacity for deviation compensation before the optimization of the grid-connected power reporting plan in this invention;
[0048] Figure 8 This is a schematic diagram of the change in energy storage capacity for deviation compensation after the optimization of the grid-connected power reporting plan in this invention. Detailed Implementation
[0049] The present invention will be further explained in detail below with reference to the accompanying drawings, so that those skilled in the art can better understand and implement the present invention. However, the following examples are only used to explain the present invention and are not intended to limit the present invention.
[0050] like Figure 1As shown, a real-time error estimation method for wind power grid-connected power prediction includes the following steps:
[0051] S1. Collect wind speed data for wind power generation over a year, normalize the wind speed data to a range of (0, 1), and divide the wind speed fluctuation process into small fluctuation process, medium fluctuation process, and large fluctuation process using the peak probability density distribution equal division method.
[0052]
[0053] Among them, v max ε0 represents the maximum peak value during a wind speed fluctuation process, ε1 and ε2 represent the judgment thresholds for the wind speed fluctuation process, and W0, W1, W2, and W3 represent the zero-output wind speed process, small fluctuation process, medium fluctuation process, and large fluctuation process, respectively.
[0054] S2. Based on the wind speed difference between time i and time i+1, the characteristics of wind speed rise and fall are derived:
[0055]
[0056] Where m is the wind speed change threshold, and m > 0; when the wind is stable at time i+1, it means that the wind speed has not changed significantly compared with the previous time, and thus the wind speed rise and fall characteristics are consistent with those of the previous time.
[0057] S3. Based on S1 and S2, determine the wind speed fluctuation process and rise / fall characteristics for day k and the nearly r days preceding day k, and statistically analyze the wind power prediction error polarity at the corresponding times; using the constraint that the sampling times are in the same fluctuation process and have the same rise / fall characteristics, divide the time series of day k and the nearly r days preceding day k into several subsequences {K1, K2, ... K}. n},{R1,R2,…R m Each subsequence consists of several time points, K n ={k1,k2,…k s}, R m ={r1,r2,…r t}; Based on the DTW algorithm for K n ,R m Similarity is calculated, and the historical time subsequences R with the smallest cumulative distance D(s,t) are selected based on their cumulative distance D(s,t). m The polarity of the prediction error is used as the time subsequence K of day k. n The polarity of the prediction error;
[0058] D(s,t)=d(s,t)+min{D(s-1,t-1),D(s-1,t),D(s,t-1)} (3);
[0059] Where s is the wind speed at time i in the time subsequence of day k, and t is the wind speed at time j in a time subsequence of the nearest r days before day k;
[0060] S4. Calculate the actual wind power output and the predicted output for the nearest r days before day k. The prediction error amplitude at each time point is:
[0061]
[0062] in, These are the actual wind power output and the day-ahead predicted wind power output at the i-th moment, respectively.
[0063] Prediction accuracy is represented by the weighted average absolute error of the N time steps preceding time i on day j.
[0064]
[0065] Where θ is the inertia factor, ranging from (0, 1), and exists between time i and time i-1;
[0066] Theoretical prediction error magnitude at time i on day j
[0067]
[0068] Among them, v i Let v be the wind speed at time i on day j. i-1 Let be the wind speed at time i-1. To contribute to the prediction at time i, This represents the actual output at time i-1;
[0069] S5. With a sampling time interval of 15 minutes, calculate the prediction accuracy for N = 2 to 96, and then calculate the Pearson correlation coefficient R between the prediction accuracy and the prediction error amplitude using 96 sample points per day. j Select the number of samples N that results in the highest correlation coefficient. opt As the optimal value for each calculation;
[0070]
[0071] Among them, E i and Let be the prediction accuracy and its average value at time i, respectively. and Let $i$ be the magnitude of the prediction error at time $i$ and $i$ is the average value of the error.
[0072] S6. Calculate the theoretical prediction error magnitude at each time point. Based on the optimal number of samples N opt Prediction accuracy Ei A model for estimating the magnitude of wind power prediction errors was constructed using multiple linear regression.
[0073]
[0074] Among them, E i Let be the prediction accuracy at time i. Let b0, b1, and b2 be the theoretical prediction error magnitude at time i, and b0, b1, and b2 be the regression coefficients. b0, b1, and b2 are solved using the least squares method.
[0075] When the wind speed increases from low to high and becomes an upwind, due to inertia, the actual rotational speed increases slowly, and the actual power generation increases slowly as a result. Therefore, the power corresponding to the wind speed during this period is lower than the standard wind power curve, and the prediction error shows a negative deviation trend. When the wind speed decreases from high to low and becomes a downwind, the prediction error shows a positive deviation trend. At the same time, in order to improve the similarity of the time series, the peak probability density distribution equal division method is used to divide the wind speed fluctuation process into small fluctuation, medium fluctuation, and large fluctuation processes. Using the wind speed fluctuation process and rise and fall characteristics as constraints, based on the DTW algorithm, the historical time subsequence most similar to each time subsequence of day k is found as the basis for judging the polarity of the prediction error.
[0076] The predicted power at time i on day j differs significantly from the predicted power at the previous N times due to changes in wind speed, resulting in a large difference in the magnitude of the prediction error. Through calculation... The value of is used to measure the similarity between the predicted power at time i and time it. The larger the value, the more similar the wind power output at the two times, and the more similar the prediction error amplitude. Therefore, As the weight of the prediction error amplitude at time i-1; at the same time, considering that the inertia of the wind turbine is large and the output of the wind turbine is similar at adjacent times, an inertia factor θ is added to the weight of the prediction error amplitude at time i-1, and the specific calculation formula is shown in equation (5).
[0077] The model for the output power of the wind turbine is shown in equation (9):
[0078]
[0079] Among them, P i a The actual output power of the wind turbine is given by ρ, air density is given by A, and the swept area of the wind turbine is given by C. p v is the wind turbine power coefficient. i Wind speed; actual output power at time i-1 Wind speed v i-1 Wind speed v at time i i The theoretical actual output power at time i can be obtained as follows: Therefore, the theoretical prediction error magnitude at time i on day j is shown in equation (6).
[0080] In practical implementation, the real-time error estimation method for wind power grid-connected power prediction according to the present invention includes the following steps:
[0081] Step 1: Collect wind speed data for one year of wind power generation, normalize the wind speed data to a range of (0, 1). Analyze the distribution of peak values and plot the probability density curve. Then, divide the data into equal parts to determine the initial threshold ε0, the threshold for small fluctuations ε1, and the threshold for large fluctuations ε2, which are 0.025, 0.181, and 0.522, respectively.
[0082] refer to Figure 2 The sampling time interval is 15 minutes. The wind speed fluctuation process diagram for 24 hours on October 1st is divided into medium fluctuation process and large fluctuation process based on the fluctuation threshold.
[0083] Step 2: Collect wind speed data for one year of wind power generation and the polarity of prediction error at the corresponding time. Calculate the wind speed difference Δv between time i and time i+1. Analyze the relationship between the change in prediction error polarity and Δv. It can be seen that the wind speed change threshold m = 0.33. Then, the rise and fall characteristics of wind speed at each time can be obtained.
[0084] Step 3: Combining Steps 1 and 2, determine the wind speed fluctuation process and rise / fall characteristics for October 1st and the preceding 273 days, and statistically analyze the polarity of wind power prediction errors at corresponding times. Using the constraint that the sampling times are within the same fluctuation process and have the same rise / fall characteristics, divide the time series for October 1st and the preceding 273 days into several subsequences {K1, K2, ... K... n},{R1,R2,…R m Each subsequence consists of several time points, K. n ={k1,k2,…k s},R m ={r1,r2,…r t Based on the DTW algorithm, K n ,R m Similarity is calculated, and the historical time subsequences are sorted by the cumulative distance D(s,t) between them. The subsequence R with the smallest cumulative distance, i.e., the one with the highest similarity, is selected. m The polarity of the prediction error as the K of the October 1 time series n The polarity of the prediction error;
[0085] refer to Figure 3 A subsequence K from October 1st n={15,16,17,18,19,20,21}, which is in a medium fluctuation process and all are downward winds. The left side is the original time subsequence, and the right side is based on the DTW algorithm to find the time subsequence with the smallest cumulative distance of 2.786 on June 18. Finally, the polarity of the prediction error is judged, and the accuracy rate is 86%.
[0086] Step 4: Collect data on actual wind power output and day-ahead predicted output for October 1st and the preceding 273 days, and calculate the theoretical prediction error amplitude at each time point. Then, calculate the prediction accuracy for N=2 to 96, and then calculate the Pearson correlation coefficient between prediction accuracy and prediction error amplitude using 96 sample points per day. Select the sample size N with the highest correlation coefficient. opt As the optimal value for each calculation, N is obtained through calculation. opt =2, with a maximum Pearson correlation coefficient of 0.722. The prediction accuracy at each time point is calculated based on the optimal number of samples.
[0087] refer to Figure 4 The correlation chart between prediction accuracy and prediction error magnitude obtained based on the optimal number of samples shows that the two have a strong correlation.
[0088] Step 5: Calculate the prediction error amplitude for each moment in the 273 days prior to October 1st. Then, based on the prediction accuracy and theoretical prediction error amplitude obtained in Step 4, construct a wind power prediction error amplitude estimation model based on multiple linear regression. Solve for the parameters using the least squares method, obtaining b0 = 11.645, b1 = 0.589, and b2 = 0.082. Therefore, the wind power prediction error amplitude estimation model is... Finally, the magnitude of the error prediction at each time point on October 1st is estimated in real time based on the wind power prediction error magnitude estimation model.
[0089] refer to Figure 5 The prediction error for each moment on October 1st achieved good estimation results in terms of both error polarity and error magnitude.
[0090] refer to Figure 6 This chart compares the power plan deviations before and after optimization of the grid-connected power reporting plan. The estimated power prediction error is added to the predicted value to obtain the planned wind power output, and the deviation from the actual power is calculated, with an average absolute error of 33.997 kW. If the predicted power is used as the reported plan, the average absolute error is 51.629 kW, effectively improving the accuracy of wind power grid-connected power reporting.
[0091] refer to Figures 7-8These figures illustrate the changes in energy storage capacity for deviation compensation before and after optimization of the grid-connected power reporting plan. Before optimization, the maximum cumulative discharge was 539.087 kWh, requiring 1797 kWh of energy storage. After optimization, the maximum cumulative discharge was -341.053 kWh, requiring 1137 kWh of energy storage, a reduction of 660 kWh compared to the former, effectively reducing the grid dispatch reserve capacity.
[0092] The specific implementation schemes described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific implementation schemes of the present invention and are not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention should fall within the scope of protection of the present invention.
Claims
1. A real-time error estimation method for wind power grid-connected power prediction, characterized in that, Includes the following steps: S1. Collect wind speed data for wind power generation over a year, normalize the wind speed data to a range of (0, 1), and divide the wind speed fluctuation process into small fluctuation process, medium fluctuation process, and large fluctuation process using the peak probability density distribution equal division method. Among them, v max ε0 represents the maximum peak value during a wind speed fluctuation process, ε1 and ε2 represent the judgment thresholds for the wind speed fluctuation process, and W0, W1, W2, and W3 represent the zero-output wind speed process, small fluctuation process, medium fluctuation process, and large fluctuation process, respectively. S2. Based on the wind speed difference between time i and time i+1, the characteristics of wind speed rise and fall are derived: Where m is the wind speed change threshold, and m > 0; when the wind is stable at time i+1, it means that the wind speed has not changed significantly compared with the previous time, and the wind speed rise and fall characteristics are consistent with those of the previous time. S3. Based on S1 and S2, determine the wind speed fluctuation process and rise / fall characteristics for day k and the nearly r days preceding day k, and statistically analyze the polarity of wind power prediction errors at corresponding times; using the constraint that the sampling times are in the same fluctuation process and have the same rise / fall characteristics, divide the time series of day k and the nearly r days preceding day k into several subsequences {K1, K2, ... K}. n },{R1,R2,…R m Each subsequence consists of several time points, K n ={k1,k2,…k s }, R m ={r1,r2,…r t }; Based on the DTW algorithm for K n ,R m Similarity is calculated, and the historical time subsequences R with the smallest cumulative distance D(s,t) are selected based on their cumulative distance D(s,t). m The polarity of the prediction error is used as the time subsequence K of day k. n The polarity of the prediction error; D(s,t)=d(s,t)+min{D(s-1,t-1),D(s-1,t),D(s,t-1)} (3); Where s is the wind speed at time i in the time subsequence of day k, and t is the wind speed at time j in a time subsequence of the nearest r days before day k; S4. Calculate the actual wind power output and the predicted output for the nearest r days before day k. The prediction error amplitude at each time point is: in, These are the actual wind power output and the day-ahead predicted wind power output at the i-th moment, respectively. Prediction accuracy is represented by the weighted average absolute error of the N time steps preceding time i on day j. Where θ is the inertia factor, ranging from (0, 1), and exists between time i and time i-1; Theoretical prediction error magnitude at time i on day j Among them, v i Let v be the wind speed at time i on day j. i-1 Let be the wind speed at time i-1. To contribute to the prediction at time i, This represents the actual output at time i-1; S5. With a sampling time interval of 15 minutes, calculate the prediction accuracy for N = 2 to 96, and then calculate the Pearson correlation coefficient R between the prediction accuracy and the prediction error amplitude using 96 sample points per day. j Select the number of samples N that results in the highest correlation coefficient. opt As the optimal value for each calculation; Among them, E i and Let be the prediction accuracy and its average value at time i, respectively. and Let $i$ be the magnitude of the prediction error at time $i$ and $i$ is the average value of the error. S6. Calculate the theoretical prediction error magnitude at each time point. Based on the optimal number of samples N opt Prediction accuracy E i A model for estimating the magnitude of wind power prediction errors was constructed using multiple linear regression. Among them, E i Let be the prediction accuracy at time i. Let b0, b1, and b2 be the theoretical prediction error magnitude at time i, and b0, b1, and b2 be the regression coefficients. b0, b1, and b2 are solved using the least squares method.
2. The real-time error estimation method for wind power grid-connected power prediction according to claim 1, characterized in that, When the wind speed increases from low to high and becomes an upwind, due to inertia, the actual rotational speed increases slowly, and the actual power generation increases slowly as a result. Therefore, the power corresponding to the wind speed during this period is lower than the standard wind power curve, and the prediction error shows a negative deviation trend. When the wind speed decreases from high to low and becomes a downwind, the prediction error shows a positive deviation trend. At the same time, in order to improve the similarity of the time series, the peak probability density distribution equal division method is used to divide the wind speed fluctuation process into small fluctuation, medium fluctuation, and large fluctuation processes. Using wind speed fluctuations and rise / fall characteristics as constraints, and based on the DTW algorithm, we find the historical time subsequences most similar to each time subsequence on day k, which serves as the basis for judging the polarity of prediction error.
3. The real-time error estimation method for wind power grid-connected power prediction according to claim 1, characterized in that, The predicted power at time i on day j differs significantly from the predicted power at the previous N times due to changes in wind speed, resulting in a large difference in the magnitude of the prediction error. By calculating 1- The value of is used to measure the similarity between the predicted power at time i and time it. The larger the value, the more similar the wind power output at the two times, and the more similar the prediction error amplitude. Therefore, 1- As the weight of the prediction error amplitude at time i-1; at the same time, considering that the inertia of the wind turbine is large and the output of the wind turbine is similar at adjacent times, an inertia factor θ is added to the weight of the prediction error amplitude at time i-1, and the specific calculation formula is shown in equation (5).
4. The real-time error estimation method for wind power grid-connected power prediction according to claim 1, characterized in that, The model for the output power of the wind turbine is shown in equation (9): Among them, P i a The actual output power of the wind turbine is given by ρ, air density is given by A, and the swept area of the wind turbine is given by C. p v is the wind turbine power coefficient. i Wind speed; actual output power at time i-1 Wind speed v i-1 Wind speed v at time i i The theoretical actual output power at time i can be obtained as follows: Therefore, the theoretical prediction error magnitude at time i on day j is shown in equation (6).